-
-
From the definition of the input variable to the activation function, that is \( z_j^l \) we have
-$$ -\frac{\partial z_j^l}{\partial w_{ij}^l} = a_i^{l-1}, -$$ - -and
-$$ -\frac{\partial z_j^l}{\partial a_i^{l-1}} = w_{ji}^l. -$$ - -With our definition of the activation function we have that (note that this function depends only on \( z_j^l \))
-$$ -\frac{\partial a_j^l}{\partial z_j^{l}} = a_j^l(1-a_j^l)=\sigma(z_j^l)(1-\sigma(z_j^l)). -$$ +
@@ -500,7 +497,7 @@ $$
-
With these definitions we can now compute the derivative of the cost function in terms of the weights.
- -Let us specialize to the output layer \( l=L \). Our cost function is
+From the definition of the input variable to the activation function, that is \( z_j^l \) we have
$$ -{\cal C}(\boldsymbol{\Theta}^L) = \frac{1}{2}\sum_{i=1}^n\left(y_i - \tilde{y}_i\right)^2=\frac{1}{2}\sum_{i=1}^n\left(a_i^L - y_i\right)^2, +\frac{\partial z_j^l}{\partial w_{ij}^l} = a_i^{l-1}, $$ -The derivative of this function with respect to the weights is
- +and
$$ -\frac{\partial{\cal C}(\boldsymbol{\Theta}^L)}{\partial w_{ij}^L} = \left(a_j^L - y_j\right)\frac{\partial a_j^L}{\partial w_{ij}^{L}}, +\frac{\partial z_j^l}{\partial a_i^{l-1}} = w_{ji}^l. $$ -The last partial derivative can easily be computed and reads (by applying the chain rule)
+With our definition of the activation function we have that (note that this function depends only on \( z_j^l \))
$$ -\frac{\partial a_j^L}{\partial w_{ij}^{L}} = \frac{\partial a_j^L}{\partial z_{j}^{L}}\frac{\partial z_j^L}{\partial w_{ij}^{L}}=a_j^L(1-a_j^L)a_i^{L-1}. +\frac{\partial a_j^l}{\partial z_j^{l}} = a_j^l(1-a_j^l)=\sigma(z_j^l)(1-\sigma(z_j^l)). $$ @@ -503,7 +507,7 @@ $$
-
We have thus
+With these definitions we can now compute the derivative of the cost function in terms of the weights.
+ +Let us specialize to the output layer \( l=L \). Our cost function is
$$ -\frac{\partial{\cal C}((\boldsymbol{\Theta}^L)}{\partial w_{ij}^L} = \left(a_j^L - y_j\right)a_j^L(1-a_j^L)a_i^{L-1}, +{\cal C}(\boldsymbol{\Theta}^L) = \frac{1}{2}\sum_{i=1}^n\left(y_i - \tilde{y}_i\right)^2=\frac{1}{2}\sum_{i=1}^n\left(a_i^L - y_i\right)^2, $$ -Defining
+The derivative of this function with respect to the weights is
+ $$ -\delta_j^L = a_j^L(1-a_j^L)\left(a_j^L - y_j\right) = \sigma'(z_j^L)\frac{\partial {\cal C}}{\partial (a_j^L)}, +\frac{\partial{\cal C}(\boldsymbol{\Theta}^L)}{\partial w_{ij}^L} = \left(a_j^L - y_j\right)\frac{\partial a_j^L}{\partial w_{ij}^{L}}, $$ -and using the Hadamard product of two vectors we can write this as
+The last partial derivative can easily be computed and reads (by applying the chain rule)
$$ -\boldsymbol{\delta}^L = \sigma'(\boldsymbol{z}^L)\circ\frac{\partial {\cal C}}{\partial (\boldsymbol{a}^L)}. +\frac{\partial a_j^L}{\partial w_{ij}^{L}} = \frac{\partial a_j^L}{\partial z_{j}^{L}}\frac{\partial z_j^L}{\partial w_{ij}^{L}}=a_j^L(1-a_j^L)a_i^{L-1}. $$ @@ -500,7 +510,7 @@ $$
-
We have thus
+$$ +\frac{\partial{\cal C}((\boldsymbol{\Theta}^L)}{\partial w_{ij}^L} = \left(a_j^L - y_j\right)a_j^L(1-a_j^L)a_i^{L-1}, +$$ + +Defining
+$$ +\delta_j^L = a_j^L(1-a_j^L)\left(a_j^L - y_j\right) = \sigma'(z_j^L)\frac{\partial {\cal C}}{\partial (a_j^L)}, +$$ + +and using the Hadamard product of two vectors we can write this as
+$$ +\boldsymbol{\delta}^L = \sigma'(\boldsymbol{z}^L)\circ\frac{\partial {\cal C}}{\partial (\boldsymbol{a}^L)}. +$$ -This is an important expression. The second term on the right handside -measures how fast the cost function is changing as a function of the $j$th -output activation. If, for example, the cost function doesn't depend -much on a particular output node \( j \), then \( \delta_j^L \) will be small, -which is what we would expect. The first term on the right, measures -how fast the activation function \( f \) is changing at a given activation -value \( z_j^L \). -
@@ -493,7 +507,7 @@ value \( z_j^L \).
-
Notice that everything in the above equations is easily computed. In -particular, we compute \( z_j^L \) while computing the behaviour of the -network, and it is only a small additional overhead to compute -\( \sigma'(z^L_j) \). The exact form of the derivative with respect to the -output depends on the form of the cost function. -However, provided the cost function is known there should be little -trouble in calculating +
This is an important expression. The second term on the right handside +measures how fast the cost function is changing as a function of the $j$th +output activation. If, for example, the cost function doesn't depend +much on a particular output node \( j \), then \( \delta_j^L \) will be small, +which is what we would expect. The first term on the right, measures +how fast the activation function \( f \) is changing at a given activation +value \( z_j^L \).
-$$ -\frac{\partial {\cal C}}{\partial (a_j^L)} -$$ - -With the definition of \( \delta_j^L \) we have a more compact definition of the derivative of the cost function in terms of the weights, namely
-$$ -\frac{\partial{\cal C}}{\partial w_{ij}^L} = \delta_j^La_i^{L-1}. -$$ - -diff --git a/doc/pub/week42/html/._week42-bs036.html b/doc/pub/week42/html/._week42-bs036.html index 94383fcbd..4c0068c55 100644 --- a/doc/pub/week42/html/._week42-bs036.html +++ b/doc/pub/week42/html/._week42-bs036.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
-
It is also easy to see that our previous equation can be written as
+Notice that everything in the above equations is easily computed. In +particular, we compute \( z_j^L \) while computing the behaviour of the +network, and it is only a small additional overhead to compute +\( \sigma'(z^L_j) \). The exact form of the derivative with respect to the +output depends on the form of the cost function. +However, provided the cost function is known there should be little +trouble in calculating +
$$ -\delta_j^L =\frac{\partial {\cal C}}{\partial z_j^L}= \frac{\partial {\cal C}}{\partial a_j^L}\frac{\partial a_j^L}{\partial z_j^L}, +\frac{\partial {\cal C}}{\partial (a_j^L)} $$ -which can also be interpreted as the partial derivative of the cost function with respect to the biases \( b_j^L \), namely
+With the definition of \( \delta_j^L \) we have a more compact definition of the derivative of the cost function in terms of the weights, namely
$$ -\delta_j^L = \frac{\partial {\cal C}}{\partial b_j^L}\frac{\partial b_j^L}{\partial z_j^L}=\frac{\partial {\cal C}}{\partial b_j^L}, +\frac{\partial{\cal C}}{\partial w_{ij}^L} = \delta_j^La_i^{L-1}. $$ -That is, the error \( \delta_j^L \) is exactly equal to the rate of change of the cost function as a function of the bias.
@@ -497,7 +510,7 @@ $$
-
We have now three equations that are essential for the computations of the derivatives of the cost function at the output layer. These equations are needed to start the algorithm and they are
+It is also easy to see that our previous equation can be written as
$$ -\begin{equation} -\frac{\partial{\cal C}(\boldsymbol{W^L})}{\partial w_{ij}^L} = \delta_j^La_i^{L-1}, -\tag{1} -\end{equation} +\delta_j^L =\frac{\partial {\cal C}}{\partial z_j^L}= \frac{\partial {\cal C}}{\partial a_j^L}\frac{\partial a_j^L}{\partial z_j^L}, $$ -and
+which can also be interpreted as the partial derivative of the cost function with respect to the biases \( b_j^L \), namely
$$ -\begin{equation} -\delta_j^L = \sigma'(z_j^L)\frac{\partial {\cal C}}{\partial (a_j^L)}, -\tag{2} -\end{equation} -$$ - -and
- -$$ -\begin{equation} -\delta_j^L = \frac{\partial {\cal C}}{\partial b_j^L}, -\tag{3} -\end{equation} +\delta_j^L = \frac{\partial {\cal C}}{\partial b_j^L}\frac{\partial b_j^L}{\partial z_j^L}=\frac{\partial {\cal C}}{\partial b_j^L}, $$ +That is, the error \( \delta_j^L \) is exactly equal to the rate of change of the cost function as a function of the bias.
@@ -511,7 +504,7 @@ $$
-
We have now three equations that are essential for the computations of the derivatives of the cost function at the output layer. These equations are needed to start the algorithm and they are
-We have that (replacing \( L \) with a general layer \( l \))
$$ -\delta_j^l =\frac{\partial {\cal C}}{\partial z_j^l}. +\begin{equation} +\frac{\partial{\cal C}(\boldsymbol{W^L})}{\partial w_{ij}^L} = \delta_j^La_i^{L-1}, +\tag{1} +\end{equation} +$$ + +and
+$$ +\begin{equation} +\delta_j^L = \sigma'(z_j^L)\frac{\partial {\cal C}}{\partial (a_j^L)}, +\tag{2} +\end{equation} +$$ + +and
+ +$$ +\begin{equation} +\delta_j^L = \frac{\partial {\cal C}}{\partial b_j^L}, +\tag{3} +\end{equation} $$ -We want to express this in terms of the equations for layer \( l+1 \).
@@ -491,7 +518,7 @@ $$
-
We obtain
+We have that (replacing \( L \) with a general layer \( l \))
$$ -\delta_j^l =\sum_k \frac{\partial {\cal C}}{\partial z_k^{l+1}}\frac{\partial z_k^{l+1}}{\partial z_j^{l}}=\sum_k \delta_k^{l+1}\frac{\partial z_k^{l+1}}{\partial z_j^{l}}, +\delta_j^l =\frac{\partial {\cal C}}{\partial z_j^l}. $$ -and recalling that
-$$ -z_j^{l+1} = \sum_{i=1}^{M_{l}}w_{ij}^{l+1}a_i^{l}+b_j^{l+1}, -$$ - -with \( M_l \) being the number of nodes in layer \( l \), we obtain
-$$ -\delta_j^l =\sum_k \delta_k^{l+1}w_{kj}^{l+1}\sigma'(z_j^l), -$$ - -This is our final equation.
- -We are now ready to set up the algorithm for back propagation and learning the weights and biases.
+We want to express this in terms of the equations for layer \( l+1 \).
@@ -503,7 +498,7 @@ $$
-
The four equations provide us with a way of computing the gradient of the cost function. Let us write this out in the form of an algorithm.
+We obtain
+$$ +\delta_j^l =\sum_k \frac{\partial {\cal C}}{\partial z_k^{l+1}}\frac{\partial z_k^{l+1}}{\partial z_j^{l}}=\sum_k \delta_k^{l+1}\frac{\partial z_k^{l+1}}{\partial z_j^{l}}, +$$ -First, we set up the input data \( \boldsymbol{x} \) and the activations -\( \boldsymbol{z}_1 \) of the input layer and compute the activation function and -the pertinent outputs \( \boldsymbol{a}^1 \). -
+and recalling that
+$$ +z_j^{l+1} = \sum_{i=1}^{M_{l}}w_{ij}^{l+1}a_i^{l}+b_j^{l+1}, +$$ -Secondly, we perform then the feed forward till we reach the output -layer and compute all \( \boldsymbol{z}_l \) of the input layer and compute the -activation function and the pertinent outputs \( \boldsymbol{a}^l \) for -\( l=1,2,3,\dots,L \). -
+with \( M_l \) being the number of nodes in layer \( l \), we obtain
+$$ +\delta_j^l =\sum_k \delta_k^{l+1}w_{kj}^{l+1}\sigma'(z_j^l), +$$ -Notation: The first hidden layer has \( l=1 \) as label and the final output layer has \( l=L \).
+This is our final equation.
+ +We are now ready to set up the algorithm for back propagation and learning the weights and biases.
@@ -499,7 +510,7 @@ activation function and the pertinent outputs \( \boldsymbol{a}^l \) for
-
Thereafter we compute the ouput error \( \boldsymbol{\delta}^L \) by computing all
-$$ -\delta_j^L = \sigma'(z_j^L)\frac{\partial {\cal C}}{\partial (a_j^L)}. -$$ +The four equations provide us with a way of computing the gradient of the cost function. Let us write this out in the form of an algorithm.
-Then we compute the back propagate error for each \( l=L-1,L-2,\dots,1 \) as
-$$ -\delta_j^l = \sum_k \delta_k^{l+1}w_{kj}^{l+1}\sigma'(z_j^l). -$$ +First, we set up the input data \( \boldsymbol{x} \) and the activations +\( \boldsymbol{z}_1 \) of the input layer and compute the activation function and +the pertinent outputs \( \boldsymbol{a}^1 \). +
+Secondly, we perform then the feed forward till we reach the output +layer and compute all \( \boldsymbol{z}_l \) of the input layer and compute the +activation function and the pertinent outputs \( \boldsymbol{a}^l \) for +\( l=1,2,3,\dots,L \). +
+ +Notation: The first hidden layer has \( l=1 \) as label and the final output layer has \( l=L \).
@@ -495,7 +506,7 @@ $$
-
Finally, we update the weights and the biases using gradient descent -for each \( l=L-1,L-2,\dots,1 \) (the first hidden layer) and update the weights and biases -according to the rules -
+Thereafter we compute the ouput error \( \boldsymbol{\delta}^L \) by computing all
$$ -w_{ij}^l\leftarrow = w_{ij}^l- \eta \delta_j^la_i^{l-1}, +\delta_j^L = \sigma'(z_j^L)\frac{\partial {\cal C}}{\partial (a_j^L)}. $$ - +Then we compute the back propagate error for each \( l=L-1,L-2,\dots,1 \) as
$$ -b_j^l \leftarrow b_j^l-\eta \frac{\partial {\cal C}}{\partial b_j^l}=b_j^l-\eta \delta_j^l, +\delta_j^l = \sum_k \delta_k^{l+1}w_{kj}^{l+1}\sigma'(z_j^l). $$ -with \( \eta \) being the learning rate.
@@ -500,7 +502,7 @@ $$
-
With the back propagate error for each \( l=L-1,L-2,\dots,1 \) as
-$$ -\delta_j^l = \sum_k \delta_k^{l+1}w_{kj}^{l+1}\sigma'(z_j^l), -$$ +Finally, we update the weights and the biases using gradient descent +for each \( l=L-1,L-2,\dots,1 \) (the first hidden layer) and update the weights and biases +according to the rules +
-we update the weights and the biases using gradient descent for each \( l=L-1,L-2,\dots,1 \) and update the weights and biases according to the rules
$$ w_{ij}^l\leftarrow = w_{ij}^l- \eta \delta_j^la_i^{l-1}, $$ @@ -474,6 +480,7 @@ $$ b_j^l \leftarrow b_j^l-\eta \frac{\partial {\cal C}}{\partial b_j^l}=b_j^l-\eta \delta_j^l, $$ +with \( \eta \) being the learning rate.
@@ -500,7 +507,7 @@ $$
-
With the back propagate error for each \( l=L-1,L-2,\dots,1 \) as
+$$ +\delta_j^l = \sum_k \delta_k^{l+1}w_{kj}^{l+1}\sigma'(z_j^l), +$$ + +we update the weights and the biases using gradient descent for each \( l=L-1,L-2,\dots,1 \) and update the weights and biases according to the rules
+$$ +w_{ij}^l\leftarrow = w_{ij}^l- \eta \delta_j^la_i^{l-1}, +$$ + + +$$ +b_j^l \leftarrow b_j^l-\eta \frac{\partial {\cal C}}{\partial b_j^l}=b_j^l-\eta \delta_j^l, +$$ -A property that characterizes a neural network, other than its -connectivity, is the choice of activation function(s). As described -in, the following restrictions are imposed on an activation function -for a FFNN to fulfill the universal approximation theorem -
-
-
The second requirement excludes all linear functions. Furthermore, in -a MLP with only linear activation functions, each layer simply -performs a linear transformation of its inputs. +
A property that characterizes a neural network, other than its +connectivity, is the choice of activation function(s). As described +in, the following restrictions are imposed on an activation function +for a FFNN to fulfill the universal approximation theorem
-Regardless of the number of layers, the output of the NN will be -nothing but a linear function of the inputs. Thus we need to introduce -some kind of non-linearity to the NN to be able to fit non-linear -functions Typical examples are the logistic Sigmoid -
- -$$ - \sigma(x) = \frac{1}{1 + e^{-x}}, -$$ - -and the hyperbolic tangent function
-$$ - \sigma(x) = \tanh(x) -$$ - - +diff --git a/doc/pub/week42/html/._week42-bs046.html b/doc/pub/week42/html/._week42-bs046.html index a4a93ab63..b42bf51ce 100644 --- a/doc/pub/week42/html/._week42-bs046.html +++ b/doc/pub/week42/html/._week42-bs046.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
-
The sigmoid function are more biologically plausible because the -output of inactive neurons are zero. Such activation function are -called one-sided. However, it has been shown that the hyperbolic -tangent performs better than the sigmoid for training MLPs. has -become the most popular for deep neural networks +
The second requirement excludes all linear functions. Furthermore, in +a MLP with only linear activation functions, each layer simply +performs a linear transformation of its inputs.
+Regardless of the number of layers, the output of the NN will be +nothing but a linear function of the inputs. Thus we need to introduce +some kind of non-linearity to the NN to be able to fit non-linear +functions Typical examples are the logistic Sigmoid +
- -"""The sigmoid function (or the logistic curve) is a
-function that takes any real number, z, and outputs a number (0,1).
-It is useful in neural networks for assigning weights on a relative scale.
-The value z is the weighted sum of parameters involved in the learning algorithm."""
+$$
+ \sigma(x) = \frac{1}{1 + e^{-x}},
+$$
-import numpy
-import matplotlib.pyplot as plt
-import math as mt
-
-z = numpy.arange(-5, 5, .1)
-sigma_fn = numpy.vectorize(lambda z: 1/(1+numpy.exp(-z)))
-sigma = sigma_fn(z)
-
-fig = plt.figure()
-ax = fig.add_subplot(111)
-ax.plot(z, sigma)
-ax.set_ylim([-0.1, 1.1])
-ax.set_xlim([-5,5])
-ax.grid(True)
-ax.set_xlabel('z')
-ax.set_title('sigmoid function')
-
-plt.show()
-
-"""Step Function"""
-z = numpy.arange(-5, 5, .02)
-step_fn = numpy.vectorize(lambda z: 1.0 if z >= 0.0 else 0.0)
-step = step_fn(z)
-
-fig = plt.figure()
-ax = fig.add_subplot(111)
-ax.plot(z, step)
-ax.set_ylim([-0.5, 1.5])
-ax.set_xlim([-5,5])
-ax.grid(True)
-ax.set_xlabel('z')
-ax.set_title('step function')
-
-plt.show()
-
-"""Sine Function"""
-z = numpy.arange(-2*mt.pi, 2*mt.pi, 0.1)
-t = numpy.sin(z)
-
-fig = plt.figure()
-ax = fig.add_subplot(111)
-ax.plot(z, t)
-ax.set_ylim([-1.0, 1.0])
-ax.set_xlim([-2*mt.pi,2*mt.pi])
-ax.grid(True)
-ax.set_xlabel('z')
-ax.set_title('sine function')
-
-plt.show()
-
-"""Plots a graph of the squashing function used by a rectified linear
-unit"""
-z = numpy.arange(-2, 2, .1)
-zero = numpy.zeros(len(z))
-y = numpy.max([zero, z], axis=0)
-
-fig = plt.figure()
-ax = fig.add_subplot(111)
-ax.plot(z, y)
-ax.set_ylim([-2.0, 2.0])
-ax.set_xlim([-2.0, 2.0])
-ax.grid(True)
-ax.set_xlabel('z')
-ax.set_title('Rectified linear unit')
-
-plt.show()
-
-and the hyperbolic tangent function
+$$ + \sigma(x) = \tanh(x) +$$@@ -585,7 +512,7 @@ plt.show()
-
The flexibility of neural networks is also one of their main -drawbacks: there are many hyperparameters to tweak. Not only can you -use any imaginable network topology (how neurons/nodes are -interconnected), but even in a simple FFNN you can change the number -of layers, the number of neurons per layer, the type of activation -function to use in each layer, the weight initialization logic, the -stochastic gradient optmized and much more. How do you know what -combination of hyperparameters is the best for your task? +
The sigmoid function are more biologically plausible because the +output of inactive neurons are zero. Such activation function are +called one-sided. However, it has been shown that the hyperbolic +tangent performs better than the sigmoid for training MLPs. has +become the most popular for deep neural networks
-However,since there are many hyperparameters to tune, and since -training a neural network on a large dataset takes a lot of time, you -will only be able to explore a tiny part of the hyperparameter space. -
-"""The sigmoid function (or the logistic curve) is a
+function that takes any real number, z, and outputs a number (0,1).
+It is useful in neural networks for assigning weights on a relative scale.
+The value z is the weighted sum of parameters involved in the learning algorithm."""
+
+import numpy
+import matplotlib.pyplot as plt
+import math as mt
+
+z = numpy.arange(-5, 5, .1)
+sigma_fn = numpy.vectorize(lambda z: 1/(1+numpy.exp(-z)))
+sigma = sigma_fn(z)
+
+fig = plt.figure()
+ax = fig.add_subplot(111)
+ax.plot(z, sigma)
+ax.set_ylim([-0.1, 1.1])
+ax.set_xlim([-5,5])
+ax.grid(True)
+ax.set_xlabel('z')
+ax.set_title('sigmoid function')
+
+plt.show()
+
+"""Step Function"""
+z = numpy.arange(-5, 5, .02)
+step_fn = numpy.vectorize(lambda z: 1.0 if z >= 0.0 else 0.0)
+step = step_fn(z)
+
+fig = plt.figure()
+ax = fig.add_subplot(111)
+ax.plot(z, step)
+ax.set_ylim([-0.5, 1.5])
+ax.set_xlim([-5,5])
+ax.grid(True)
+ax.set_xlabel('z')
+ax.set_title('step function')
+
+plt.show()
+
+"""Sine Function"""
+z = numpy.arange(-2*mt.pi, 2*mt.pi, 0.1)
+t = numpy.sin(z)
+
+fig = plt.figure()
+ax = fig.add_subplot(111)
+ax.plot(z, t)
+ax.set_ylim([-1.0, 1.0])
+ax.set_xlim([-2*mt.pi,2*mt.pi])
+ax.grid(True)
+ax.set_xlabel('z')
+ax.set_title('sine function')
+
+plt.show()
+
+"""Plots a graph of the squashing function used by a rectified linear
+unit"""
+z = numpy.arange(-2, 2, .1)
+zero = numpy.zeros(len(z))
+y = numpy.max([zero, z], axis=0)
+
+fig = plt.figure()
+ax = fig.add_subplot(111)
+ax.plot(z, y)
+ax.set_ylim([-2.0, 2.0])
+ax.set_xlim([-2.0, 2.0])
+ax.grid(True)
+ax.set_xlabel('z')
+ax.set_title('Rectified linear unit')
+
+plt.show()
+
+
-
For many problems you can start with just one or two hidden layers and -it will work just fine. For the MNIST data set you ca easily get a -high accuracy using just one hidden layer with a few hundred neurons. -You can reach for this data set above 98% accuracy using two hidden -layers with the same total amount of neurons, in roughly the same -amount of training time. +
The flexibility of neural networks is also one of their main +drawbacks: there are many hyperparameters to tweak. Not only can you +use any imaginable network topology (how neurons/nodes are +interconnected), but even in a simple FFNN you can change the number +of layers, the number of neurons per layer, the type of activation +function to use in each layer, the weight initialization logic, the +stochastic gradient optmized and much more. How do you know what +combination of hyperparameters is the best for your task?
-For more complex problems, you can gradually ramp up the number of -hidden layers, until you start overfitting the training set. Very -complex tasks, such as large image classification or speech -recognition, typically require networks with dozens of layers and they -need a huge amount of training data. However, you will rarely have to -train such networks from scratch: it is much more common to reuse -parts of a pretrained state-of-the-art network that performs a similar -task. +
However,since there are many hyperparameters to tune, and since +training a neural network on a large dataset takes a lot of time, you +will only be able to explore a tiny part of the hyperparameter space.
+diff --git a/doc/pub/week42/html/._week42-bs049.html b/doc/pub/week42/html/._week42-bs049.html index 744a0b520..d459c5de5 100644 --- a/doc/pub/week42/html/._week42-bs049.html +++ b/doc/pub/week42/html/._week42-bs049.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
- -
The Back propagation algorithm we derived above works by going from -the output layer to the input layer, propagating the error gradient on -the way. Once the algorithm has computed the gradient of the cost -function with regards to each parameter in the network, it uses these -gradients to update each parameter with a Gradient Descent (GD) step. +
For many problems you can start with just one or two hidden layers and +it will work just fine. For the MNIST data set you ca easily get a +high accuracy using just one hidden layer with a few hundred neurons. +You can reach for this data set above 98% accuracy using two hidden +layers with the same total amount of neurons, in roughly the same +amount of training time.
-Unfortunately for us, the gradients often get smaller and smaller as -the algorithm progresses down to the first hidden layers. As a result, -the GD update leaves the lower layer connection weights virtually -unchanged, and training never converges to a good solution. This is -known in the literature as the vanishing gradients problem. +
For more complex problems, you can gradually ramp up the number of +hidden layers, until you start overfitting the training set. Very +complex tasks, such as large image classification or speech +recognition, typically require networks with dozens of layers and they +need a huge amount of training data. However, you will rarely have to +train such networks from scratch: it is much more common to reuse +parts of a pretrained state-of-the-art network that performs a similar +task.
@@ -498,7 +509,7 @@ known in the literature as the vanishing gradients problem.
- -
In other cases, the opposite can happen, namely the the gradients can -grow bigger and bigger. The result is that many of the layers get -large updates of the weights the algorithm diverges. This is the -exploding gradients problem, which is mostly encountered in -recurrent neural networks. More generally, deep neural networks suffer -from unstable gradients, different layers may learn at widely -different speeds +
The Back propagation algorithm we derived above works by going from +the output layer to the input layer, propagating the error gradient on +the way. Once the algorithm has computed the gradient of the cost +function with regards to each parameter in the network, it uses these +gradients to update each parameter with a Gradient Descent (GD) step. +
+ +Unfortunately for us, the gradients often get smaller and smaller as +the algorithm progresses down to the first hidden layers. As a result, +the GD update leaves the lower layer connection weights virtually +unchanged, and training never converges to a good solution. This is +known in the literature as the vanishing gradients problem.
@@ -493,7 +505,7 @@ different speeds
- -
Although this unfortunate behavior has been empirically observed for -quite a while (it was one of the reasons why deep neural networks were -mostly abandoned for a long time), it is only around 2010 that -significant progress was made in understanding it. -
- -A paper titled Understanding the Difficulty of Training Deep -Feedforward Neural Networks by Xavier Glorot and Yoshua Bengio found that -the problems with the popular logistic -sigmoid activation function and the weight initialization technique -that was most popular at the time, namely random initialization using -a normal distribution with a mean of 0 and a standard deviation of -1. +
In other cases, the opposite can happen, namely the the gradients can +grow bigger and bigger. The result is that many of the layers get +large updates of the weights the algorithm diverges. This is the +exploding gradients problem, which is mostly encountered in +recurrent neural networks. More generally, deep neural networks suffer +from unstable gradients, different layers may learn at widely +different speeds
@@ -499,7 +500,7 @@ a normal distribution with a mean of 0 and a standard deviation of
- -
They showed that with this activation function and this -initialization scheme, the variance of the outputs of each layer is -much greater than the variance of its inputs. Going forward in the -network, the variance keeps increasing after each layer until the -activation function saturates at the top layers. This is actually made -worse by the fact that the logistic function has a mean of 0.5, not 0 -(the hyperbolic tangent function has a mean of 0 and behaves slightly -better than the logistic function in deep networks). +
Although this unfortunate behavior has been empirically observed for +quite a while (it was one of the reasons why deep neural networks were +mostly abandoned for a long time), it is only around 2010 that +significant progress was made in understanding it. +
+ +A paper titled Understanding the Difficulty of Training Deep +Feedforward Neural Networks by Xavier Glorot and Yoshua Bengio found that +the problems with the popular logistic +sigmoid activation function and the weight initialization technique +that was most popular at the time, namely random initialization using +a normal distribution with a mean of 0 and a standard deviation of +1.
@@ -494,7 +506,7 @@ better than the logistic function in deep networks).
-
Looking at the logistic activation function, when inputs become large -(negative or positive), the function saturates at 0 or 1, with a -derivative extremely close to 0. Thus when backpropagation kicks in, -it has virtually no gradient to propagate back through the network, -and what little gradient exists keeps getting diluted as -backpropagation progresses down through the top layers, so there is -really nothing left for the lower layers. -
- -In their paper, Glorot and Bengio propose a way to significantly -alleviate this problem. We need the signal to flow properly in both -directions: in the forward direction when making predictions, and in -the reverse direction when backpropagating gradients. We don’t want -the signal to die out, nor do we want it to explode and saturate. For -the signal to flow properly, the authors argue that we need the -variance of the outputs of each layer to be equal to the variance of -its inputs, and we also need the gradients to have equal variance -before and after flowing through a layer in the reverse direction. +
They showed that with this activation function and this +initialization scheme, the variance of the outputs of each layer is +much greater than the variance of its inputs. Going forward in the +network, the variance keeps increasing after each layer until the +activation function saturates at the top layers. This is actually made +worse by the fact that the logistic function has a mean of 0.5, not 0 +(the hyperbolic tangent function has a mean of 0 and behaves slightly +better than the logistic function in deep networks).
@@ -504,7 +501,7 @@ before and after flowing through a layer in the reverse direction.
-
One of the insights in the 2010 paper by Glorot and Bengio was that -the vanishing/exploding gradients problems were in part due to a poor -choice of activation function. Until then most people had assumed that -if Nature had chosen to use roughly sigmoid activation functions in -biological neurons, they must be an excellent choice. But it turns out -that other activation functions behave much better in deep neural -networks, in particular the ReLU activation function, mostly because -it does not saturate for positive values (and also because it is quite -fast to compute). +
Looking at the logistic activation function, when inputs become large +(negative or positive), the function saturates at 0 or 1, with a +derivative extremely close to 0. Thus when backpropagation kicks in, +it has virtually no gradient to propagate back through the network, +and what little gradient exists keeps getting diluted as +backpropagation progresses down through the top layers, so there is +really nothing left for the lower layers. +
+ +In their paper, Glorot and Bengio propose a way to significantly +alleviate this problem. We need the signal to flow properly in both +directions: in the forward direction when making predictions, and in +the reverse direction when backpropagating gradients. We don’t want +the signal to die out, nor do we want it to explode and saturate. For +the signal to flow properly, the authors argue that we need the +variance of the outputs of each layer to be equal to the variance of +its inputs, and we also need the gradients to have equal variance +before and after flowing through a layer in the reverse direction.
@@ -495,7 +511,7 @@ fast to compute).
-
The ReLU activation function suffers from a problem known as the dying -ReLUs: during training, some neurons effectively die, meaning they -stop outputting anything other than 0. -
- -In some cases, you may find that half of your network’s neurons are -dead, especially if you used a large learning rate. During training, -if a neuron’s weights get updated such that the weighted sum of the -neuron’s inputs is negative, it will start outputting 0. When this -happen, the neuron is unlikely to come back to life since the gradient -of the ReLU function is 0 when its input is negative. +
One of the insights in the 2010 paper by Glorot and Bengio was that +the vanishing/exploding gradients problems were in part due to a poor +choice of activation function. Until then most people had assumed that +if Nature had chosen to use roughly sigmoid activation functions in +biological neurons, they must be an excellent choice. But it turns out +that other activation functions behave much better in deep neural +networks, in particular the ReLU activation function, mostly because +it does not saturate for positive values (and also because it is quite +fast to compute).
@@ -497,7 +502,7 @@ of the ReLU function is 0 when its input is negative.
-
To solve this problem, nowadays practitioners use a variant of the -ReLU function, such as the leaky ReLU discussed above or the so-called -exponential linear unit (ELU) function +
The ReLU activation function suffers from a problem known as the dying +ReLUs: during training, some neurons effectively die, meaning they +stop outputting anything other than 0.
-$$ -ELU(z) = \left\{\begin{array}{cc} \alpha\left( \exp{(z)}-1\right) & z < 0,\\ z & z \ge 0.\end{array}\right. -$$ - +In some cases, you may find that half of your network’s neurons are +dead, especially if you used a large learning rate. During training, +if a neuron’s weights get updated such that the weighted sum of the +neuron’s inputs is negative, it will start outputting 0. When this +happen, the neuron is unlikely to come back to life since the gradient +of the ReLU function is 0 when its input is negative. +
@@ -494,7 +504,7 @@ $$
-
In general it seems that the ELU activation function is better than -the leaky ReLU function (and its variants), which is better than -ReLU. ReLU performs better than \( \tanh \) which in turn performs better -than the logistic function. +
To solve this problem, nowadays practitioners use a variant of the +ReLU function, such as the leaky ReLU discussed above or the so-called +exponential linear unit (ELU) function
-If runtime performance is an issue, then you may opt for the leaky -ReLU function over the ELU function If you don’t want to tweak yet -another hyperparameter, you may just use the default \( \alpha \) of -\( 0.01 \) for the leaky ReLU, and \( 1 \) for ELU. If you have spare time and -computing power, you can use cross-validation or bootstrap to evaluate -other activation functions. -
+$$ +ELU(z) = \left\{\begin{array}{cc} \alpha\left( \exp{(z)}-1\right) & z < 0,\\ z & z \ge 0.\end{array}\right. +$$ +@@ -498,7 +501,7 @@ other activation functions.
-
In most cases you can use the ReLU activation function in the hidden -layers (or one of its variants). +
In general it seems that the ELU activation function is better than +the leaky ReLU function (and its variants), which is better than +ReLU. ReLU performs better than \( \tanh \) which in turn performs better +than the logistic function.
-It is a bit faster to compute than other activation functions, and the -gradient descent optimization does in general not get stuck. +
If runtime performance is an issue, then you may opt for the leaky +ReLU function over the ELU function If you don’t want to tweak yet +another hyperparameter, you may just use the default \( \alpha \) of +\( 0.01 \) for the leaky ReLU, and \( 1 \) for ELU. If you have spare time and +computing power, you can use cross-validation or bootstrap to evaluate +other activation functions.
-For the output layer: - -
-
Batch Normalization aims to address the vanishing/exploding gradients -problems, and more generally the problem that the distribution of each -layer’s inputs changes during training, as the parameters of the -previous layers change. +
In most cases you can use the ReLU activation function in the hidden +layers (or one of its variants).
-The technique consists of adding an operation in the model just before -the activation function of each layer, simply zero-centering and -normalizing the inputs, then scaling and shifting the result using two -new parameters per layer (one for scaling, the other for shifting). In -other words, this operation lets the model learn the optimal scale and -mean of the inputs for each layer. In order to zero-center and -normalize the inputs, the algorithm needs to estimate the inputs’ mean -and standard deviation. It does so by evaluating the mean and standard -deviation of the inputs over the current mini-batch, from this the -name batch normalization. +
It is a bit faster to compute than other activation functions, and the +gradient descent optimization does in general not get stuck.
+For the output layer: + +diff --git a/doc/pub/week42/html/._week42-bs060.html b/doc/pub/week42/html/._week42-bs060.html index a5815a79e..d7284fa32 100644 --- a/doc/pub/week42/html/._week42-bs060.html +++ b/doc/pub/week42/html/._week42-bs060.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
-
It is a fairly simple algorithm: at every training step, every neuron -(including the input neurons but excluding the output neurons) has a -probability \( p \) of being temporarily dropped out, meaning it will be -entirely ignored during this training step, but it may be active -during the next step. +
Batch Normalization aims to address the vanishing/exploding gradients +problems, and more generally the problem that the distribution of each +layer’s inputs changes during training, as the parameters of the +previous layers change.
-The hyperparameter \( p \) is called the dropout rate, and it is typically -set to 50%. After training, the neurons are not dropped anymore. It -is viewed as one of the most popular regularization techniques. +
The technique consists of adding an operation in the model just before +the activation function of each layer, simply zero-centering and +normalizing the inputs, then scaling and shifting the result using two +new parameters per layer (one for scaling, the other for shifting). In +other words, this operation lets the model learn the optimal scale and +mean of the inputs for each layer. In order to zero-center and +normalize the inputs, the algorithm needs to estimate the inputs’ mean +and standard deviation. It does so by evaluating the mean and standard +deviation of the inputs over the current mini-batch, from this the +name batch normalization.
@@ -496,7 +509,7 @@ is viewed as one of the most popular regularization techniques.
-
A popular technique to lessen the exploding gradients problem is to -simply clip the gradients during backpropagation so that they never -exceed some threshold (this is mostly useful for recurrent neural -networks). +
It is a fairly simple algorithm: at every training step, every neuron +(including the input neurons but excluding the output neurons) has a +probability \( p \) of being temporarily dropped out, meaning it will be +entirely ignored during this training step, but it may be active +during the next step.
-This technique is called Gradient Clipping.
- -In general however, Batch -Normalization is preferred. +
The hyperparameter \( p \) is called the dropout rate, and it is typically +set to 50%. After training, the neurons are not dropped anymore. It +is viewed as one of the most popular regularization techniques.
@@ -496,7 +503,7 @@ Normalization is preferred.
- -
The first thing we would like to do is divide the data into two or -three parts. A training set, a validation or dev (development) set, -and a test set. The test set is the data on which we want to make -predictions. The dev set is a subset of the training data we use to -check how well we are doing out-of-sample, after training the model on -the training dataset. We use the validation error as a proxy for the -test error in order to make tweaks to our model. It is crucial that we -do not use any of the test data to train the algorithm. This is a -cardinal sin in ML. Then: +
A popular technique to lessen the exploding gradients problem is to +simply clip the gradients during backpropagation so that they never +exceed some threshold (this is mostly useful for recurrent neural +networks). +
+ +This technique is called Gradient Clipping.
+ +In general however, Batch +Normalization is preferred.
-diff --git a/doc/pub/week42/html/._week42-bs063.html b/doc/pub/week42/html/._week42-bs063.html index 076540ce0..c2070555b 100644 --- a/doc/pub/week42/html/._week42-bs063.html +++ b/doc/pub/week42/html/._week42-bs063.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
- -
If the validation and test sets are drawn from the same distributions, -then a good performance on the validation set should lead to similarly -good performance on the test set. -
- -However, sometimes -the training data and test data differ in subtle ways because, for -example, they are collected using slightly different methods, or -because it is cheaper to collect data in one way versus another. In -this case, there can be a mismatch between the training and test -data. This can lead to the neural network overfitting these small -differences between the test and training sets, and a poor performance -on the test set despite having a good performance on the validation -set. To rectify this, Andrew Ng suggests making two validation or dev -sets, one constructed from the training data and one constructed from -the test data. The difference between the performance of the algorithm -on these two validation sets quantifies the train-test mismatch. This -can serve as another important diagnostic when using DNNs for -supervised learning. +
The first thing we would like to do is divide the data into two or +three parts. A training set, a validation or dev (development) set, +and a test set. The test set is the data on which we want to make +predictions. The dev set is a subset of the training data we use to +check how well we are doing out-of-sample, after training the model on +the training dataset. We use the validation error as a proxy for the +test error in order to make tweaks to our model. It is crucial that we +do not use any of the test data to train the algorithm. This is a +cardinal sin in ML. Then:
+diff --git a/doc/pub/week42/html/._week42-bs064.html b/doc/pub/week42/html/._week42-bs064.html index d8df2b23a..037a0c524 100644 --- a/doc/pub/week42/html/._week42-bs064.html +++ b/doc/pub/week42/html/._week42-bs064.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
-
Like all statistical methods, supervised learning using neural -networks has important limitations. This is especially important when -one seeks to apply these methods, especially to physics problems. Like -all tools, DNNs are not a universal solution. Often, the same or -better performance on a task can be achieved by using a few -hand-engineered features (or even a collection of random -features). +
If the validation and test sets are drawn from the same distributions, +then a good performance on the validation set should lead to similarly +good performance on the test set. +
+ +However, sometimes +the training data and test data differ in subtle ways because, for +example, they are collected using slightly different methods, or +because it is cheaper to collect data in one way versus another. In +this case, there can be a mismatch between the training and test +data. This can lead to the neural network overfitting these small +differences between the test and training sets, and a poor performance +on the test set despite having a good performance on the validation +set. To rectify this, Andrew Ng suggests making two validation or dev +sets, one constructed from the training data and one constructed from +the test data. The difference between the performance of the algorithm +on these two validation sets quantifies the train-test mismatch. This +can serve as another important diagnostic when using DNNs for +supervised learning.
@@ -493,7 +512,7 @@ features).
-
Here we list some of the important limitations of supervised neural network based models.
+Like all statistical methods, supervised learning using neural +networks has important limitations. This is especially important when +one seeks to apply these methods, especially to physics problems. Like +all tools, DNNs are not a universal solution. Often, the same or +better performance on a task can be achieved by using a few +hand-engineered features (or even a collection of random +features). +
-diff --git a/doc/pub/week42/html/._week42-bs066.html b/doc/pub/week42/html/._week42-bs066.html index a3bba4859..e0d93fa69 100644 --- a/doc/pub/week42/html/._week42-bs066.html +++ b/doc/pub/week42/html/._week42-bs066.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
-
Here we list some of the important limitations of supervised neural network based models.
@@ -487,7 +497,7 @@ MathJax.Hub.Config({
-
Some of these remarks are particular to DNNs, others are shared by all supervised learning methods. This motivates the use of unsupervised methods which in part circumvent these problems.
-diff --git a/doc/pub/week42/html/._week42-bs068.html b/doc/pub/week42/html/._week42-bs068.html index fd594f43f..c46fd7545 100644 --- a/doc/pub/week42/html/._week42-bs068.html +++ b/doc/pub/week42/html/._week42-bs068.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
-
Let us write this out in the form of an algorithm.
- -First, we set up the input data \( \boldsymbol{x} \) and the activations -\( \boldsymbol{z}_1 \) of the input layer and compute the activation function and -the pertinent outputs \( \boldsymbol{a}^1 \). -
-Secondly, we perform then the feed forward till we reach the output -layer and compute all \( \boldsymbol{z}_l \) of the input layer and compute the -activation function and the pertinent outputs \( \boldsymbol{a}^l \) for -\( l=2,3,\dots,L \). -
-Thereafter we compute the ouput error \( \boldsymbol{\delta}^L \) by computing all
-$$ -\delta_j^L = f'(z_j^L)\frac{\partial {\cal C}}{\partial (a_j^L)}. -$$ -Then we compute the back propagate error for each \( l=L-1,L-2,\dots,2 \) as
-$$ -\delta_j^l = \sum_k \delta_k^{l+1}w_{kj}^{l+1}\sigma'(z_j^l). -$$ -Finally, we update the weights and the biases using gradient descent for each \( l=L-1,L-2,\dots,2 \) and update the weights and biases according to the rules
-$$ -w_{ij}^l\leftarrow = w_{ij}^l- \eta \delta_j^la_i^{l-1}, -$$ - - -$$ -b_j^l \leftarrow b_j^l-\eta \frac{\partial {\cal C}}{\partial b_j^l}=b_j^l-\eta \delta_j^l, -$$ -The parameter \( \eta \) is the learning parameter discussed in connection with the gradient descent methods. -Here it is convenient to use stochastic gradient descent (see the examples below) with mini-batches with an outer loop that steps through multiple epochs of training. -
+Some of these remarks are particular to DNNs, others are shared by all supervised learning methods. This motivates the use of unsupervised methods which in part circumvent these problems.
@@ -551,7 +496,7 @@ Here it is convenient to use stochastic gradient descent (see the examples below
- -
We are now gong to develop an example based on the MNIST data -base. This is a classification problem and we need to use our -cross-entropy function we discussed in connection with logistic -regression. The cross-entropy defines our cost function for the -classificaton problems with neural networks. -
+Let us write this out in the form of an algorithm.
-In binary classification with two classes \( (0, 1) \) we define the -logistic/sigmoid function as the probability that a particular input -is in class \( 0 \) or \( 1 \). This is possible because the logistic -function takes any input from the real numbers and inputs a number -between 0 and 1, and can therefore be interpreted as a probability. It -also has other nice properties, such as a derivative that is simple to -calculate. +
First, we set up the input data \( \boldsymbol{x} \) and the activations +\( \boldsymbol{z}_1 \) of the input layer and compute the activation function and +the pertinent outputs \( \boldsymbol{a}^1 \).
+For an input \( \boldsymbol{a} \) from the hidden layer, the probability that the input \( \boldsymbol{x} \) -is in class 0 or 1 is just. We let \( \theta \) represent the unknown weights and biases to be adjusted by our equations). The variable \( x \) -represents our activation values \( z \). We have + +
Secondly, we perform then the feed forward till we reach the output +layer and compute all \( \boldsymbol{z}_l \) of the input layer and compute the +activation function and the pertinent outputs \( \boldsymbol{a}^l \) for +\( l=2,3,\dots,L \).
+Thereafter we compute the ouput error \( \boldsymbol{\delta}^L \) by computing all
$$ -P(y = 0 \mid \boldsymbol{x}, \boldsymbol{\theta}) = \frac{1}{1 + \exp{(- \boldsymbol{x}})} , +\delta_j^L = f'(z_j^L)\frac{\partial {\cal C}}{\partial (a_j^L)}. +$$ +Then we compute the back propagate error for each \( l=L-1,L-2,\dots,2 \) as
+$$ +\delta_j^l = \sum_k \delta_k^{l+1}w_{kj}^{l+1}\sigma'(z_j^l). +$$ +Finally, we update the weights and the biases using gradient descent for each \( l=L-1,L-2,\dots,2 \) and update the weights and biases according to the rules
+$$ +w_{ij}^l\leftarrow = w_{ij}^l- \eta \delta_j^la_i^{l-1}, $$ -and
-$$ -P(y = 1 \mid \boldsymbol{x}, \boldsymbol{\theta}) = 1 - P(y = 0 \mid \boldsymbol{x}, \boldsymbol{\theta}) , -$$ -where \( y \in \{0, 1\} \) and \( \boldsymbol{\theta} \) represents the weights and biases -of our network. +$$ +b_j^l \leftarrow b_j^l-\eta \frac{\partial {\cal C}}{\partial b_j^l}=b_j^l-\eta \delta_j^l, +$$ +
The parameter \( \eta \) is the learning parameter discussed in connection with the gradient descent methods. +Here it is convenient to use stochastic gradient descent (see the examples below) with mini-batches with an outer loop that steps through multiple epochs of training.
-diff --git a/doc/pub/week42/html/._week42-bs070.html b/doc/pub/week42/html/._week42-bs070.html index 3356c508b..6fbaa97c1 100644 --- a/doc/pub/week42/html/._week42-bs070.html +++ b/doc/pub/week42/html/._week42-bs070.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
- -
Our cost function is given as (see the Logistic regression lectures)
-$$ -\mathcal{C}(\boldsymbol{\theta}) = - \ln P(\mathcal{D} \mid \boldsymbol{\theta}) = - \sum_{i=1}^n -y_i \ln[P(y_i = 0)] + (1 - y_i) \ln [1 - P(y_i = 0)] = \sum_{i=1}^n \mathcal{L}_i(\boldsymbol{\theta}) . -$$ - -This last equality means that we can interpret our cost function as a sum over the loss function -for each point in the dataset \( \mathcal{L}_i(\boldsymbol{\theta}) \). -The negative sign is just so that we can think about our algorithm as minimizing a positive number, rather -than maximizing a negative number. +
We are now gong to develop an example based on the MNIST data +base. This is a classification problem and we need to use our +cross-entropy function we discussed in connection with logistic +regression. The cross-entropy defines our cost function for the +classificaton problems with neural networks.
-In multiclass classification it is common to treat each integer label as a so called one-hot vector:
- -\( y = 5 \quad \rightarrow \quad \boldsymbol{y} = (0, 0, 0, 0, 0, 1, 0, 0, 0, 0) , \) and
- -\( y = 1 \quad \rightarrow \quad \boldsymbol{y} = (0, 1, 0, 0, 0, 0, 0, 0, 0, 0) , \) - -i.e. a binary bit string of length \( C \), where \( C = 10 \) is the number of classes in the MNIST dataset (numbers from \( 0 \) to \( 9 \))..
- -If \( \boldsymbol{x}_i \) is the \( i \)-th input (image), \( y_{ic} \) refers to the \( c \)-th component of the \( i \)-th -output vector \( \boldsymbol{y}_i \). -The probability of \( \boldsymbol{x}_i \) being in class \( c \) will be given by the softmax function: +
In binary classification with two classes \( (0, 1) \) we define the +logistic/sigmoid function as the probability that a particular input +is in class \( 0 \) or \( 1 \). This is possible because the logistic +function takes any input from the real numbers and inputs a number +between 0 and 1, and can therefore be interpreted as a probability. It +also has other nice properties, such as a derivative that is simple to +calculate.
+For an input \( \boldsymbol{a} \) from the hidden layer, the probability that the input \( \boldsymbol{x} \) +is in class 0 or 1 is just. We let \( \theta \) represent the unknown weights and biases to be adjusted by our equations). The variable \( x \) +represents our activation values \( z \). We have +
$$ -P(y_{ic} = 1 \mid \boldsymbol{x}_i, \boldsymbol{\theta}) = \frac{\exp{((\boldsymbol{a}_i^{hidden})^T \boldsymbol{w}_c)}} -{\sum_{c'=0}^{C-1} \exp{((\boldsymbol{a}_i^{hidden})^T \boldsymbol{w}_{c'})}} , +P(y = 0 \mid \boldsymbol{x}, \boldsymbol{\theta}) = \frac{1}{1 + \exp{(- \boldsymbol{x}})} , $$ -which reduces to the logistic function in the binary case. -The likelihood of this \( C \)-class classifier -is now given as: +
and
+$$ +P(y = 1 \mid \boldsymbol{x}, \boldsymbol{\theta}) = 1 - P(y = 0 \mid \boldsymbol{x}, \boldsymbol{\theta}) , +$$ + +where \( y \in \{0, 1\} \) and \( \boldsymbol{\theta} \) represents the weights and biases +of our network.
-$$ -P(\mathcal{D} \mid \boldsymbol{\theta}) = \prod_{i=1}^n \prod_{c=0}^{C-1} [P(y_{ic} = 1)]^{y_{ic}} . -$$ - -Again we take the negative log-likelihood to define our cost function:
- -$$ -\mathcal{C}(\boldsymbol{\theta}) = - \log{P(\mathcal{D} \mid \boldsymbol{\theta})}. -$$ - -See the logistic regression lectures for a full definition of the cost function.
- -The back propagation equations need now only a small change, namely the definition of a new cost function. We are thus ready to use the same equations as before!
@@ -533,7 +525,7 @@ $$
-
As an example of the above, relevant for project 2 as well, let us consider a binary class. As discussed in our logistic regression lectures, we defined a cost function in terms of the parameters \( \beta \) as
+Our cost function is given as (see the Logistic regression lectures)
$$ -\mathcal{C}(\boldsymbol{\beta}) = - \sum_{i=1}^n \left(y_i\log{p(y_i \vert x_i,\boldsymbol{\beta})}+(1-y_i)\log{1-p(y_i \vert x_i,\boldsymbol{\beta})}\right), +\mathcal{C}(\boldsymbol{\theta}) = - \ln P(\mathcal{D} \mid \boldsymbol{\theta}) = - \sum_{i=1}^n +y_i \ln[P(y_i = 0)] + (1 - y_i) \ln [1 - P(y_i = 0)] = \sum_{i=1}^n \mathcal{L}_i(\boldsymbol{\theta}) . $$ -where we had defined the logistic (sigmoid) function
-$$ -p(y_i =1\vert x_i,\boldsymbol{\beta})=\frac{\exp{(\beta_0+\beta_1 x_i)}}{1+\exp{(\beta_0+\beta_1 x_i)}}, -$$ - -and
-$$ -p(y_i =0\vert x_i,\boldsymbol{\beta})=1-p(y_i =1\vert x_i,\boldsymbol{\beta}). -$$ - -The parameters \( \boldsymbol{\beta} \) were defined using a minimization method like gradient descent or Newton-Raphson's method.
- -Now we replace \( x_i \) with the activation \( z_i^l \) for a given layer \( l \) and the outputs as \( y_i=a_i^l=f(z_i^l) \), with \( z_i^l \) now being a function of the weights \( w_{ij}^l \) and biases \( b_i^l \). -We have then +
This last equality means that we can interpret our cost function as a sum over the loss function +for each point in the dataset \( \mathcal{L}_i(\boldsymbol{\theta}) \). +The negative sign is just so that we can think about our algorithm as minimizing a positive number, rather +than maximizing a negative number.
-$$ -a_i^l = y_i = \frac{\exp{(z_i^l)}}{1+\exp{(z_i^l)}}, -$$ -with
-$$ -z_i^l = \sum_{j}w_{ij}^l a_j^{l-1}+b_i^l, -$$ +In multiclass classification it is common to treat each integer label as a so called one-hot vector:
-where the superscript \( l-1 \) indicates that these are the outputs from layer \( l-1 \). -Our cost function at the final layer \( l=L \) is now +
\( y = 5 \quad \rightarrow \quad \boldsymbol{y} = (0, 0, 0, 0, 0, 1, 0, 0, 0, 0) , \) and
+ +\( y = 1 \quad \rightarrow \quad \boldsymbol{y} = (0, 1, 0, 0, 0, 0, 0, 0, 0, 0) , \) + +i.e. a binary bit string of length \( C \), where \( C = 10 \) is the number of classes in the MNIST dataset (numbers from \( 0 \) to \( 9 \))..
+ +If \( \boldsymbol{x}_i \) is the \( i \)-th input (image), \( y_{ic} \) refers to the \( c \)-th component of the \( i \)-th +output vector \( \boldsymbol{y}_i \). +The probability of \( \boldsymbol{x}_i \) being in class \( c \) will be given by the softmax function:
+ $$ -\mathcal{C}(\boldsymbol{W}) = - \sum_{i=1}^n \left(t_i\log{a_i^L}+(1-t_i)\log{(1-a_i^L)}\right), +P(y_{ic} = 1 \mid \boldsymbol{x}_i, \boldsymbol{\theta}) = \frac{\exp{((\boldsymbol{a}_i^{hidden})^T \boldsymbol{w}_c)}} +{\sum_{c'=0}^{C-1} \exp{((\boldsymbol{a}_i^{hidden})^T \boldsymbol{w}_{c'})}} , $$ -where we have defined the targets \( t_i \). The derivatives of the cost function with respect to the output \( a_i^L \) are then easily calculated and we get
+which reduces to the logistic function in the binary case. +The likelihood of this \( C \)-class classifier +is now given as: +
+ $$ -\frac{\partial \mathcal{C}(\boldsymbol{W})}{\partial a_i^L} = \frac{a_i^L-t_i}{a_i^L(1-a_i^L)}. +P(\mathcal{D} \mid \boldsymbol{\theta}) = \prod_{i=1}^n \prod_{c=0}^{C-1} [P(y_{ic} = 1)]^{y_{ic}} . $$ -In case we use another activation function than the logistic one, we need to evaluate other derivatives.
+Again we take the negative log-likelihood to define our cost function:
+ +$$ +\mathcal{C}(\boldsymbol{\theta}) = - \log{P(\mathcal{D} \mid \boldsymbol{\theta})}. +$$ + +See the logistic regression lectures for a full definition of the cost function.
+ +The back propagation equations need now only a small change, namely the definition of a new cost function. We are thus ready to use the same equations as before!
@@ -527,7 +540,7 @@ $$
-
In case we employ the more general case given by the Softmax equation, we need to evaluate the derivative of the activation function with respect to the activation \( z_i^l \), that is we need
+As an example of the above, relevant for project 2 as well, let us consider a binary class. As discussed in our logistic regression lectures, we defined a cost function in terms of the parameters \( \beta \) as
$$ -\frac{\partial f(z_i^l)}{\partial w_{jk}^l} = -\frac{\partial f(z_i^l)}{\partial z_j^l} \frac{\partial z_j^l}{\partial w_{jk}^l}= \frac{\partial f(z_i^l)}{\partial z_j^l}a_k^{l-1}. +\mathcal{C}(\boldsymbol{\beta}) = - \sum_{i=1}^n \left(y_i\log{p(y_i \vert x_i,\boldsymbol{\beta})}+(1-y_i)\log{1-p(y_i \vert x_i,\boldsymbol{\beta})}\right), $$ -For the Softmax function we have
+where we had defined the logistic (sigmoid) function
$$ -f(z_i^l) = \frac{\exp{(z_i^l)}}{\sum_{m=1}^K\exp{(z_m^l)}}. +p(y_i =1\vert x_i,\boldsymbol{\beta})=\frac{\exp{(\beta_0+\beta_1 x_i)}}{1+\exp{(\beta_0+\beta_1 x_i)}}, $$ -Its derivative with respect to \( z_j^l \) gives
+and
$$ -\frac{\partial f(z_i^l)}{\partial z_j^l}= f(z_i^l)\left(\delta_{ij}-f(z_j^l)\right), +p(y_i =0\vert x_i,\boldsymbol{\beta})=1-p(y_i =1\vert x_i,\boldsymbol{\beta}). $$ -which in case of the simply binary model reduces to having \( i=j \).
+The parameters \( \boldsymbol{\beta} \) were defined using a minimization method like gradient descent or Newton-Raphson's method.
+ +Now we replace \( x_i \) with the activation \( z_i^l \) for a given layer \( l \) and the outputs as \( y_i=a_i^l=f(z_i^l) \), with \( z_i^l \) now being a function of the weights \( w_{ij}^l \) and biases \( b_i^l \). +We have then +
+$$ +a_i^l = y_i = \frac{\exp{(z_i^l)}}{1+\exp{(z_i^l)}}, +$$ + +with
+$$ +z_i^l = \sum_{j}w_{ij}^l a_j^{l-1}+b_i^l, +$$ + +where the superscript \( l-1 \) indicates that these are the outputs from layer \( l-1 \). +Our cost function at the final layer \( l=L \) is now +
+$$ +\mathcal{C}(\boldsymbol{W}) = - \sum_{i=1}^n \left(t_i\log{a_i^L}+(1-t_i)\log{(1-a_i^L)}\right), +$$ + +where we have defined the targets \( t_i \). The derivatives of the cost function with respect to the output \( a_i^L \) are then easily calculated and we get
+$$ +\frac{\partial \mathcal{C}(\boldsymbol{W})}{\partial a_i^L} = \frac{a_i^L-t_i}{a_i^L(1-a_i^L)}. +$$ + +In case we use another activation function than the logistic one, we need to evaluate other derivatives.
@@ -501,7 +534,7 @@ $$
- -
In case we employ the more general case given by the Softmax equation, we need to evaluate the derivative of the activation function with respect to the activation \( z_i^l \), that is we need
+$$ +\frac{\partial f(z_i^l)}{\partial w_{jk}^l} = +\frac{\partial f(z_i^l)}{\partial z_j^l} \frac{\partial z_j^l}{\partial w_{jk}^l}= \frac{\partial f(z_i^l)}{\partial z_j^l}a_k^{l-1}. +$$ -One can identify a set of key steps when using neural networks to solve supervised learning problems:
+For the Softmax function we have
+$$ +f(z_i^l) = \frac{\exp{(z_i^l)}}{\sum_{m=1}^K\exp{(z_m^l)}}. +$$ + +Its derivative with respect to \( z_j^l \) gives
+$$ +\frac{\partial f(z_i^l)}{\partial z_j^l}= f(z_i^l)\left(\delta_{ij}-f(z_j^l)\right), +$$ + +which in case of the simply binary model reduces to having \( i=j \).
-diff --git a/doc/pub/week42/html/._week42-bs074.html b/doc/pub/week42/html/._week42-bs074.html index 499922228..0ad12f712 100644 --- a/doc/pub/week42/html/._week42-bs074.html +++ b/doc/pub/week42/html/._week42-bs074.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
- -
Here we will be using the MNIST dataset, which is readily available through the scikit-learn -package. You may also find it for example here. -The MNIST (Modified National Institute of Standards and Technology) database is a large database -of handwritten digits that is commonly used for training various image processing systems. -The MNIST dataset consists of 70 000 images of size \( 28\times 28 \) pixels, each labeled from 0 to 9. -The scikit-learn dataset we will use consists of a selection of 1797 images of size \( 8\times 8 \) collected and processed from this database. -
- -To feed data into a feed-forward neural network we need to represent -the inputs as a design/feature matrix \( X = (n_{inputs}, n_{features}) \). Each -row represents an input, in this case a handwritten digit, and -each column represents a feature, in this case a pixel. The -correct answers, also known as labels or targets are -represented as a 1D array of integers -\( Y = (n_{inputs}) = (5, 3, 1, 8,...) \). -
- -As an example, say we want to build a neural network using supervised learning to predict Body-Mass Index (BMI) from -measurements of height (in m) -and weight (in kg). If we have measurements of 5 people the design/feature matrix could be for example: -
- -$$ X = \begin{bmatrix} -1.85 & 81\\ -1.71 & 65\\ -1.95 & 103\\ -1.55 & 42\\ -1.63 & 56 -\end{bmatrix} ,$$ -
- -and the targets would be:
- -$$ Y = (23.7, 22.2, 27.1, 17.5, 21.1) $$
- -Since each input image is a 2D matrix, we need to flatten the image -(i.e. "unravel" the 2D matrix into a 1D array) to turn the data into a -design/feature matrix. This means we lose all spatial information in the -image, such as locality and translational invariance. More complicated -architectures such as Convolutional Neural Networks can take advantage -of such information, and are most commonly applied when analyzing -images. -
- - - -# import necessary packages
-import numpy as np
-import matplotlib.pyplot as plt
-from sklearn import datasets
-
-
-# ensure the same random numbers appear every time
-np.random.seed(0)
-
-# display images in notebook
-%matplotlib inline
-plt.rcParams['figure.figsize'] = (12,12)
-
-
-# download MNIST dataset
-digits = datasets.load_digits()
-
-# define inputs and labels
-inputs = digits.images
-labels = digits.target
-
-print("inputs = (n_inputs, pixel_width, pixel_height) = " + str(inputs.shape))
-print("labels = (n_inputs) = " + str(labels.shape))
-
-
-# flatten the image
-# the value -1 means dimension is inferred from the remaining dimensions: 8x8 = 64
-n_inputs = len(inputs)
-inputs = inputs.reshape(n_inputs, -1)
-print("X = (n_inputs, n_features) = " + str(inputs.shape))
-
-
-# choose some random images to display
-indices = np.arange(n_inputs)
-random_indices = np.random.choice(indices, size=5)
-
-for i, image in enumerate(digits.images[random_indices]):
- plt.subplot(1, 5, i+1)
- plt.axis('off')
- plt.imshow(image, cmap=plt.cm.gray_r, interpolation='nearest')
- plt.title("Label: %d" % digits.target[random_indices[i]])
-plt.show()
-
-One can identify a set of key steps when using neural networks to solve supervised learning problems:
+diff --git a/doc/pub/week42/html/._week42-bs075.html b/doc/pub/week42/html/._week42-bs075.html index 1288b7b87..2d8b05efd 100644 --- a/doc/pub/week42/html/._week42-bs075.html +++ b/doc/pub/week42/html/._week42-bs075.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
-
Performing analysis before partitioning the dataset is a major error, that can lead to incorrect conclusions.
+Here we will be using the MNIST dataset, which is readily available through the scikit-learn +package. You may also find it for example here. +The MNIST (Modified National Institute of Standards and Technology) database is a large database +of handwritten digits that is commonly used for training various image processing systems. +The MNIST dataset consists of 70 000 images of size \( 28\times 28 \) pixels, each labeled from 0 to 9. +The scikit-learn dataset we will use consists of a selection of 1797 images of size \( 8\times 8 \) collected and processed from this database. +
-We will reserve \( 80 \% \) of our dataset for training and \( 20 \% \) for testing.
+To feed data into a feed-forward neural network we need to represent +the inputs as a design/feature matrix \( X = (n_{inputs}, n_{features}) \). Each +row represents an input, in this case a handwritten digit, and +each column represents a feature, in this case a pixel. The +correct answers, also known as labels or targets are +represented as a 1D array of integers +\( Y = (n_{inputs}) = (5, 3, 1, 8,...) \). +
-It is important that the train and test datasets are drawn randomly from our dataset, to ensure -no bias in the sampling. -Say you are taking measurements of weather data to predict the weather in the coming 5 days. -You don't want to train your model on measurements taken from the hours 00.00 to 12.00, and then test it on data -collected from 12.00 to 24.00. +
As an example, say we want to build a neural network using supervised learning to predict Body-Mass Index (BMI) from +measurements of height (in m) +and weight (in kg). If we have measurements of 5 people the design/feature matrix could be for example: +
+ +$$ X = \begin{bmatrix} +1.85 & 81\\ +1.71 & 65\\ +1.95 & 103\\ +1.55 & 42\\ +1.63 & 56 +\end{bmatrix} ,$$ +
+ +and the targets would be:
+ +$$ Y = (23.7, 22.2, 27.1, 17.5, 21.1) $$
+ +Since each input image is a 2D matrix, we need to flatten the image +(i.e. "unravel" the 2D matrix into a 1D array) to turn the data into a +design/feature matrix. This means we lose all spatial information in the +image, such as locality and translational invariance. More complicated +architectures such as Convolutional Neural Networks can take advantage +of such information, and are most commonly applied when analyzing +images.
@@ -477,33 +517,48 @@ collected from 12.00 to 24.00.from sklearn.model_selection import train_test_split
+ # import necessary packages
+import numpy as np
+import matplotlib.pyplot as plt
+from sklearn import datasets
-# one-liner from scikit-learn library
-train_size = 0.8
-test_size = 1 - train_size
-X_train, X_test, Y_train, Y_test = train_test_split(inputs, labels, train_size=train_size,
- test_size=test_size)
-# equivalently in numpy
-def train_test_split_numpy(inputs, labels, train_size, test_size):
- n_inputs = len(inputs)
- inputs_shuffled = inputs.copy()
- labels_shuffled = labels.copy()
-
- np.random.shuffle(inputs_shuffled)
- np.random.shuffle(labels_shuffled)
-
- train_end = int(n_inputs*train_size)
- X_train, X_test = inputs_shuffled[:train_end], inputs_shuffled[train_end:]
- Y_train, Y_test = labels_shuffled[:train_end], labels_shuffled[train_end:]
-
- return X_train, X_test, Y_train, Y_test
+# ensure the same random numbers appear every time
+np.random.seed(0)
-#X_train, X_test, Y_train, Y_test = train_test_split_numpy(inputs, labels, train_size, test_size)
+# display images in notebook
+%matplotlib inline
+plt.rcParams['figure.figsize'] = (12,12)
-print("Number of training images: " + str(len(X_train)))
-print("Number of test images: " + str(len(X_test)))
+
+# download MNIST dataset
+digits = datasets.load_digits()
+
+# define inputs and labels
+inputs = digits.images
+labels = digits.target
+
+print("inputs = (n_inputs, pixel_width, pixel_height) = " + str(inputs.shape))
+print("labels = (n_inputs) = " + str(labels.shape))
+
+
+# flatten the image
+# the value -1 means dimension is inferred from the remaining dimensions: 8x8 = 64
+n_inputs = len(inputs)
+inputs = inputs.reshape(n_inputs, -1)
+print("X = (n_inputs, n_features) = " + str(inputs.shape))
+
+
+# choose some random images to display
+indices = np.arange(n_inputs)
+random_indices = np.random.choice(indices, size=5)
+
+for i, image in enumerate(digits.images[random_indices]):
+ plt.subplot(1, 5, i+1)
+ plt.axis('off')
+ plt.imshow(image, cmap=plt.cm.gray_r, interpolation='nearest')
+ plt.title("Label: %d" % digits.target[random_indices[i]])
+plt.show()
-
Our simple feed-forward neural network will consist of an input layer, a single hidden layer and an output layer. The activation \( y \) of each neuron is a weighted sum of inputs, passed through an activation function. In case of the simple perceptron model we have
+Performing analysis before partitioning the dataset is a major error, that can lead to incorrect conclusions.
-$$ z = \sum_{i=1}^n w_i a_i ,$$
+We will reserve \( 80 \% \) of our dataset for training and \( 20 \% \) for testing.
-$$ y = f(z) ,$$
- -where \( f \) is the activation function, \( a_i \) represents input from neuron \( i \) in the preceding layer -and \( w_i \) is the weight to input \( i \). -The activation of the neurons in the input layer is just the features (e.g. a pixel value). +
It is important that the train and test datasets are drawn randomly from our dataset, to ensure +no bias in the sampling. +Say you are taking measurements of weather data to predict the weather in the coming 5 days. +You don't want to train your model on measurements taken from the hours 00.00 to 12.00, and then test it on data +collected from 12.00 to 24.00.
-The simplest activation function for a neuron is the Heaviside function:
-$$ f(z) = -\begin{cases} -1, & z > 0\\ -0, & \text{otherwise} -\end{cases} -$$ -
+ +from sklearn.model_selection import train_test_split
-A feed-forward neural network with this activation is known as a perceptron.
-For a binary classifier (i.e. two classes, 0 or 1, dog or not-dog) we can also use this in our output layer.
-This activation can be generalized to \( k \) classes (using e.g. the one-against-all strategy),
-and we call these architectures multiclass perceptrons.
-
+# one-liner from scikit-learn library
+train_size = 0.8
+test_size = 1 - train_size
+X_train, X_test, Y_train, Y_test = train_test_split(inputs, labels, train_size=train_size,
+ test_size=test_size)
-However, it is now common to use the terms Single Layer Perceptron (SLP) (1 hidden layer) and
-Multilayer Perceptron (MLP) (2 or more hidden layers) to refer to feed-forward neural networks with any activation function.
-
+# equivalently in numpy
+def train_test_split_numpy(inputs, labels, train_size, test_size):
+ n_inputs = len(inputs)
+ inputs_shuffled = inputs.copy()
+ labels_shuffled = labels.copy()
+
+ np.random.shuffle(inputs_shuffled)
+ np.random.shuffle(labels_shuffled)
+
+ train_end = int(n_inputs*train_size)
+ X_train, X_test = inputs_shuffled[:train_end], inputs_shuffled[train_end:]
+ Y_train, Y_test = labels_shuffled[:train_end], labels_shuffled[train_end:]
+
+ return X_train, X_test, Y_train, Y_test
-Typical choices for activation functions include the sigmoid function, hyperbolic tangent, and Rectified Linear Unit (ReLU).
-We will be using the sigmoid function \( \sigma(x) \):
-
+#X_train, X_test, Y_train, Y_test = train_test_split_numpy(inputs, labels, train_size, test_size)
-$$ f(x) = \sigma(x) = \frac{1}{1 + e^{-x}} ,$$
+print("Number of training images: " + str(len(X_train)))
+print("Number of test images: " + str(len(X_test)))
+
+which is inspired by probability theory (see logistic regression) and was most commonly used until about 2011. See the discussion below concerning other activation functions.
@@ -523,7 +552,7 @@ We will be using the sigmoid function \( \sigma(x) \):
- -
Since each input image has 8x8 = 64 pixels or features, we have an input layer of 64 neurons.
+Our simple feed-forward neural network will consist of an input layer, a single hidden layer and an output layer. The activation \( y \) of each neuron is a weighted sum of inputs, passed through an activation function. In case of the simple perceptron model we have
-We will use 50 neurons in the hidden layer receiving input from the neurons in the input layer. -Since each neuron in the hidden layer is connected to the 64 inputs we have 64x50 = 3200 weights to the hidden layer. +
$$ z = \sum_{i=1}^n w_i a_i ,$$
+ +$$ y = f(z) ,$$
+ +where \( f \) is the activation function, \( a_i \) represents input from neuron \( i \) in the preceding layer +and \( w_i \) is the weight to input \( i \). +The activation of the neurons in the input layer is just the features (e.g. a pixel value).
-If we were building a binary classifier, it would be sufficient with a single neuron in the output layer, -which could output 0 or 1 according to the Heaviside function. This would be an example of a hard classifier, meaning it outputs the class of the input directly. However, if we are dealing with noisy data it is often beneficial to use a soft classifier, which outputs the probability of being in class 0 or 1. +
The simplest activation function for a neuron is the Heaviside function:
+ +$$ f(z) = +\begin{cases} +1, & z > 0\\ +0, & \text{otherwise} +\end{cases} +$$
-For a soft binary classifier, we could use a single neuron and interpret the output as either being the probability of being in class 0 or the probability of being in class 1. Alternatively we could use 2 neurons, and interpret each neuron as the probability of being in each class.
- -Since we are doing multiclass classification, with 10 categories, it is natural to use 10 neurons in the output layer. We number the neurons \( j = 0,1,...,9 \). The activation of each output neuron \( j \) will be according to the softmax function:
- -$$ P(\text{class \( j \)} \mid \text{input \( \boldsymbol{a} \)}) = \frac{\exp{(\boldsymbol{a}^T \boldsymbol{w}_j)}} -{\sum_{c=0}^{9} \exp{(\boldsymbol{a}^T \boldsymbol{w}_c)}} ,$$ +
A feed-forward neural network with this activation is known as a perceptron. +For a binary classifier (i.e. two classes, 0 or 1, dog or not-dog) we can also use this in our output layer. +This activation can be generalized to \( k \) classes (using e.g. the one-against-all strategy), +and we call these architectures multiclass perceptrons.
-i.e. each neuron \( j \) outputs the probability of being in class \( j \) given an input from the hidden layer \( \boldsymbol{a} \), with \( \boldsymbol{w}_j \) the weights of neuron \( j \) to the inputs. -The denominator is a normalization factor to ensure the outputs (probabilities) sum up to 1. -The exponent is just the weighted sum of inputs as before: +
However, it is now common to use the terms Single Layer Perceptron (SLP) (1 hidden layer) and +Multilayer Perceptron (MLP) (2 or more hidden layers) to refer to feed-forward neural networks with any activation function.
-$$ z_j = \sum_{i=1}^n w_ {ij} a_i+b_j.$$
- -Since each neuron in the output layer is connected to the 50 inputs from the hidden layer we have 50x10 = 500 -weights to the output layer. +
Typical choices for activation functions include the sigmoid function, hyperbolic tangent, and Rectified Linear Unit (ReLU). +We will be using the sigmoid function \( \sigma(x) \):
+$$ f(x) = \sigma(x) = \frac{1}{1 + e^{-x}} ,$$
+ +which is inspired by probability theory (see logistic regression) and was most commonly used until about 2011. See the discussion below concerning other activation functions.
+diff --git a/doc/pub/week42/html/._week42-bs078.html b/doc/pub/week42/html/._week42-bs078.html index c7264104f..22d0a88ad 100644 --- a/doc/pub/week42/html/._week42-bs078.html +++ b/doc/pub/week42/html/._week42-bs078.html @@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d 2, None, 'setting-up-the-equations-for-a-neural-network'), - ('Layout of a neural network with three hidden layers', + ('Layout of a neural network with three hidden layers (last ' + 'later = $l=L=4$, first layer $l=0$)', 2, None, - 'layout-of-a-neural-network-with-three-hidden-layers'), + 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'), ('Definitions', 2, None, 'definitions'), ('Inputs to the activation function', 2, None, 'inputs-to-the-activation-function'), + ('Layout of input to first hidden layer $l=1$ from input layer ' + '$l=0$', + 2, + None, + 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'), ('Derivatives and the chain rule', 2, None, @@ -369,83 +375,84 @@ MathJax.Hub.Config({
-
Typically weights are initialized with small values distributed around zero, drawn from a uniform -or normal distribution. Setting all weights to zero means all neurons give the same output, making the network useless. +
Since each input image has 8x8 = 64 pixels or features, we have an input layer of 64 neurons.
+ +We will use 50 neurons in the hidden layer receiving input from the neurons in the input layer. +Since each neuron in the hidden layer is connected to the 64 inputs we have 64x50 = 3200 weights to the hidden layer.
-Adding a bias value to the weighted sum of inputs allows the neural network to represent a greater range -of values. Without it, any input with the value 0 will be mapped to zero (before being passed through the activation). The bias unit has an output of 1, and a weight to each neuron \( j \), \( b_j \): +
If we were building a binary classifier, it would be sufficient with a single neuron in the output layer, +which could output 0 or 1 according to the Heaviside function. This would be an example of a hard classifier, meaning it outputs the class of the input directly. However, if we are dealing with noisy data it is often beneficial to use a soft classifier, which outputs the probability of being in class 0 or 1.
-$$ z_j = \sum_{i=1}^n w_ {ij} a_i + b_j.$$
+For a soft binary classifier, we could use a single neuron and interpret the output as either being the probability of being in class 0 or the probability of being in class 1. Alternatively we could use 2 neurons, and interpret each neuron as the probability of being in each class.
-The bias weights \( \boldsymbol{b} \) are often initialized to zero, but a small value like \( 0.01 \) ensures all neurons have some output which can be backpropagated in the first training cycle.
+Since we are doing multiclass classification, with 10 categories, it is natural to use 10 neurons in the output layer. We number the neurons \( j = 0,1,...,9 \). The activation of each output neuron \( j \) will be according to the softmax function:
- -# building our neural network
+$$ P(\text{class \( j \)} \mid \text{input \( \boldsymbol{a} \)}) = \frac{\exp{(\boldsymbol{a}^T \boldsymbol{w}_j)}}
+{\sum_{c=0}^{9} \exp{(\boldsymbol{a}^T \boldsymbol{w}_c)}} ,$$
+
-n_inputs, n_features = X_train.shape
-n_hidden_neurons = 50
-n_categories = 10
+i.e. each neuron \( j \) outputs the probability of being in class \( j \) given an input from the hidden layer \( \boldsymbol{a} \), with \( \boldsymbol{w}_j \) the weights of neuron \( j \) to the inputs.
+The denominator is a normalization factor to ensure the outputs (probabilities) sum up to 1.
+The exponent is just the weighted sum of inputs as before:
+
-# we make the weights normally distributed using numpy.random.randn
-
-# weights and bias in the hidden layer
-hidden_weights = np.random.randn(n_features, n_hidden_neurons)
-hidden_bias = np.zeros(n_hidden_neurons) + 0.01
-
-# weights and bias in the output layer
-output_weights = np.random.randn(n_hidden_neurons, n_categories)
-output_bias = np.zeros(n_categories) + 0.01
-
-$$ z_j = \sum_{i=1}^n w_ {ij} a_i+b_j.$$
+Since each neuron in the output layer is connected to the 50 inputs from the hidden layer we have 50x10 = 500 +weights to the output layer. +
@@ -533,7 +529,7 @@ output_bias = np87
Denote \( F \) the number of features, \( H \) the number of hidden neurons and \( C \) the number of categories.
-For each input image we calculate a weighted sum of input features (pixel values) to each neuron \( j \) in the hidden layer \( l \):
+ Typically weights are initialized with small values distributed around zero, drawn from a uniform
+or normal distribution. Setting all weights to zero means all neurons give the same output, making the network useless.
$$ z_{j}^{l} = \sum_{i=1}^{F} w_{ij}^{l} x_i + b_{j}^{l},$$ this is then passed through our activation function $$ a_{j}^{l} = f(z_{j}^{l}) .$$ We calculate a weighted sum of inputs (activations in the hidden layer) to each neuron \( j \) in the output layer: $$ z_{j}^{L} = \sum_{i=1}^{H} w_{ij}^{L} a_{i}^{l} + b_{j}^{L}.$$ Finally we calculate the output of neuron \( j \) in the output layer using the softmax function: $$ a_{j}^{L} = \frac{\exp{(z_j^{L})}}
-{\sum_{c=0}^{C-1} \exp{(z_c^{L})}} .$$
+ Adding a bias value to the weighted sum of inputs allows the neural network to represent a greater range
+of values. Without it, any input with the value 0 will be mapped to zero (before being passed through the activation). The bias unit has an output of 1, and a weight to each neuron \( j \), \( b_j \):
$$ z_j = \sum_{i=1}^n w_ {ij} a_i + b_j.$$ The bias weights \( \boldsymbol{b} \) are often initialized to zero, but a small value like \( 0.01 \) ensures all neurons have some output which can be backpropagated in the first training cycle.
Since our data has the dimensions \( X = (n_{inputs}, n_{features}) \) and our weights to the hidden
-layer have the dimensions
-\( W_{hidden} = (n_{features}, n_{hidden}) \),
-we can easily feed the network all our training data in one go by taking the matrix product
+ Denote \( F \) the number of features, \( H \) the number of hidden neurons and \( C \) the number of categories.
+For each input image we calculate a weighted sum of input features (pixel values) to each neuron \( j \) in the hidden layer \( l \):
$$ X W^{h} = (n_{inputs}, n_{hidden}),$$ $$ z_{j}^{l} = \sum_{i=1}^{F} w_{ij}^{l} x_i + b_{j}^{l},$$ and obtain a matrix that holds the weighted sum of inputs to the hidden layer
-for each input image and each hidden neuron.
-We also add the bias to obtain a matrix of weighted sums to the hidden layer \( Z^{h} \):
+ this is then passed through our activation function $$ a_{j}^{l} = f(z_{j}^{l}) .$$ We calculate a weighted sum of inputs (activations in the hidden layer) to each neuron \( j \) in the output layer: $$ z_{j}^{L} = \sum_{i=1}^{H} w_{ij}^{L} a_{i}^{l} + b_{j}^{L}.$$ Finally we calculate the output of neuron \( j \) in the output layer using the softmax function: $$ a_{j}^{L} = \frac{\exp{(z_j^{L})}}
+{\sum_{c=0}^{C-1} \exp{(z_c^{L})}} .$$
$$ \boldsymbol{z}^{l} = \boldsymbol{X} \boldsymbol{W}^{l} + \boldsymbol{b}^{l} ,$$ meaning the same bias (1D array with size equal number of hidden neurons) is added to each input image.
-This is then passed through the activation:
- $$ \boldsymbol{a}^{l} = f(\boldsymbol{z}^l) .$$ This is fed to the output layer: $$ \boldsymbol{z}^{L} = \boldsymbol{a}^{L} \boldsymbol{W}^{L} + \boldsymbol{b}^{L} .$$ Finally we receive our output values for each image and each category by passing it through the softmax function: $$ output = softmax (\boldsymbol{z}^{L}) = (n_{inputs}, n_{categories}) .$$
To measure how well our neural network is doing we need to introduce a cost function.
-We will call the function that gives the error of a single sample output the loss function, and the function
-that gives the total error of our network across all samples the cost function.
-A typical choice for multiclass classification is the cross-entropy loss, also known as the negative log likelihood.
+ Since our data has the dimensions \( X = (n_{inputs}, n_{features}) \) and our weights to the hidden
+layer have the dimensions
+\( W_{hidden} = (n_{features}, n_{hidden}) \),
+we can easily feed the network all our training data in one go by taking the matrix product
In multiclass classification it is common to treat each integer label as a so called one-hot vector: $$ X W^{h} = (n_{inputs}, n_{hidden}),$$ $$ y = 5 \quad \rightarrow \quad \boldsymbol{y} = (0, 0, 0, 0, 0, 1, 0, 0, 0, 0) ,$$ $$ y = 1 \quad \rightarrow \quad \boldsymbol{y} = (0, 1, 0, 0, 0, 0, 0, 0, 0, 0) ,$$ i.e. a binary bit string of length \( C \), where \( C = 10 \) is the number of classes in the MNIST dataset. Let \( y_{ic} \) denote the \( c \)-th component of the \( i \)-th one-hot vector.
-We define the cost function \( \mathcal{C} \) as a sum over the cross-entropy loss for each point \( \boldsymbol{x}_i \) in the dataset.
+ and obtain a matrix that holds the weighted sum of inputs to the hidden layer
+for each input image and each hidden neuron.
+We also add the bias to obtain a matrix of weighted sums to the hidden layer \( Z^{h} \):
In the one-hot representation only one of the terms in the loss function is non-zero, namely the
-probability of the correct category \( c' \)
-(i.e. the category \( c' \) such that \( y_{ic'} = 1 \)). This means that the cross entropy loss only punishes you for how wrong
-you got the correct label. The probability of category \( c \) is given by the softmax function. The vector \( \boldsymbol{\theta} \) represents the parameters of our network, i.e. all the weights and biases.
+ $$ \boldsymbol{z}^{l} = \boldsymbol{X} \boldsymbol{W}^{l} + \boldsymbol{b}^{l} ,$$ meaning the same bias (1D array with size equal number of hidden neurons) is added to each input image.
+This is then passed through the activation:
$$ \boldsymbol{a}^{l} = f(\boldsymbol{z}^l) .$$ This is fed to the output layer: $$ \boldsymbol{z}^{L} = \boldsymbol{a}^{L} \boldsymbol{W}^{L} + \boldsymbol{b}^{L} .$$ Finally we receive our output values for each image and each category by passing it through the softmax function: $$ output = softmax (\boldsymbol{z}^{L}) = (n_{inputs}, n_{categories}) .$$
@@ -509,7 +578,7 @@ you got the correct label. The probability of category \( c \) is given by the s
The network is trained by finding the weights and biases that minimize the cost function. One of the most widely used classes of methods is gradient descent and its generalizations. The idea behind gradient descent
-is simply to adjust the weights in the direction where the gradient of the cost function is large and negative. This ensures we flow toward a local minimum of the cost function.
-Each parameter \( \theta \) is iteratively adjusted according to the rule
+ To measure how well our neural network is doing we need to introduce a cost function.
+We will call the function that gives the error of a single sample output the loss function, and the function
+that gives the total error of our network across all samples the cost function.
+A typical choice for multiclass classification is the cross-entropy loss, also known as the negative log likelihood.
$$ \theta_{i+1} = \theta_i - \eta \nabla \mathcal{C}(\theta_i) ,$$ In multiclass classification it is common to treat each integer label as a so called one-hot vector: where \( \eta \) is known as the learning rate, which controls how big a step we take towards the minimum.
-This update can be repeated for any number of iterations, or until we are satisfied with the result.
+ $$ y = 5 \quad \rightarrow \quad \boldsymbol{y} = (0, 0, 0, 0, 0, 1, 0, 0, 0, 0) ,$$ $$ y = 1 \quad \rightarrow \quad \boldsymbol{y} = (0, 1, 0, 0, 0, 0, 0, 0, 0, 0) ,$$ i.e. a binary bit string of length \( C \), where \( C = 10 \) is the number of classes in the MNIST dataset. Let \( y_{ic} \) denote the \( c \)-th component of the \( i \)-th one-hot vector.
+We define the cost function \( \mathcal{C} \) as a sum over the cross-entropy loss for each point \( \boldsymbol{x}_i \) in the dataset.
A simple and effective improvement is a variant called Batch Gradient Descent.
-Instead of calculating the gradient on the whole dataset, we calculate an approximation of the gradient
-on a subset of the data called a minibatch.
-If there are \( N \) data points and we have a minibatch size of \( M \), the total number of batches
-is \( N/M \).
-We denote each minibatch \( B_k \), with \( k = 1, 2,...,N/M \). The gradient then becomes:
+ In the one-hot representation only one of the terms in the loss function is non-zero, namely the
+probability of the correct category \( c' \)
+(i.e. the category \( c' \) such that \( y_{ic'} = 1 \)). This means that the cross entropy loss only punishes you for how wrong
+you got the correct label. The probability of category \( c \) is given by the softmax function. The vector \( \boldsymbol{\theta} \) represents the parameters of our network, i.e. all the weights and biases.
$$ \nabla \mathcal{C}(\theta) = \frac{1}{N} \sum_{i=1}^N \nabla \mathcal{L}_i(\theta) \quad \rightarrow \quad
-\frac{1}{M} \sum_{i \in B_k} \nabla \mathcal{L}_i(\theta) ,$$
- i.e. instead of averaging the loss over the entire dataset, we average over a minibatch. This has two important benefits: The various optmization methods, with codes and algorithms, are discussed in our lectures on Gradient descent approaches.
@@ -516,7 +516,7 @@ We denote each minibatch \( B_k \), with \( k = 1, 2,...,N/M \). The gradient th
It is common to add an extra term to the cost function, proportional
-to the size of the weights. This is equivalent to constraining the
-size of the weights, so that they do not grow out of control.
-Constraining the size of the weights means that the weights cannot
-grow arbitrarily large to fit the training data, and in this way
-reduces overfitting.
+ The network is trained by finding the weights and biases that minimize the cost function. One of the most widely used classes of methods is gradient descent and its generalizations. The idea behind gradient descent
+is simply to adjust the weights in the direction where the gradient of the cost function is large and negative. This ensures we flow toward a local minimum of the cost function.
+Each parameter \( \theta \) is iteratively adjusted according to the rule
We will measure the size of the weights using the so called L2-norm, meaning our cost function becomes: $$ \theta_{i+1} = \theta_i - \eta \nabla \mathcal{C}(\theta_i) ,$$ $$ \mathcal{C}(\theta) = \frac{1}{N} \sum_{i=1}^N \mathcal{L}_i(\theta) \quad \rightarrow \quad
-\frac{1}{N} \sum_{i=1}^N \mathcal{L}_i(\theta) + \lambda \lvert \lvert \boldsymbol{w} \rvert \rvert_2^2
-= \frac{1}{N} \sum_{i=1}^N \mathcal{L}(\theta) + \lambda \sum_{ij} w_{ij}^2,$$
+ where \( \eta \) is known as the learning rate, which controls how big a step we take towards the minimum.
+This update can be repeated for any number of iterations, or until we are satisfied with the result.
i.e. we sum up all the weights squared. The factor \( \lambda \) is known as a regularization parameter. In order to train the model, we need to calculate the derivative of
-the cost function with respect to every bias and weight in the
-network. In total our network has \( (64 + 1)\times 50=3250 \) weights in
-the hidden layer and \( (50 + 1)\times 10=510 \) weights to the output
-layer (\( +1 \) for the bias), and the gradient must be calculated for
-every parameter. We use the backpropagation algorithm discussed
-above. This is a clever use of the chain rule that allows us to
-calculate the gradient efficently.
+ A simple and effective improvement is a variant called Batch Gradient Descent.
+Instead of calculating the gradient on the whole dataset, we calculate an approximation of the gradient
+on a subset of the data called a minibatch.
+If there are \( N \) data points and we have a minibatch size of \( M \), the total number of batches
+is \( N/M \).
+We denote each minibatch \( B_k \), with \( k = 1, 2,...,N/M \). The gradient then becomes:
$$ \nabla \mathcal{C}(\theta) = \frac{1}{N} \sum_{i=1}^N \nabla \mathcal{L}_i(\theta) \quad \rightarrow \quad
+\frac{1}{M} \sum_{i \in B_k} \nabla \mathcal{L}_i(\theta) ,$$
+ i.e. instead of averaging the loss over the entire dataset, we average over a minibatch. This has two important benefits: The various optmization methods, with codes and algorithms, are discussed in our lectures on Gradient descent approaches.
To more efficently train our network these equations are implemented using matrix operations.
-The error in the output layer is calculated simply as, with \( \boldsymbol{t} \) being our targets,
+ It is common to add an extra term to the cost function, proportional
+to the size of the weights. This is equivalent to constraining the
+size of the weights, so that they do not grow out of control.
+Constraining the size of the weights means that the weights cannot
+grow arbitrarily large to fit the training data, and in this way
+reduces overfitting.
$$ \delta_L = \boldsymbol{t} - \boldsymbol{y} = (n_{inputs}, n_{categories}) .$$ We will measure the size of the weights using the so called L2-norm, meaning our cost function becomes: The gradient for the output weights is calculated as $$ \nabla W_{L} = \boldsymbol{a}^T \delta_L = (n_{hidden}, n_{categories}) ,$$ where \( \boldsymbol{a} = (n_{inputs}, n_{hidden}) \). This simply means that we are summing up the gradients for each input.
-Since we are going backwards we have to transpose the activation matrix.
+ $$ \mathcal{C}(\theta) = \frac{1}{N} \sum_{i=1}^N \mathcal{L}_i(\theta) \quad \rightarrow \quad
+\frac{1}{N} \sum_{i=1}^N \mathcal{L}_i(\theta) + \lambda \lvert \lvert \boldsymbol{w} \rvert \rvert_2^2
+= \frac{1}{N} \sum_{i=1}^N \mathcal{L}(\theta) + \lambda \sum_{ij} w_{ij}^2,$$
The gradient with respect to the output bias is then i.e. we sum up all the weights squared. The factor \( \lambda \) is known as a regularization parameter. $$ \nabla \boldsymbol{b}_{L} = \sum_{i=1}^{n_{inputs}} \delta_L = (n_{categories}) .$$ The error in the hidden layer is $$ \Delta_h = \delta_L W_{L}^T \circ f'(z_{h}) = \delta_L W_{L}^T \circ a_{h} \circ (1 - a_{h}) = (n_{inputs}, n_{hidden}) ,$$ where \( f'(a_{h}) \) is the derivative of the activation in the hidden layer. The matrix products mean
-that we are summing up the products for each neuron in the output layer. The symbol \( \circ \) denotes
-the Hadamard product, meaning element-wise multiplication.
+ In order to train the model, we need to calculate the derivative of
+the cost function with respect to every bias and weight in the
+network. In total our network has \( (64 + 1)\times 50=3250 \) weights in
+the hidden layer and \( (50 + 1)\times 10=510 \) weights to the output
+layer (\( +1 \) for the bias), and the gradient must be calculated for
+every parameter. We use the backpropagation algorithm discussed
+above. This is a clever use of the chain rule that allows us to
+calculate the gradient efficently.
This again gives us the gradients in the hidden layer: $$ \nabla W_{h} = X^T \delta_h = (n_{features}, n_{hidden}) ,$$ $$ \nabla b_{h} = \sum_{i=1}^{n_{inputs}} \delta_h = (n_{hidden}) .$$
As we can see the network does not seem to be learning at all. It seems to be just guessing the label for each image.
-In order to obtain a network that does something useful, we will have to do a bit more work.
+ To more efficently train our network these equations are implemented using matrix operations.
+The error in the output layer is calculated simply as, with \( \boldsymbol{t} \) being our targets,
The choice of hyperparameters such as learning rate and regularization parameter is hugely influential for the performance of the network. Typically a grid-search is performed, wherein we test different hyperparameters separated by orders of magnitude. For example we could test the learning rates \( \eta = 10^{-6}, 10^{-5},...,10^{-1} \) with different regularization parameters \( \lambda = 10^{-6},...,10^{-0} \). $$ \delta_L = \boldsymbol{t} - \boldsymbol{y} = (n_{inputs}, n_{categories}) .$$ Next, we haven't implemented minibatching yet, which introduces stochasticity and is though to act as an important regularizer on the weights. We call a feed-forward + backward pass with a minibatch an iteration, and a full training period
-going through the entire dataset (\( n/M \) batches) an epoch.
+ The gradient for the output weights is calculated as $$ \nabla W_{L} = \boldsymbol{a}^T \delta_L = (n_{hidden}, n_{categories}) ,$$ where \( \boldsymbol{a} = (n_{inputs}, n_{hidden}) \). This simply means that we are summing up the gradients for each input.
+Since we are going backwards we have to transpose the activation matrix.
If this does not improve network performance, you may want to consider altering the network architecture, adding more neurons or hidden layers.
-Andrew Ng goes through some of these considerations in this video. You can find a summary of the video here.
+ The gradient with respect to the output bias is then $$ \nabla \boldsymbol{b}_{L} = \sum_{i=1}^{n_{inputs}} \delta_L = (n_{categories}) .$$ The error in the hidden layer is $$ \Delta_h = \delta_L W_{L}^T \circ f'(z_{h}) = \delta_L W_{L}^T \circ a_{h} \circ (1 - a_{h}) = (n_{inputs}, n_{hidden}) ,$$ where \( f'(a_{h}) \) is the derivative of the activation in the hidden layer. The matrix products mean
+that we are summing up the products for each neuron in the output layer. The symbol \( \circ \) denotes
+the Hadamard product, meaning element-wise multiplication.
This again gives us the gradients in the hidden layer: $$ \nabla W_{h} = X^T \delta_h = (n_{features}, n_{hidden}) ,$$ $$ \nabla b_{h} = \sum_{i=1}^{n_{inputs}} \delta_h = (n_{hidden}) .$$
It is very natural to think of the network as an object, with specific instances of the network
-being realizations of this object with different hyperparameters. An implementation using Python classes provides a clean structure and interface, and the full implementation of our neural network is given below.
+ As we can see the network does not seem to be learning at all. It seems to be just guessing the label for each image.
+In order to obtain a network that does something useful, we will have to do a bit more work.
The choice of hyperparameters such as learning rate and regularization parameter is hugely influential for the performance of the network. Typically a grid-search is performed, wherein we test different hyperparameters separated by orders of magnitude. For example we could test the learning rates \( \eta = 10^{-6}, 10^{-5},...,10^{-1} \) with different regularization parameters \( \lambda = 10^{-6},...,10^{-0} \). Next, we haven't implemented minibatching yet, which introduces stochasticity and is though to act as an important regularizer on the weights. We call a feed-forward + backward pass with a minibatch an iteration, and a full training period
+going through the entire dataset (\( n/M \) batches) an epoch.
+ If this does not improve network performance, you may want to consider altering the network architecture, adding more neurons or hidden layers.
+Andrew Ng goes through some of these considerations in this video. You can find a summary of the video here.
+
@@ -610,7 +505,7 @@ being realizations of this object with different hyperparameters. An implementat
To measure the performance of our network we evaluate how well it does it data it has never seen before, i.e. the test data.
-We measure the performance of the network using the accuracy score.
-The accuracy is as you would expect just the number of images correctly labeled divided by the total number of images. A perfect classifier will have an accuracy score of \( 1 \).
+ It is very natural to think of the network as an object, with specific instances of the network
+being realizations of this object with different hyperparameters. An implementation using Python classes provides a clean structure and interface, and the full implementation of our neural network is given below.
$$ \text{Accuracy} = \frac{\sum_{i=1}^n I(\tilde{y}_i = y_i)}{n} ,$$ where \( I \) is the indicator function, \( 1 \) if \( \tilde{y}_i = y_i \) and \( 0 \) otherwise. We now perform a grid search to find the optimal hyperparameters for the network.
-Note that we are only using 1 layer with 50 neurons, and human performance is estimated to be around \( 98\% \) (\( 2\% \) error rate).
+ To measure the performance of our network we evaluate how well it does it data it has never seen before, i.e. the test data.
+We measure the performance of the network using the accuracy score.
+The accuracy is as you would expect just the number of images correctly labeled divided by the total number of images. A perfect classifier will have an accuracy score of \( 1 \).
$$ \text{Accuracy} = \frac{\sum_{i=1}^n I(\tilde{y}_i = y_i)}{n} ,$$ where \( I \) is the indicator function, \( 1 \) if \( \tilde{y}_i = y_i \) and \( 0 \) otherwise. We now perform a grid search to find the optimal hyperparameters for the network.
+Note that we are only using 1 layer with 50 neurons, and human performance is estimated to be around \( 98\% \) (\( 2\% \) error rate).
+ scikit-learn focuses more
-on traditional machine learning methods, such as regression,
-clustering, decision trees, etc. As such, it has only two types of
-neural networks: Multi Layer Perceptron outputting continuous values,
-MPLRegressor, and Multi Layer Perceptron outputting labels,
-MLPClassifier. We will see how simple it is to use these classes.
- scikit-learn implements a few improvements from our neural network,
-such as early stopping, a varying learning rate, different
-optimization methods, etc. We would therefore expect a better
-performance overall.
- scikit-learn focuses more
+on traditional machine learning methods, such as regression,
+clustering, decision trees, etc. As such, it has only two types of
+neural networks: Multi Layer Perceptron outputting continuous values,
+MPLRegressor, and Multi Layer Perceptron outputting labels,
+MLPClassifier. We will see how simple it is to use these classes.
+ scikit-learn implements a few improvements from our neural network,
+such as early stopping, a varying learning rate, different
+optimization methods, etc. We would therefore expect a better
+performance overall.
+ Now we want to build on the experience gained from our neural network implementation in NumPy and scikit-learn
-and use it to construct a neural network in Tensorflow. Once we have constructed a neural network in NumPy
-and Tensorflow, building one in Keras is really quite trivial, though the performance may suffer.
- In our previous example we used only one hidden layer, and in this we will use two. From this it should be quite
-clear how to build one using an arbitrary number of hidden layers, using data structures such as Python lists or
-NumPy arrays.
-
@@ -490,6 +543,7 @@ NumPy arrays.
Figure 1: Figure 1: Figure 1:
Feed-forward pass
+
+Weights and biases
-
+
+
+
+
+
+
+
+
+
+
+ # building our neural network
+
+n_inputs, n_features = X_train.shape
+n_hidden_neurons = 50
+n_categories = 10
+
+# we make the weights normally distributed using numpy.random.randn
+
+# weights and bias in the hidden layer
+hidden_weights = np.random.randn(n_features, n_hidden_neurons)
+hidden_bias = np.zeros(n_hidden_neurons) + 0.01
+
+# weights and bias in the output layer
+output_weights = np.random.randn(n_hidden_neurons, n_categories)
+output_bias = np.zeros(n_categories) + 0.01
+
+
+
+
+
+
+
+
+
@@ -504,7 +540,7 @@ For each input image we calculate a weighted sum of input features (pixel values
diff --git a/doc/pub/week42/html/._week42-bs080.html b/doc/pub/week42/html/._week42-bs080.html
index d6899189d..e9ea00896 100644
--- a/doc/pub/week42/html/._week42-bs080.html
+++ b/doc/pub/week42/html/._week42-bs080.html
@@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d
2,
None,
'setting-up-the-equations-for-a-neural-network'),
- ('Layout of a neural network with three hidden layers',
+ ('Layout of a neural network with three hidden layers (last '
+ 'later = $l=L=4$, first layer $l=0$)',
2,
None,
- 'layout-of-a-neural-network-with-three-hidden-layers'),
+ 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'),
('Definitions', 2, None, 'definitions'),
('Inputs to the activation function',
2,
None,
'inputs-to-the-activation-function'),
+ ('Layout of input to first hidden layer $l=1$ from input layer '
+ '$l=0$',
+ 2,
+ None,
+ 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'),
('Derivatives and the chain rule',
2,
None,
@@ -369,83 +375,84 @@ MathJax.Hub.Config({
Matrix multiplications
+
+Feed-forward pass
-
-
-
-
-
-
-
-
-
-
-
- # setup the feed-forward pass, subscript h = hidden layer
-
-def sigmoid(x):
- return 1/(1 + np.exp(-x))
-
-def feed_forward(X):
- # weighted sum of inputs to the hidden layer
- z_h = np.matmul(X, hidden_weights) + hidden_bias
- # activation in the hidden layer
- a_h = sigmoid(z_h)
-
- # weighted sum of inputs to the output layer
- z_o = np.matmul(a_h, output_weights) + output_bias
- # softmax output
- # axis 0 holds each input and axis 1 the probabilities of each category
- exp_term = np.exp(z_o)
- probabilities = exp_term / np.sum(exp_term, axis=1, keepdims=True)
-
- return probabilities
-
-probabilities = feed_forward(X_train)
-print("probabilities = (n_inputs, n_categories) = " + str(probabilities.shape))
-print("probability that image 0 is in category 0,1,2,...,9 = \n" + str(probabilities[0]))
-print("probabilities sum up to: " + str(probabilities[0].sum()))
-print()
-
-# we obtain a prediction by taking the class with the highest likelihood
-def predict(X):
- probabilities = feed_forward(X)
- return np.argmax(probabilities, axis=1)
-
-predictions = predict(X_train)
-print("predictions = (n_inputs) = " + str(predictions.shape))
-print("prediction for image 0: " + str(predictions[0]))
-print("correct label for image 0: " + str(Y_train[0]))
-
-
-
-
-
-
-
-
-
Choose cost function and optimizer
+
+Matrix multiplications
-
+
+
+
+
+
+
+
+
+
+ # setup the feed-forward pass, subscript h = hidden layer
+
+def sigmoid(x):
+ return 1/(1 + np.exp(-x))
+
+def feed_forward(X):
+ # weighted sum of inputs to the hidden layer
+ z_h = np.matmul(X, hidden_weights) + hidden_bias
+ # activation in the hidden layer
+ a_h = sigmoid(z_h)
+
+ # weighted sum of inputs to the output layer
+ z_o = np.matmul(a_h, output_weights) + output_bias
+ # softmax output
+ # axis 0 holds each input and axis 1 the probabilities of each category
+ exp_term = np.exp(z_o)
+ probabilities = exp_term / np.sum(exp_term, axis=1, keepdims=True)
+
+ return probabilities
+
+probabilities = feed_forward(X_train)
+print("probabilities = (n_inputs, n_categories) = " + str(probabilities.shape))
+print("probability that image 0 is in category 0,1,2,...,9 = \n" + str(probabilities[0]))
+print("probabilities sum up to: " + str(probabilities[0].sum()))
+print()
+
+# we obtain a prediction by taking the class with the highest likelihood
+def predict(X):
+ probabilities = feed_forward(X)
+ return np.argmax(probabilities, axis=1)
+
+predictions = predict(X_train)
+print("predictions = (n_inputs) = " + str(predictions.shape))
+print("prediction for image 0: " + str(predictions[0]))
+print("correct label for image 0: " + str(Y_train[0]))
+
+
+
+
+
+
+
+
+ Optimizing the cost function
+Choose cost function and optimizer
-
-
-
Regularization
+
+Optimizing the cost function
-
+
+
Matrix multiplication
+
+Regularization
-
-
-
-
-
-
-
-
-
-
-
- # to categorical turns our integer vector into a onehot representation
-from sklearn.metrics import accuracy_score
-
-# one-hot in numpy
-def to_categorical_numpy(integer_vector):
- n_inputs = len(integer_vector)
- n_categories = np.max(integer_vector) + 1
- onehot_vector = np.zeros((n_inputs, n_categories))
- onehot_vector[range(n_inputs), integer_vector] = 1
-
- return onehot_vector
-
-#Y_train_onehot, Y_test_onehot = to_categorical(Y_train), to_categorical(Y_test)
-Y_train_onehot, Y_test_onehot = to_categorical_numpy(Y_train), to_categorical_numpy(Y_test)
-
-def feed_forward_train(X):
- # weighted sum of inputs to the hidden layer
- z_h = np.matmul(X, hidden_weights) + hidden_bias
- # activation in the hidden layer
- a_h = sigmoid(z_h)
-
- # weighted sum of inputs to the output layer
- z_o = np.matmul(a_h, output_weights) + output_bias
- # softmax output
- # axis 0 holds each input and axis 1 the probabilities of each category
- exp_term = np.exp(z_o)
- probabilities = exp_term / np.sum(exp_term, axis=1, keepdims=True)
-
- # for backpropagation need activations in hidden and output layers
- return a_h, probabilities
-
-def backpropagation(X, Y):
- a_h, probabilities = feed_forward_train(X)
-
- # error in the output layer
- error_output = probabilities - Y
- # error in the hidden layer
- error_hidden = np.matmul(error_output, output_weights.T) * a_h * (1 - a_h)
-
- # gradients for the output layer
- output_weights_gradient = np.matmul(a_h.T, error_output)
- output_bias_gradient = np.sum(error_output, axis=0)
-
- # gradient for the hidden layer
- hidden_weights_gradient = np.matmul(X.T, error_hidden)
- hidden_bias_gradient = np.sum(error_hidden, axis=0)
-
- return output_weights_gradient, output_bias_gradient, hidden_weights_gradient, hidden_bias_gradient
-
-print("Old accuracy on training data: " + str(accuracy_score(predict(X_train), Y_train)))
-
-eta = 0.01
-lmbd = 0.01
-for i in range(1000):
- # calculate gradients
- dWo, dBo, dWh, dBh = backpropagation(X_train, Y_train_onehot)
-
- # regularization term gradients
- dWo += lmbd * output_weights
- dWh += lmbd * hidden_weights
-
- # update weights and biases
- output_weights -= eta * dWo
- output_bias -= eta * dBo
- hidden_weights -= eta * dWh
- hidden_bias -= eta * dBh
-
-print("New accuracy on training data: " + str(accuracy_score(predict(X_train), Y_train)))
-
-
-
-
-
-
-
-
- Improving performance
+Matrix multiplication
-
+
+
+
+
+
+
+
+
+
+
+ # to categorical turns our integer vector into a onehot representation
+from sklearn.metrics import accuracy_score
+
+# one-hot in numpy
+def to_categorical_numpy(integer_vector):
+ n_inputs = len(integer_vector)
+ n_categories = np.max(integer_vector) + 1
+ onehot_vector = np.zeros((n_inputs, n_categories))
+ onehot_vector[range(n_inputs), integer_vector] = 1
+
+ return onehot_vector
+
+#Y_train_onehot, Y_test_onehot = to_categorical(Y_train), to_categorical(Y_test)
+Y_train_onehot, Y_test_onehot = to_categorical_numpy(Y_train), to_categorical_numpy(Y_test)
+
+def feed_forward_train(X):
+ # weighted sum of inputs to the hidden layer
+ z_h = np.matmul(X, hidden_weights) + hidden_bias
+ # activation in the hidden layer
+ a_h = sigmoid(z_h)
+
+ # weighted sum of inputs to the output layer
+ z_o = np.matmul(a_h, output_weights) + output_bias
+ # softmax output
+ # axis 0 holds each input and axis 1 the probabilities of each category
+ exp_term = np.exp(z_o)
+ probabilities = exp_term / np.sum(exp_term, axis=1, keepdims=True)
+
+ # for backpropagation need activations in hidden and output layers
+ return a_h, probabilities
+
+def backpropagation(X, Y):
+ a_h, probabilities = feed_forward_train(X)
+
+ # error in the output layer
+ error_output = probabilities - Y
+ # error in the hidden layer
+ error_hidden = np.matmul(error_output, output_weights.T) * a_h * (1 - a_h)
+
+ # gradients for the output layer
+ output_weights_gradient = np.matmul(a_h.T, error_output)
+ output_bias_gradient = np.sum(error_output, axis=0)
+
+ # gradient for the hidden layer
+ hidden_weights_gradient = np.matmul(X.T, error_hidden)
+ hidden_bias_gradient = np.sum(error_hidden, axis=0)
+
+ return output_weights_gradient, output_bias_gradient, hidden_weights_gradient, hidden_bias_gradient
+
+print("Old accuracy on training data: " + str(accuracy_score(predict(X_train), Y_train)))
+
+eta = 0.01
+lmbd = 0.01
+for i in range(1000):
+ # calculate gradients
+ dWo, dBo, dWh, dBh = backpropagation(X_train, Y_train_onehot)
+
+ # regularization term gradients
+ dWo += lmbd * output_weights
+ dWh += lmbd * hidden_weights
+
+ # update weights and biases
+ output_weights -= eta * dWo
+ output_bias -= eta * dBo
+ hidden_weights -= eta * dWh
+ hidden_bias -= eta * dBh
+
+print("New accuracy on training data: " + str(accuracy_score(predict(X_train), Y_train)))
+
+
+
+
+
+
+
+
+
@@ -498,7 +615,7 @@ Andrew Ng goes through some of these considerations in this 94
diff --git a/doc/pub/week42/html/._week42-bs086.html b/doc/pub/week42/html/._week42-bs086.html
index 3b0d16fa2..0bfab2a0c 100644
--- a/doc/pub/week42/html/._week42-bs086.html
+++ b/doc/pub/week42/html/._week42-bs086.html
@@ -104,15 +104,21 @@ doconce format html week42.do.txt --html_style=bootstrap --pygments_html_style=d
2,
None,
'setting-up-the-equations-for-a-neural-network'),
- ('Layout of a neural network with three hidden layers',
+ ('Layout of a neural network with three hidden layers (last '
+ 'later = $l=L=4$, first layer $l=0$)',
2,
None,
- 'layout-of-a-neural-network-with-three-hidden-layers'),
+ 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'),
('Definitions', 2, None, 'definitions'),
('Inputs to the activation function',
2,
None,
'inputs-to-the-activation-function'),
+ ('Layout of input to first hidden layer $l=1$ from input layer '
+ '$l=0$',
+ 2,
+ None,
+ 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'),
('Derivatives and the chain rule',
2,
None,
@@ -369,83 +375,84 @@ MathJax.Hub.Config({
Full object-oriented implementation
+Improving performance
-
-
+
-
-
-
-
-
-
-
- class NeuralNetwork:
- def __init__(
- self,
- X_data,
- Y_data,
- n_hidden_neurons=50,
- n_categories=10,
- epochs=10,
- batch_size=100,
- eta=0.1,
- lmbd=0.0):
-
- self.X_data_full = X_data
- self.Y_data_full = Y_data
-
- self.n_inputs = X_data.shape[0]
- self.n_features = X_data.shape[1]
- self.n_hidden_neurons = n_hidden_neurons
- self.n_categories = n_categories
-
- self.epochs = epochs
- self.batch_size = batch_size
- self.iterations = self.n_inputs // self.batch_size
- self.eta = eta
- self.lmbd = lmbd
-
- self.create_biases_and_weights()
-
- def create_biases_and_weights(self):
- self.hidden_weights = np.random.randn(self.n_features, self.n_hidden_neurons)
- self.hidden_bias = np.zeros(self.n_hidden_neurons) + 0.01
-
- self.output_weights = np.random.randn(self.n_hidden_neurons, self.n_categories)
- self.output_bias = np.zeros(self.n_categories) + 0.01
-
- def feed_forward(self):
- # feed-forward for training
- self.z_h = np.matmul(self.X_data, self.hidden_weights) + self.hidden_bias
- self.a_h = sigmoid(self.z_h)
-
- self.z_o = np.matmul(self.a_h, self.output_weights) + self.output_bias
-
- exp_term = np.exp(self.z_o)
- self.probabilities = exp_term / np.sum(exp_term, axis=1, keepdims=True)
-
- def feed_forward_out(self, X):
- # feed-forward for output
- z_h = np.matmul(X, self.hidden_weights) + self.hidden_bias
- a_h = sigmoid(z_h)
-
- z_o = np.matmul(a_h, self.output_weights) + self.output_bias
-
- exp_term = np.exp(z_o)
- probabilities = exp_term / np.sum(exp_term, axis=1, keepdims=True)
- return probabilities
-
- def backpropagation(self):
- error_output = self.probabilities - self.Y_data
- error_hidden = np.matmul(error_output, self.output_weights.T) * self.a_h * (1 - self.a_h)
-
- self.output_weights_gradient = np.matmul(self.a_h.T, error_output)
- self.output_bias_gradient = np.sum(error_output, axis=0)
-
- self.hidden_weights_gradient = np.matmul(self.X_data.T, error_hidden)
- self.hidden_bias_gradient = np.sum(error_hidden, axis=0)
-
- if self.lmbd > 0.0:
- self.output_weights_gradient += self.lmbd * self.output_weights
- self.hidden_weights_gradient += self.lmbd * self.hidden_weights
-
- self.output_weights -= self.eta * self.output_weights_gradient
- self.output_bias -= self.eta * self.output_bias_gradient
- self.hidden_weights -= self.eta * self.hidden_weights_gradient
- self.hidden_bias -= self.eta * self.hidden_bias_gradient
-
- def predict(self, X):
- probabilities = self.feed_forward_out(X)
- return np.argmax(probabilities, axis=1)
-
- def predict_probabilities(self, X):
- probabilities = self.feed_forward_out(X)
- return probabilities
-
- def train(self):
- data_indices = np.arange(self.n_inputs)
-
- for i in range(self.epochs):
- for j in range(self.iterations):
- # pick datapoints with replacement
- chosen_datapoints = np.random.choice(
- data_indices, size=self.batch_size, replace=False
- )
-
- # minibatch training data
- self.X_data = self.X_data_full[chosen_datapoints]
- self.Y_data = self.Y_data_full[chosen_datapoints]
-
- self.feed_forward()
- self.backpropagation()
-
-
-
-
-
-
-
-
- Evaluate model performance on test data
+Full object-oriented implementation
-
@@ -475,22 +477,105 @@ The accuracy is as you would expect just the number of images correctly labeled
@@ -532,7 +617,7 @@ test_predict = dnn96
-
epochs = 100
-batch_size = 100
+
class NeuralNetwork:
+ def __init__(
+ self,
+ X_data,
+ Y_data,
+ n_hidden_neurons=50,
+ n_categories=10,
+ epochs=10,
+ batch_size=100,
+ eta=0.1,
+ lmbd=0.0):
-dnn = NeuralNetwork(X_train, Y_train_onehot, eta=eta, lmbd=lmbd, epochs=epochs, batch_size=batch_size,
- n_hidden_neurons=n_hidden_neurons, n_categories=n_categories)
-dnn.train()
-test_predict = dnn.predict(X_test)
+ self.X_data_full = X_data
+ self.Y_data_full = Y_data
-# accuracy score from scikit library
-print("Accuracy score on test set: ", accuracy_score(Y_test, test_predict))
+ self.n_inputs = X_data.shape[0]
+ self.n_features = X_data.shape[1]
+ self.n_hidden_neurons = n_hidden_neurons
+ self.n_categories = n_categories
-# equivalent in numpy
-def accuracy_score_numpy(Y_test, Y_pred):
- return np.sum(Y_test == Y_pred) / len(Y_test)
+ self.epochs = epochs
+ self.batch_size = batch_size
+ self.iterations = self.n_inputs // self.batch_size
+ self.eta = eta
+ self.lmbd = lmbd
-#print("Accuracy score on test set: ", accuracy_score_numpy(Y_test, test_predict))
+ self.create_biases_and_weights()
+
+ def create_biases_and_weights(self):
+ self.hidden_weights = np.random.randn(self.n_features, self.n_hidden_neurons)
+ self.hidden_bias = np.zeros(self.n_hidden_neurons) + 0.01
+
+ self.output_weights = np.random.randn(self.n_hidden_neurons, self.n_categories)
+ self.output_bias = np.zeros(self.n_categories) + 0.01
+
+ def feed_forward(self):
+ # feed-forward for training
+ self.z_h = np.matmul(self.X_data, self.hidden_weights) + self.hidden_bias
+ self.a_h = sigmoid(self.z_h)
+
+ self.z_o = np.matmul(self.a_h, self.output_weights) + self.output_bias
+
+ exp_term = np.exp(self.z_o)
+ self.probabilities = exp_term / np.sum(exp_term, axis=1, keepdims=True)
+
+ def feed_forward_out(self, X):
+ # feed-forward for output
+ z_h = np.matmul(X, self.hidden_weights) + self.hidden_bias
+ a_h = sigmoid(z_h)
+
+ z_o = np.matmul(a_h, self.output_weights) + self.output_bias
+
+ exp_term = np.exp(z_o)
+ probabilities = exp_term / np.sum(exp_term, axis=1, keepdims=True)
+ return probabilities
+
+ def backpropagation(self):
+ error_output = self.probabilities - self.Y_data
+ error_hidden = np.matmul(error_output, self.output_weights.T) * self.a_h * (1 - self.a_h)
+
+ self.output_weights_gradient = np.matmul(self.a_h.T, error_output)
+ self.output_bias_gradient = np.sum(error_output, axis=0)
+
+ self.hidden_weights_gradient = np.matmul(self.X_data.T, error_hidden)
+ self.hidden_bias_gradient = np.sum(error_hidden, axis=0)
+
+ if self.lmbd > 0.0:
+ self.output_weights_gradient += self.lmbd * self.output_weights
+ self.hidden_weights_gradient += self.lmbd * self.hidden_weights
+
+ self.output_weights -= self.eta * self.output_weights_gradient
+ self.output_bias -= self.eta * self.output_bias_gradient
+ self.hidden_weights -= self.eta * self.hidden_weights_gradient
+ self.hidden_bias -= self.eta * self.hidden_bias_gradient
+
+ def predict(self, X):
+ probabilities = self.feed_forward_out(X)
+ return np.argmax(probabilities, axis=1)
+
+ def predict_probabilities(self, X):
+ probabilities = self.feed_forward_out(X)
+ return probabilities
+
+ def train(self):
+ data_indices = np.arange(self.n_inputs)
+
+ for i in range(self.epochs):
+ for j in range(self.iterations):
+ # pick datapoints with replacement
+ chosen_datapoints = np.random.choice(
+ data_indices, size=self.batch_size, replace=False
+ )
+
+ # minibatch training data
+ self.X_data = self.X_data_full[chosen_datapoints]
+ self.Y_data = self.Y_data_full[chosen_datapoints]
+
+ self.feed_forward()
+ self.backpropagation()
Adjust hyperparameters
+Evaluate model performance on test data
-
@@ -470,26 +482,22 @@ Note that we are only using 1 layer with 50 neurons, and human performance is es
@@ -531,7 +539,7 @@ DNN_numpy = np.
-
eta_vals = np.logspace(-5, 1, 7)
-lmbd_vals = np.logspace(-5, 1, 7)
-# store the models for later use
-DNN_numpy = np.zeros((len(eta_vals), len(lmbd_vals)), dtype=object)
+
epochs = 100
+batch_size = 100
-# grid search
-for i, eta in enumerate(eta_vals):
- for j, lmbd in enumerate(lmbd_vals):
- dnn = NeuralNetwork(X_train, Y_train_onehot, eta=eta, lmbd=lmbd, epochs=epochs, batch_size=batch_size,
- n_hidden_neurons=n_hidden_neurons, n_categories=n_categories)
- dnn.train()
-
- DNN_numpy[i][j] = dnn
-
- test_predict = dnn.predict(X_test)
-
- print("Learning rate = ", eta)
- print("Lambda = ", lmbd)
- print("Accuracy score on test set: ", accuracy_score(Y_test, test_predict))
- print()
+dnn = NeuralNetwork(X_train, Y_train_onehot, eta=eta, lmbd=lmbd, epochs=epochs, batch_size=batch_size,
+ n_hidden_neurons=n_hidden_neurons, n_categories=n_categories)
+dnn.train()
+test_predict = dnn.predict(X_test)
+
+# accuracy score from scikit library
+print("Accuracy score on test set: ", accuracy_score(Y_test, test_predict))
+
+# equivalent in numpy
+def accuracy_score_numpy(Y_test, Y_pred):
+ return np.sum(Y_test == Y_pred) / len(Y_test)
+
+#print("Accuracy score on test set: ", accuracy_score_numpy(Y_test, test_predict))
Visualization
+Adjust hyperparameters
+
+
@@ -539,6 +537,8 @@ plt.show()
-
# visual representation of grid search
-# uses seaborn heatmap, you can also do this with matplotlib imshow
-import seaborn as sns
+
eta_vals = np.logspace(-5, 1, 7)
+lmbd_vals = np.logspace(-5, 1, 7)
+# store the models for later use
+DNN_numpy = np.zeros((len(eta_vals), len(lmbd_vals)), dtype=object)
-sns.set()
-
-train_accuracy = np.zeros((len(eta_vals), len(lmbd_vals)))
-test_accuracy = np.zeros((len(eta_vals), len(lmbd_vals)))
-
-for i in range(len(eta_vals)):
- for j in range(len(lmbd_vals)):
- dnn = DNN_numpy[i][j]
+# grid search
+for i, eta in enumerate(eta_vals):
+ for j, lmbd in enumerate(lmbd_vals):
+ dnn = NeuralNetwork(X_train, Y_train_onehot, eta=eta, lmbd=lmbd, epochs=epochs, batch_size=batch_size,
+ n_hidden_neurons=n_hidden_neurons, n_categories=n_categories)
+ dnn.train()
- train_pred = dnn.predict(X_train)
- test_pred = dnn.predict(X_test)
-
- train_accuracy[i][j] = accuracy_score(Y_train, train_pred)
- test_accuracy[i][j] = accuracy_score(Y_test, test_pred)
-
+ DNN_numpy[i][j] = dnn
-fig, ax = plt.subplots(figsize = (10, 10))
-sns.heatmap(train_accuracy, annot=True, ax=ax, cmap="viridis")
-ax.set_title("Training Accuracy")
-ax.set_ylabel("$\eta$")
-ax.set_xlabel("$\lambda$")
-plt.show()
-
-fig, ax = plt.subplots(figsize = (10, 10))
-sns.heatmap(test_accuracy, annot=True, ax=ax, cmap="viridis")
-ax.set_title("Test Accuracy")
-ax.set_ylabel("$\eta$")
-ax.set_xlabel("$\lambda$")
-plt.show()
+ test_predict = dnn.predict(X_test)
+
+ print("Learning rate = ", eta)
+ print("Lambda = ", lmbd)
+ print("Accuracy score on test set: ", accuracy_score(Y_test, test_predict))
+ print()
scikit-learn implementation
-
-Visualization
@@ -480,22 +473,39 @@ performance overall.
@@ -535,6 +545,7 @@ DNN_scikit = np
-
from sklearn.neural_network import MLPClassifier
-# store models for later use
-DNN_scikit = np.zeros((len(eta_vals), len(lmbd_vals)), dtype=object)
+
# visual representation of grid search
+# uses seaborn heatmap, you can also do this with matplotlib imshow
+import seaborn as sns
-for i, eta in enumerate(eta_vals):
- for j, lmbd in enumerate(lmbd_vals):
- dnn = MLPClassifier(hidden_layer_sizes=(n_hidden_neurons), activation='logistic',
- alpha=lmbd, learning_rate_init=eta, max_iter=epochs)
- dnn.fit(X_train, Y_train)
+sns.set()
+
+train_accuracy = np.zeros((len(eta_vals), len(lmbd_vals)))
+test_accuracy = np.zeros((len(eta_vals), len(lmbd_vals)))
+
+for i in range(len(eta_vals)):
+ for j in range(len(lmbd_vals)):
+ dnn = DNN_numpy[i][j]
- DNN_scikit[i][j] = dnn
+ train_pred = dnn.predict(X_train)
+ test_pred = dnn.predict(X_test)
+
+ train_accuracy[i][j] = accuracy_score(Y_train, train_pred)
+ test_accuracy[i][j] = accuracy_score(Y_test, test_pred)
+
- print("Learning rate = ", eta)
- print("Lambda = ", lmbd)
- print("Accuracy score on test set: ", dnn.score(X_test, Y_test))
- print()
+fig, ax = plt.subplots(figsize = (10, 10))
+sns.heatmap(train_accuracy, annot=True, ax=ax, cmap="viridis")
+ax.set_title("Training Accuracy")
+ax.set_ylabel("$\eta$")
+ax.set_xlabel("$\lambda$")
+plt.show()
+
+fig, ax = plt.subplots(figsize = (10, 10))
+sns.heatmap(test_accuracy, annot=True, ax=ax, cmap="viridis")
+ax.set_title("Test Accuracy")
+ax.set_ylabel("$\eta$")
+ax.set_xlabel("$\lambda$")
+plt.show()
Visualization
+scikit-learn implementation
+
+
@@ -465,40 +487,22 @@ MathJax.Hub.Config({
@@ -537,6 +541,7 @@ plt.show()
-
# optional
-# visual representation of grid search
-# uses seaborn heatmap, could probably do this in matplotlib
-import seaborn as sns
+
from sklearn.neural_network import MLPClassifier
+# store models for later use
+DNN_scikit = np.zeros((len(eta_vals), len(lmbd_vals)), dtype=object)
-sns.set()
-
-train_accuracy = np.zeros((len(eta_vals), len(lmbd_vals)))
-test_accuracy = np.zeros((len(eta_vals), len(lmbd_vals)))
-
-for i in range(len(eta_vals)):
- for j in range(len(lmbd_vals)):
- dnn = DNN_scikit[i][j]
+for i, eta in enumerate(eta_vals):
+ for j, lmbd in enumerate(lmbd_vals):
+ dnn = MLPClassifier(hidden_layer_sizes=(n_hidden_neurons), activation='logistic',
+ alpha=lmbd, learning_rate_init=eta, max_iter=epochs)
+ dnn.fit(X_train, Y_train)
- train_pred = dnn.predict(X_train)
- test_pred = dnn.predict(X_test)
-
- train_accuracy[i][j] = accuracy_score(Y_train, train_pred)
- test_accuracy[i][j] = accuracy_score(Y_test, test_pred)
-
+ DNN_scikit[i][j] = dnn
-fig, ax = plt.subplots(figsize = (10, 10))
-sns.heatmap(train_accuracy, annot=True, ax=ax, cmap="viridis")
-ax.set_title("Training Accuracy")
-ax.set_ylabel("$\eta$")
-ax.set_xlabel("$\lambda$")
-plt.show()
-
-fig, ax = plt.subplots(figsize = (10, 10))
-sns.heatmap(test_accuracy, annot=True, ax=ax, cmap="viridis")
-ax.set_title("Test Accuracy")
-ax.set_ylabel("$\eta$")
-ax.set_xlabel("$\lambda$")
-plt.show()
+ print("Learning rate = ", eta)
+ print("Lambda = ", lmbd)
+ print("Accuracy score on test set: ", dnn.score(X_test, Y_test))
+ print()
Building neural networks in Tensorflow and Keras
+Visualization
-
+
-
+
+
+
+
+
+
+
+ # optional
+# visual representation of grid search
+# uses seaborn heatmap, could probably do this in matplotlib
+import seaborn as sns
+
+sns.set()
+
+train_accuracy = np.zeros((len(eta_vals), len(lmbd_vals)))
+test_accuracy = np.zeros((len(eta_vals), len(lmbd_vals)))
+
+for i in range(len(eta_vals)):
+ for j in range(len(lmbd_vals)):
+ dnn = DNN_scikit[i][j]
+
+ train_pred = dnn.predict(X_train)
+ test_pred = dnn.predict(X_test)
+
+ train_accuracy[i][j] = accuracy_score(Y_train, train_pred)
+ test_accuracy[i][j] = accuracy_score(Y_test, test_pred)
+
+
+fig, ax = plt.subplots(figsize = (10, 10))
+sns.heatmap(train_accuracy, annot=True, ax=ax, cmap="viridis")
+ax.set_title("Training Accuracy")
+ax.set_ylabel("$\eta$")
+ax.set_xlabel("$\lambda$")
+plt.show()
+
+fig, ax = plt.subplots(figsize = (10, 10))
+sns.heatmap(test_accuracy, annot=True, ax=ax, cmap="viridis")
+ax.set_title("Test Accuracy")
+ax.set_ylabel("$\eta$")
+ax.set_xlabel("$\lambda$")
+plt.show()
+
+
+
+
+
+
+
+
+ Layout of a neural network with three hidden layers
+Layout of a neural network with three hidden layers (last later = \( l=L=4 \), first layer \( l=0 \))
+Layout of input to first hidden layer \( l=1 \) from input layer \( l=0 \)
+
+
+
+Derivatives and the chain rule
diff --git a/doc/pub/week42/html/week42-solarized.html b/doc/pub/week42/html/week42-solarized.html
index 100d78c42..f8e3db8b1 100644
--- a/doc/pub/week42/html/week42-solarized.html
+++ b/doc/pub/week42/html/week42-solarized.html
@@ -131,15 +131,21 @@ div.toc p,a {
2,
None,
'setting-up-the-equations-for-a-neural-network'),
- ('Layout of a neural network with three hidden layers',
+ ('Layout of a neural network with three hidden layers (last '
+ 'later = $l=L=4$, first layer $l=0$)',
2,
None,
- 'layout-of-a-neural-network-with-three-hidden-layers'),
+ 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'),
('Definitions', 2, None, 'definitions'),
('Inputs to the activation function',
2,
None,
'inputs-to-the-activation-function'),
+ ('Layout of input to first hidden layer $l=1$ from input layer '
+ '$l=0$',
+ 2,
+ None,
+ 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'),
('Derivatives and the chain rule',
2,
None,
@@ -936,11 +942,11 @@ all inputs \( \boldsymbol{x} \) are given by \( \boldsymbol{\tilde{y}}_i \).
-Layout of a neural network with three hidden layers
+Layout of a neural network with three hidden layers (last later = \( l=L=4 \), first layer \( l=0 \))
@@ -985,6 +991,15 @@ a_j^l = \sigma(z_j^l) = \frac{1}{1+\exp{-(z_j^l)}}.
$$
+
+Layout of input to first hidden layer \( l=1 \) from input layer \( l=0 \)
+
+
+
+
Derivatives and the chain rule
diff --git a/doc/pub/week42/html/week42.html b/doc/pub/week42/html/week42.html
index 483203b5c..5311f40b0 100644
--- a/doc/pub/week42/html/week42.html
+++ b/doc/pub/week42/html/week42.html
@@ -208,15 +208,21 @@ div.toc p,a {
2,
None,
'setting-up-the-equations-for-a-neural-network'),
- ('Layout of a neural network with three hidden layers',
+ ('Layout of a neural network with three hidden layers (last '
+ 'later = $l=L=4$, first layer $l=0$)',
2,
None,
- 'layout-of-a-neural-network-with-three-hidden-layers'),
+ 'layout-of-a-neural-network-with-three-hidden-layers-last-later-l-l-4-first-layer-l-0'),
('Definitions', 2, None, 'definitions'),
('Inputs to the activation function',
2,
None,
'inputs-to-the-activation-function'),
+ ('Layout of input to first hidden layer $l=1$ from input layer '
+ '$l=0$',
+ 2,
+ None,
+ 'layout-of-input-to-first-hidden-layer-l-1-from-input-layer-l-0'),
('Derivatives and the chain rule',
2,
None,
@@ -1013,11 +1019,11 @@ all inputs \( \boldsymbol{x} \) are given by \( \boldsymbol{\tilde{y}}_i \).
-Layout of a neural network with three hidden layers
+Layout of a neural network with three hidden layers (last later = \( l=L=4 \), first layer \( l=0 \))
@@ -1062,6 +1068,15 @@ a_j^l = \sigma(z_j^l) = \frac{1}{1+\exp{-(z_j^l)}}.
$$
+
+Layout of input to first hidden layer \( l=1 \) from input layer \( l=0 \)
+
+
+
+
Derivatives and the chain rule
diff --git a/doc/pub/week42/ipynb/ipynb-week42-src.tar.gz b/doc/pub/week42/ipynb/ipynb-week42-src.tar.gz
index 98f481598..026cc35e2 100644
Binary files a/doc/pub/week42/ipynb/ipynb-week42-src.tar.gz and b/doc/pub/week42/ipynb/ipynb-week42-src.tar.gz differ
diff --git a/doc/pub/week42/ipynb/week42.ipynb b/doc/pub/week42/ipynb/week42.ipynb
index 209f12073..581ead8a7 100644
--- a/doc/pub/week42/ipynb/week42.ipynb
+++ b/doc/pub/week42/ipynb/week42.ipynb
@@ -2,7 +2,7 @@
"cells": [
{
"cell_type": "markdown",
- "id": "71143de3",
+ "id": "092c6813",
"metadata": {
"editable": true
},
@@ -14,7 +14,7 @@
},
{
"cell_type": "markdown",
- "id": "35db3d31",
+ "id": "f44f930c",
"metadata": {
"editable": true
},
@@ -27,7 +27,7 @@
},
{
"cell_type": "markdown",
- "id": "3a41a07e",
+ "id": "52ac4c14",
"metadata": {
"editable": true
},
@@ -56,7 +56,7 @@
},
{
"cell_type": "markdown",
- "id": "e4274067",
+ "id": "2c1630dd",
"metadata": {
"editable": true
},
@@ -73,7 +73,7 @@
},
{
"cell_type": "markdown",
- "id": "38ccf7d0",
+ "id": "49d3cbcf",
"metadata": {
"editable": true
},
@@ -92,7 +92,7 @@
},
{
"cell_type": "markdown",
- "id": "bf1932a3",
+ "id": "9b42371e",
"metadata": {
"editable": true
},
@@ -108,7 +108,7 @@
},
{
"cell_type": "markdown",
- "id": "1cb1eb45",
+ "id": "b8f7a026",
"metadata": {
"editable": true
},
@@ -119,7 +119,7 @@
},
{
"cell_type": "markdown",
- "id": "0f07a5df",
+ "id": "012dd6a1",
"metadata": {
"editable": true
},
@@ -133,7 +133,7 @@
},
{
"cell_type": "markdown",
- "id": "8671fb13",
+ "id": "2847e093",
"metadata": {
"editable": true
},
@@ -148,7 +148,7 @@
},
{
"cell_type": "markdown",
- "id": "12ed4396",
+ "id": "45fef288",
"metadata": {
"editable": true
},
@@ -160,7 +160,7 @@
},
{
"cell_type": "markdown",
- "id": "c1516595",
+ "id": "de152393",
"metadata": {
"editable": true
},
@@ -174,7 +174,7 @@
},
{
"cell_type": "markdown",
- "id": "2949b61a",
+ "id": "621fb2b9",
"metadata": {
"editable": true
},
@@ -186,7 +186,7 @@
},
{
"cell_type": "markdown",
- "id": "46e01f35",
+ "id": "c418fb3c",
"metadata": {
"editable": true
},
@@ -202,7 +202,7 @@
},
{
"cell_type": "markdown",
- "id": "9c6ea29d",
+ "id": "800aecac",
"metadata": {
"editable": true
},
@@ -218,7 +218,7 @@
},
{
"cell_type": "markdown",
- "id": "6c66c7ee",
+ "id": "2e973cf8",
"metadata": {
"editable": true
},
@@ -230,7 +230,7 @@
},
{
"cell_type": "markdown",
- "id": "f8846711",
+ "id": "95e54e97",
"metadata": {
"editable": true
},
@@ -240,7 +240,7 @@
},
{
"cell_type": "markdown",
- "id": "e4b9e408",
+ "id": "6fd99322",
"metadata": {
"editable": true
},
@@ -252,7 +252,7 @@
},
{
"cell_type": "markdown",
- "id": "b8f62859",
+ "id": "2b2ebf34",
"metadata": {
"editable": true
},
@@ -262,7 +262,7 @@
},
{
"cell_type": "markdown",
- "id": "f2ec677d",
+ "id": "d488d19a",
"metadata": {
"editable": true
},
@@ -274,7 +274,7 @@
},
{
"cell_type": "markdown",
- "id": "31099b72",
+ "id": "af6e84a4",
"metadata": {
"editable": true
},
@@ -284,7 +284,7 @@
},
{
"cell_type": "markdown",
- "id": "cc6de265",
+ "id": "c9407fa7",
"metadata": {
"editable": true
},
@@ -296,7 +296,7 @@
},
{
"cell_type": "markdown",
- "id": "de05db8e",
+ "id": "b33546c2",
"metadata": {
"editable": true
},
@@ -312,7 +312,7 @@
},
{
"cell_type": "markdown",
- "id": "a5fcef62",
+ "id": "e2b7759f",
"metadata": {
"editable": true
},
@@ -324,7 +324,7 @@
},
{
"cell_type": "markdown",
- "id": "38f87147",
+ "id": "5e1062d7",
"metadata": {
"editable": true
},
@@ -336,7 +336,7 @@
},
{
"cell_type": "markdown",
- "id": "92822d7d",
+ "id": "2733bbc4",
"metadata": {
"editable": true
},
@@ -346,7 +346,7 @@
},
{
"cell_type": "markdown",
- "id": "7d08d013",
+ "id": "b72c5008",
"metadata": {
"editable": true
},
@@ -358,7 +358,7 @@
},
{
"cell_type": "markdown",
- "id": "8ca04a0a",
+ "id": "b219dc77",
"metadata": {
"editable": true
},
@@ -368,7 +368,7 @@
},
{
"cell_type": "markdown",
- "id": "cd2b820f",
+ "id": "5313f959",
"metadata": {
"editable": true
},
@@ -384,7 +384,7 @@
},
{
"cell_type": "markdown",
- "id": "74a4bf55",
+ "id": "fe80b0ed",
"metadata": {
"editable": true
},
@@ -396,7 +396,7 @@
},
{
"cell_type": "markdown",
- "id": "b6a6ea2c",
+ "id": "fdd61c89",
"metadata": {
"editable": true
},
@@ -408,7 +408,7 @@
},
{
"cell_type": "markdown",
- "id": "f569e2c4",
+ "id": "e73defca",
"metadata": {
"editable": true
},
@@ -420,7 +420,7 @@
},
{
"cell_type": "markdown",
- "id": "11dd5f28",
+ "id": "6a10c098",
"metadata": {
"editable": true
},
@@ -432,7 +432,7 @@
},
{
"cell_type": "markdown",
- "id": "e79bd0c6",
+ "id": "5e45bea0",
"metadata": {
"editable": true
},
@@ -444,7 +444,7 @@
},
{
"cell_type": "markdown",
- "id": "0994dc8a",
+ "id": "345aecc6",
"metadata": {
"editable": true
},
@@ -454,7 +454,7 @@
},
{
"cell_type": "markdown",
- "id": "a5fb5850",
+ "id": "65142a0a",
"metadata": {
"editable": true
},
@@ -470,7 +470,7 @@
},
{
"cell_type": "markdown",
- "id": "75ef56f2",
+ "id": "5faab1bb",
"metadata": {
"editable": true
},
@@ -482,7 +482,7 @@
},
{
"cell_type": "markdown",
- "id": "2cba4793",
+ "id": "476fcd72",
"metadata": {
"editable": true
},
@@ -494,7 +494,7 @@
},
{
"cell_type": "markdown",
- "id": "002f872b",
+ "id": "8c8120ee",
"metadata": {
"editable": true
},
@@ -504,7 +504,7 @@
},
{
"cell_type": "markdown",
- "id": "ca9db79a",
+ "id": "5f3741cc",
"metadata": {
"editable": true
},
@@ -516,7 +516,7 @@
},
{
"cell_type": "markdown",
- "id": "91c832ea",
+ "id": "129deee5",
"metadata": {
"editable": true
},
@@ -530,7 +530,7 @@
},
{
"cell_type": "markdown",
- "id": "497597b5",
+ "id": "9991265c",
"metadata": {
"editable": true
},
@@ -548,7 +548,7 @@
{
"cell_type": "code",
"execution_count": 1,
- "id": "31785b91",
+ "id": "d70cb287",
"metadata": {
"collapsed": false,
"editable": true
@@ -621,7 +621,7 @@
},
{
"cell_type": "markdown",
- "id": "fb3ba38a",
+ "id": "0c995e02",
"metadata": {
"editable": true
},
@@ -631,7 +631,7 @@
},
{
"cell_type": "markdown",
- "id": "f39aa22a",
+ "id": "22b79621",
"metadata": {
"editable": true
},
@@ -648,7 +648,7 @@
},
{
"cell_type": "markdown",
- "id": "07786f01",
+ "id": "554f6b51",
"metadata": {
"editable": true
},
@@ -660,7 +660,7 @@
},
{
"cell_type": "markdown",
- "id": "855ae860",
+ "id": "5f747812",
"metadata": {
"editable": true
},
@@ -670,7 +670,7 @@
},
{
"cell_type": "markdown",
- "id": "f982b39a",
+ "id": "6f9a465a",
"metadata": {
"editable": true
},
@@ -682,7 +682,7 @@
},
{
"cell_type": "markdown",
- "id": "5b0b9a6a",
+ "id": "8071b978",
"metadata": {
"editable": true
},
@@ -698,7 +698,7 @@
},
{
"cell_type": "markdown",
- "id": "f74b7c7d",
+ "id": "bf0242dd",
"metadata": {
"editable": true
},
@@ -710,7 +710,7 @@
},
{
"cell_type": "markdown",
- "id": "dbaa42b5",
+ "id": "5ffe91c1",
"metadata": {
"editable": true
},
@@ -722,7 +722,7 @@
},
{
"cell_type": "markdown",
- "id": "8f7dc89c",
+ "id": "f5512cb1",
"metadata": {
"editable": true
},
@@ -733,7 +733,7 @@
},
{
"cell_type": "markdown",
- "id": "adfec9e2",
+ "id": "fa7c53f0",
"metadata": {
"editable": true
},
@@ -745,7 +745,7 @@
},
{
"cell_type": "markdown",
- "id": "1a91248a",
+ "id": "b500870f",
"metadata": {
"editable": true
},
@@ -758,7 +758,7 @@
},
{
"cell_type": "markdown",
- "id": "f8dd5dde",
+ "id": "79cffbd0",
"metadata": {
"editable": true
},
@@ -770,7 +770,7 @@
},
{
"cell_type": "markdown",
- "id": "806e1970",
+ "id": "75bca176",
"metadata": {
"editable": true
},
@@ -780,7 +780,7 @@
},
{
"cell_type": "markdown",
- "id": "887c1cc5",
+ "id": "e6aa2766",
"metadata": {
"editable": true
},
@@ -792,7 +792,7 @@
},
{
"cell_type": "markdown",
- "id": "4bf40527",
+ "id": "47f01305",
"metadata": {
"editable": true
},
@@ -804,7 +804,7 @@
},
{
"cell_type": "markdown",
- "id": "a5d878ba",
+ "id": "f242e593",
"metadata": {
"editable": true
},
@@ -816,7 +816,7 @@
},
{
"cell_type": "markdown",
- "id": "318c3664",
+ "id": "54f82c8e",
"metadata": {
"editable": true
},
@@ -826,7 +826,7 @@
},
{
"cell_type": "markdown",
- "id": "c4fc9951",
+ "id": "04835564",
"metadata": {
"editable": true
},
@@ -838,7 +838,7 @@
},
{
"cell_type": "markdown",
- "id": "3d187d65",
+ "id": "b364a288",
"metadata": {
"editable": true
},
@@ -854,7 +854,7 @@
},
{
"cell_type": "markdown",
- "id": "24ebe80e",
+ "id": "7ce5168e",
"metadata": {
"editable": true
},
@@ -866,7 +866,7 @@
},
{
"cell_type": "markdown",
- "id": "640153c7",
+ "id": "7d7d5872",
"metadata": {
"editable": true
},
@@ -876,7 +876,7 @@
},
{
"cell_type": "markdown",
- "id": "9f00a103",
+ "id": "c0e39147",
"metadata": {
"editable": true
},
@@ -888,7 +888,7 @@
},
{
"cell_type": "markdown",
- "id": "c313a954",
+ "id": "b086049b",
"metadata": {
"editable": true
},
@@ -898,7 +898,7 @@
},
{
"cell_type": "markdown",
- "id": "8f9899f9",
+ "id": "094989b6",
"metadata": {
"editable": true
},
@@ -910,7 +910,7 @@
},
{
"cell_type": "markdown",
- "id": "fe894efa",
+ "id": "2742317a",
"metadata": {
"editable": true
},
@@ -922,7 +922,7 @@
},
{
"cell_type": "markdown",
- "id": "ff3f394f",
+ "id": "0e2c957d",
"metadata": {
"editable": true
},
@@ -935,7 +935,7 @@
},
{
"cell_type": "markdown",
- "id": "91783800",
+ "id": "d81132fa",
"metadata": {
"editable": true
},
@@ -945,7 +945,7 @@
},
{
"cell_type": "markdown",
- "id": "b9199d44",
+ "id": "9c779d7e",
"metadata": {
"editable": true
},
@@ -957,7 +957,7 @@
},
{
"cell_type": "markdown",
- "id": "0cc20449",
+ "id": "d1318243",
"metadata": {
"editable": true
},
@@ -967,7 +967,7 @@
},
{
"cell_type": "markdown",
- "id": "a376f16a",
+ "id": "4f9977f9",
"metadata": {
"editable": true
},
@@ -979,7 +979,7 @@
},
{
"cell_type": "markdown",
- "id": "9cf47847",
+ "id": "fdc2aba9",
"metadata": {
"editable": true
},
@@ -990,7 +990,7 @@
},
{
"cell_type": "markdown",
- "id": "c1bd9901",
+ "id": "36334b10",
"metadata": {
"editable": true
},
@@ -1002,7 +1002,7 @@
},
{
"cell_type": "markdown",
- "id": "addef0ec",
+ "id": "d7c74c36",
"metadata": {
"editable": true
},
@@ -1012,7 +1012,7 @@
},
{
"cell_type": "markdown",
- "id": "a5b267fc",
+ "id": "9258fbf2",
"metadata": {
"editable": true
},
@@ -1024,7 +1024,7 @@
},
{
"cell_type": "markdown",
- "id": "53d64a33",
+ "id": "4703fbac",
"metadata": {
"editable": true
},
@@ -1034,7 +1034,7 @@
},
{
"cell_type": "markdown",
- "id": "765d3576",
+ "id": "0b451589",
"metadata": {
"editable": true
},
@@ -1046,7 +1046,7 @@
},
{
"cell_type": "markdown",
- "id": "680e6d32",
+ "id": "75530820",
"metadata": {
"editable": true
},
@@ -1058,7 +1058,7 @@
},
{
"cell_type": "markdown",
- "id": "6b5a3f94",
+ "id": "1103fe71",
"metadata": {
"editable": true
},
@@ -1070,7 +1070,7 @@
},
{
"cell_type": "markdown",
- "id": "c0b94fbc",
+ "id": "f2ea48ee",
"metadata": {
"editable": true
},
@@ -1080,7 +1080,7 @@
},
{
"cell_type": "markdown",
- "id": "a8bc0cb4",
+ "id": "a508fd86",
"metadata": {
"editable": true
},
@@ -1092,7 +1092,7 @@
},
{
"cell_type": "markdown",
- "id": "fc5ba87a",
+ "id": "fc4448cf",
"metadata": {
"editable": true
},
@@ -1102,7 +1102,7 @@
},
{
"cell_type": "markdown",
- "id": "d7928417",
+ "id": "7585a56f",
"metadata": {
"editable": true
},
@@ -1114,7 +1114,7 @@
},
{
"cell_type": "markdown",
- "id": "38fcd712",
+ "id": "b2b6e1f2",
"metadata": {
"editable": true
},
@@ -1126,7 +1126,7 @@
},
{
"cell_type": "markdown",
- "id": "a68a28b5",
+ "id": "2e3f4fd2",
"metadata": {
"editable": true
},
@@ -1138,7 +1138,7 @@
},
{
"cell_type": "markdown",
- "id": "81b5762d",
+ "id": "7506d23b",
"metadata": {
"editable": true
},
@@ -1148,7 +1148,7 @@
},
{
"cell_type": "markdown",
- "id": "6485ea35",
+ "id": "c53a7ad0",
"metadata": {
"editable": true
},
@@ -1160,7 +1160,7 @@
},
{
"cell_type": "markdown",
- "id": "56f15a69",
+ "id": "e741bcd1",
"metadata": {
"editable": true
},
@@ -1170,7 +1170,7 @@
},
{
"cell_type": "markdown",
- "id": "7f3597fa",
+ "id": "3f40c8b4",
"metadata": {
"editable": true
},
@@ -1183,7 +1183,7 @@
},
{
"cell_type": "markdown",
- "id": "f3af219f",
+ "id": "38133e07",
"metadata": {
"editable": true
},
@@ -1195,7 +1195,7 @@
},
{
"cell_type": "markdown",
- "id": "7a21d5ea",
+ "id": "a48e0d24",
"metadata": {
"editable": true
},
@@ -1205,7 +1205,7 @@
},
{
"cell_type": "markdown",
- "id": "d0c3898f",
+ "id": "c2527d7c",
"metadata": {
"editable": true
},
@@ -1217,7 +1217,7 @@
},
{
"cell_type": "markdown",
- "id": "34325b5c",
+ "id": "5b70545f",
"metadata": {
"editable": true
},
@@ -1227,7 +1227,7 @@
},
{
"cell_type": "markdown",
- "id": "918bb1fa",
+ "id": "02d7a474",
"metadata": {
"editable": true
},
@@ -1239,7 +1239,7 @@
},
{
"cell_type": "markdown",
- "id": "894b36e6",
+ "id": "28f26a48",
"metadata": {
"editable": true
},
@@ -1249,7 +1249,7 @@
},
{
"cell_type": "markdown",
- "id": "55f20c6b",
+ "id": "7cc8726f",
"metadata": {
"editable": true
},
@@ -1261,7 +1261,7 @@
},
{
"cell_type": "markdown",
- "id": "0d783805",
+ "id": "123de6da",
"metadata": {
"editable": true
},
@@ -1271,7 +1271,7 @@
},
{
"cell_type": "markdown",
- "id": "f64324de",
+ "id": "518455da",
"metadata": {
"editable": true
},
@@ -1288,7 +1288,7 @@
},
{
"cell_type": "markdown",
- "id": "c860274d",
+ "id": "c6591a2b",
"metadata": {
"editable": true
},
@@ -1300,7 +1300,7 @@
},
{
"cell_type": "markdown",
- "id": "266ed951",
+ "id": "e8b6fcf1",
"metadata": {
"editable": true
},
@@ -1312,23 +1312,23 @@
},
{
"cell_type": "markdown",
- "id": "e3597760",
+ "id": "ced47734",
"metadata": {
"editable": true
},
"source": [
- "## Layout of a neural network with three hidden layers\n",
+ "## Layout of a neural network with three hidden layers (last later = $l=L=4$, first layer $l=0$)\n",
"\n",
- "\n",
+ "\n",
"\n",
"\n",
- "![](\"figures/nn1.png\")
![](\"figures/nn2.png\")
![](\"figures/structure.png\")