# My presentations on ‘Elements of Neural Networks & Deep Learning’ -Parts 4,5

This is the next set of presentations on “Elements of Neural Networks and Deep Learning”.  In the 4th presentation I discuss and derive the generalized equations for a multi-unit, multi-layer Deep Learning network.  The 5th presentation derives the equations for a Deep Learning network when performing multi-class classification along with the derivations for cross-entropy loss. The corresponding implementations are available in vectorized R, Python and Octave are available in my book ‘Deep Learning from first principles:Second edition- In vectorized Python, R and Octave

Important note: Do check out my later version of these videos at Take 4+: Presentations on ‘Elements of Neural Networks and Deep Learning’ – Parts 1-8 . These have more content and also include some corrections. Check it out!

1. Elements of Neural Network and Deep Learning – Part 4
This presentation is a continuation of my 3rd presentation in which I derived the equations for a simple 3 layer Neural Network with 1 hidden layer. In this video presentation, I discuss step-by-step the derivations for a L-Layer, multi-unit Deep Learning Network, with any activation function g(z)

The implementations of L-Layer, multi-unit Deep Learning Network in vectorized R, Python and Octave are available in my post Deep Learning from first principles in Python, R and Octave – Part 3

2. Elements of Neural Network and Deep Learning – Part 5
This presentation discusses multi-class classification using the Softmax function. The detailed derivation for the Jacobian of the Softmax is discussed, and subsequently the derivative of cross-entropy loss is also discussed in detail. Finally the final set of equations for a Neural Network with multi-class classification is derived.

The corresponding implementations in vectorized R, Python and Octave are available in the following posts
a. Deep Learning from first principles in Python, R and Octave – Part 4
b. Deep Learning from first principles in Python, R and Octave – Part 5

To be continued. Watch this space!

Checkout my book ‘Deep Learning from first principles: Second Edition – In vectorized Python, R and Octave’. My book starts with the implementation of a simple 2-layer Neural Network and works its way to a generic L-Layer Deep Learning Network, with all the bells and whistles. The derivations have been discussed in detail. The code has been extensively commented and included in its entirety in the Appendix sections. My book is available on Amazon as paperback ($18.99) and in kindle version($9.99/Rs449).

To see all posts click Index of Posts

# My book ‘Deep Learning from first principles:Second Edition’ now on Amazon

The second edition of my book ‘Deep Learning from first principles:Second Edition- In vectorized Python, R and Octave’, is now available on Amazon, in both paperback ($18.99) and kindle ($9.99/Rs449/-)  versions. Since this book is almost 70% code, all functions, and code snippets have been formatted to use the fixed-width font ‘Lucida Console’. In addition line numbers have been added to all code snippets. This makes the code more organized and much more readable. I have also fixed typos in the book

The book includes the following chapters

Table of Contents
Preface 4
Introduction 6
1. Logistic Regression as a Neural Network 8
2. Implementing a simple Neural Network 23
3. Building a L- Layer Deep Learning Network 48
4. Deep Learning network with the Softmax 85
5. MNIST classification with Softmax 103
6. Initialization, regularization in Deep Learning 121
7. Gradient Descent Optimization techniques 167
8. Gradient Check in Deep Learning 197
1. Appendix A 214
2. Appendix 1 – Logistic Regression as a Neural Network 220
3. Appendix 2 - Implementing a simple Neural Network 227
4. Appendix 3 - Building a L- Layer Deep Learning Network 240
5. Appendix 4 - Deep Learning network with the Softmax 259
6. Appendix 5 - MNIST classification with Softmax 269
7. Appendix 6 - Initialization, regularization in Deep Learning 302
8. Appendix 7 - Gradient Descent Optimization techniques 344
9. Appendix 8 – Gradient Check 405
References 475

To see posts click Index of Posts

# My book “Deep Learning from first principles” now on Amazon

Note: The 2nd edition of this book is now available on Amazon

My 4th book(self-published), “Deep Learning from first principles – In vectorized Python, R and Octave” (557 pages), is now available on Amazon in both paperback ($18.99) and kindle ($9.99/Rs449). The book starts with the most primitive 2-layer Neural Network and works  its way to a generic L-layer Deep Learning Network, with all the bells and whistles.  The book includes detailed derivations and vectorized implementations in Python, R and Octave.  The code has been extensively  commented and has been included in the Appendix section.

# Deep Learning from first principles in Python, R and Octave – Part 8

## 1. Introduction

You don’t understand anything until you learn it more than one way. Marvin Minsky
No computer has ever been designed that is ever aware of what it’s doing; but most of the time, we aren’t either. Marvin Minsky
A wealth of information creates a poverty of attention. Herbert Simon

This post, Deep Learning from first Principles in Python, R and Octave-Part8, is my final post in my Deep Learning from first principles series. In this post, I discuss and implement a key functionality needed while building Deep Learning networks viz. ‘Gradient Checking’. Gradient Checking is an important method to check the correctness of your implementation, specifically the forward propagation and the backward propagation cycles of an implementation. In addition I also discuss some tips for tuning hyper-parameters of a Deep Learning network based on my experience.

My post in this  ‘Deep Learning Series’ so far were
1. Deep Learning from first principles in Python, R and Octave – Part 1 In part 1, I implement logistic regression as a neural network in vectorized Python, R and Octave
2. Deep Learning from first principles in Python, R and Octave – Part 2 In the second part I implement a simple Neural network with just 1 hidden layer and a sigmoid activation output function
3. Deep Learning from first principles in Python, R and Octave – Part 3 The 3rd part implemented a multi-layer Deep Learning Network with sigmoid activation output in vectorized Python, R and Octave
4. Deep Learning from first principles in Python, R and Octave – Part 4 The 4th part deals with multi-class classification. Specifically, I derive the Jacobian of the Softmax function and enhance my L-Layer DL network to include Softmax output function in addition to Sigmoid activation
5. Deep Learning from first principles in Python, R and Octave – Part 5 This post uses the Softmax classifier implemented to classify MNIST digits using a L-layer Deep Learning network
6. Deep Learning from first principles in Python, R and Octave – Part 6 The 6th part adds more bells and whistles to my L-Layer DL network, by including different initialization types namely He and Xavier. Besides L2 Regularization and random dropout is added.
7. Deep Learning from first principles in Python, R and Octave – Part 7 The 7th part deals with Stochastic Gradient Descent Optimization methods including momentum, RMSProp and Adam
8. Deep Learning from first principles in Python, R and Octave – Part 8 – This post implements a critical function for ensuring the correctness of a L-Layer Deep Learning network implementation using Gradient Checking

Checkout my book ‘Deep Learning from first principles: Second Edition – In vectorized Python, R and Octave’. My book starts with the implementation of a simple 2-layer Neural Network and works its way to a generic L-Layer Deep Learning Network, with all the bells and whistles. The derivations have been discussed in detail. The code has been extensively commented and included in its entirety in the Appendix sections. My book is available on Amazon as paperback ($18.99) and in kindle version($9.99/Rs449).

You may also like my companion book “Practical Machine Learning with R and Python- Machine Learning in stereo” available in Amazon in paperback($9.99) and Kindle($6.99) versions. This book is ideal for a quick reference of the various ML functions and associated measurements in both R and Python which are essential to delve deep into Deep Learning.

Gradient Checking is based on the following approach. One iteration of Gradient Descent computes and updates the parameters $\theta$ by doing
$\theta := \theta - \frac{d}{d\theta}J(\theta)$.
To minimize the cost we will need to minimize $J(\theta)$
Let $g(\theta)$ be a function that computes the derivative $\frac {d}{d\theta}J(\theta)$. Gradient Checking allows us to numerically evaluate the implementation of the function $g(\theta)$ and verify its correctness.
We know the derivative of a function is given by
$\frac {d}{d\theta}J(\theta) = lim->0 \frac {J(\theta +\epsilon) - J(\theta -\epsilon)} {2*\epsilon}$
Note: The above derivative is based on the 2 sided derivative. The 1-sided derivative  is given by $\frac {d}{d\theta}J(\theta) = lim->0 \frac {J(\theta +\epsilon) - J(\theta)} {\epsilon}$
Gradient Checking is based on the 2-sided derivative because the error is of the order $O(\epsilon^{2})$ as opposed $O(\epsilon)$ for the 1-sided derivative.
Hence Gradient Check uses the 2 sided derivative as follows.
$g(\theta) = lim->0 \frac {J(\theta +\epsilon) - J(\theta -\epsilon)} {2*\epsilon}$

In Gradient Check the following is done
A) Run one normal cycle of your implementation by doing the following
a) Compute the output activation by running 1 cycle of forward propagation
b) Compute the cost using the output activation

B) Perform gradient check steps as below
a) Set $\theta$ . Flatten all ‘weights’ and ‘bias’ matrices and vectors to a column vector.
b) Initialize $\theta+$ by bumping up $\theta$ by adding $\epsilon$ ($\theta + \epsilon$)
c) Perform forward propagation with $\theta+$
d) Compute cost with $\theta+$ i.e. $J(\theta+)$
e) Initialize  $\theta-$ by bumping down $\theta$ by subtracting $\epsilon$ $(\theta - \epsilon)$
f) Perform forward propagation with $\theta-$
g) Compute cost with $\theta-$ i.e.  $J(\theta-)$
h) Compute $\frac {d} {d\theta} J(\theta)$ or ‘gradapprox’ as$\frac {J(\theta+) - J(\theta-) } {2\epsilon}$using the 2 sided derivative.
i) Compute L2norm or the Euclidean distance between ‘grad’ and ‘gradapprox’. If the
diference is of the order of $10^{-5}$ or $10^{-7}$ the implementation is correct. In the Deep Learning Specialization Prof Andrew Ng mentions that if the difference is of the order of $10^{-7}$ then the implementation is correct. A difference of $10^{-5}$ is also ok. Anything more than that is a cause of worry and you should look at your code more closely. To see more details click Gradient checking and advanced optimization

After spending a better part of 3 days, I now realize how critical Gradient Check is for ensuring the correctness of you implementation. Initially I was getting very high difference and did not know how to understand the results or debug my implementation. After many hours of staring at the results, I  was able to finally arrive at a way, to localize issues in the implementation. In fact, I did catch a small bug in my Python code, which did not exist in the R and Octave implementations. I will demonstrate this below

## 1.1a Gradient Check – Sigmoid Activation – Python

import numpy as np
import matplotlib

train_X, train_Y, test_X, test_Y = load_dataset()
#Set layer dimensions
layersDimensions = [2,4,1]
parameters = initializeDeepModel(layersDimensions)
#Perform forward prop
AL, caches, dropoutMat = forwardPropagationDeep(train_X, parameters, keep_prob=1, hiddenActivationFunc="relu",outputActivationFunc="sigmoid")
#Compute cost
cost = computeCost(AL, train_Y, outputActivationFunc="sigmoid")
print("cost=",cost)
gradients = backwardPropagationDeep(AL, train_Y, caches, dropoutMat, lambd=0, keep_prob=1,                                   hiddenActivationFunc="relu",outputActivationFunc="sigmoid")

epsilon = 1e-7
outputActivationFunc="sigmoid"

# Set-up variables
# Flatten parameters to a vector
parameters_values, _ = dictionary_to_vector(parameters)
num_parameters = parameters_values.shape[0]
#Initialize
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))

# Compute gradapprox using 2 sided derivative
for i in range(num_parameters):
# Compute J_plus[i].
thetaplus = np.copy(parameters_values)
thetaplus[i][0] = thetaplus[i][0] + epsilon
AL, caches, dropoutMat = forwardPropagationDeep(train_X, vector_to_dictionary(parameters,thetaplus), keep_prob=1,
hiddenActivationFunc="relu",outputActivationFunc=outputActivationFunc)
J_plus[i] = computeCost(AL, train_Y, outputActivationFunc=outputActivationFunc)

# Compute J_minus[i].
thetaminus = np.copy(parameters_values)
thetaminus[i][0] = thetaminus[i][0] - epsilon
AL, caches, dropoutMat  = forwardPropagationDeep(train_X, vector_to_dictionary(parameters,thetaminus), keep_prob=1,
hiddenActivationFunc="relu",outputActivationFunc=outputActivationFunc)
J_minus[i] = computeCost(AL, train_Y, outputActivationFunc=outputActivationFunc)

difference =  numerator/denominator

#Check the difference
if difference > 1e-5:
print ("\033[93m" + "There is a mistake in the backward propagation! difference = " + str(difference) + "\033[0m")
else:
print ("\033[92m" + "Your backward propagation works perfectly fine! difference = " + str(difference) + "\033[0m")
print(difference)
print("\n")
# The technique below can be used to identify
# which of the parameters are in error
print(m)
print("\n")
print(n)

## (300, 2)
## (300,)
## cost= 0.6931455556341791
## [92mYour backward propagation works perfectly fine! difference = 1.1604150683743381e-06[0m
## 1.1604150683743381e-06
##
##
## {'dW1': array([[-6.19439955e-06, -2.06438046e-06],
##        [-1.50165447e-05,  7.50401672e-05],
##        [ 1.33435433e-04,  1.74112143e-04],
##        [-3.40909024e-05, -1.38363681e-04]]), 'db1': array([[ 7.31333221e-07],
##        [ 7.98425950e-06],
##        [ 8.15002817e-08],
##        [-5.69821155e-08]]), 'dW2': array([[2.73416304e-04, 2.96061451e-04, 7.51837363e-05, 1.01257729e-04]]), 'db2': array([[-7.22232235e-06]])}
##
##
## {'dW1': array([[-6.19448937e-06, -2.06501483e-06],
##        [-1.50168766e-05,  7.50399742e-05],
##        [ 1.33435485e-04,  1.74112391e-04],
##        [-3.40910633e-05, -1.38363765e-04]]), 'db1': array([[ 7.31081862e-07],
##        [ 7.98472399e-06],
##        [ 8.16013923e-08],
##        [-5.71764858e-08]]), 'dW2': array([[2.73416290e-04, 2.96061509e-04, 7.51831930e-05, 1.01257891e-04]]), 'db2': array([[-7.22255589e-06]])}

## 1.1b Gradient Check – Softmax Activation – Python (Error!!)

In the code below I show, how I managed to spot a bug in your implementation

import numpy as np
N = 100 # number of points per class
D = 2 # dimensionality
K = 3 # number of classes
X = np.zeros((N*K,D)) # data matrix (each row = single example)
y = np.zeros(N*K, dtype='uint8') # class labels
for j in range(K):
ix = range(N*j,N*(j+1))
t = np.linspace(j*4,(j+1)*4,N) + np.random.randn(N)*0.2 # theta
X[ix] = np.c_[r*np.sin(t), r*np.cos(t)]
y[ix] = j

# Plot the data
#plt.scatter(X[:, 0], X[:, 1], c=y, s=40, cmap=plt.cm.Spectral)
layersDimensions = [2,3,3]
y1=y.reshape(-1,1).T
train_X=X.T
train_Y=y1

parameters = initializeDeepModel(layersDimensions)
#Compute forward prop
AL, caches, dropoutMat = forwardPropagationDeep(train_X, parameters, keep_prob=1,
hiddenActivationFunc="relu",outputActivationFunc="softmax")
#Compute cost
cost = computeCost(AL, train_Y, outputActivationFunc="softmax")
print("cost=",cost)
gradients = backwardPropagationDeep(AL, train_Y, caches, dropoutMat, lambd=0, keep_prob=1,
hiddenActivationFunc="relu",outputActivationFunc="softmax")
# Note the transpose of the gradients for Softmax has to be taken
L= len(parameters)//2
print(L)
gradient_check_n(parameters, gradients, train_X, train_Y, epsilon = 1e-7,outputActivationFunc="softmax")

cost= 1.0986187818144022
2
There is a mistake in the backward propagation! difference = 0.7100295155692544
0.7100295155692544

{'dW1': array([[ 0.00050125,  0.00045194],
[ 0.00096392,  0.00039641],
[-0.00014276, -0.00045639]]), 'db1': array([[ 0.00070082],
[-0.00224399],
[ 0.00052305]]), 'dW2': array([[-8.40953794e-05, -9.52657769e-04, -1.10269379e-04],
[-7.45469382e-04,  9.49795606e-04,  2.29045434e-04],
[ 8.29564761e-04,  2.86216305e-06, -1.18776055e-04]]),
'db2': array([[-0.00253808],
[-0.00505508],
[ 0.00759315]])}

{'dW1': array([[ 0.00050125,  0.00045194],
[ 0.00096392,  0.00039641],
[-0.00014276, -0.00045639]]), 'db1': array([[ 0.00070082],
[-0.00224399],
[ 0.00052305]]), 'dW2': array([[-8.40960634e-05, -9.52657953e-04, -1.10268461e-04],
[-7.45469242e-04,  9.49796908e-04,  2.29045671e-04],
[ 8.29565305e-04,  2.86104473e-06, -1.18776100e-04]]),
'db2': array([[-8.46211989e-06],
[-1.68487446e-05],
[ 2.53108645e-05]])}

Gradient Check gives a high value of the difference of 0.7100295. Inspecting the Gradients and Gradapprox we can see there is a very big discrepancy in db2. After I went over my code I discovered that I my computation in the function layerActivationBackward for Softmax was


# Erroneous code
if activationFunc == 'softmax':
dW = 1/numtraining * np.dot(A_prev,dZ)
db = np.sum(dZ, axis=0, keepdims=True)
dA_prev = np.dot(dZ,W)
# Fixed code
if activationFunc == 'softmax':
dW = 1/numtraining * np.dot(A_prev,dZ)
db = 1/numtraining *  np.sum(dZ, axis=0, keepdims=True)
dA_prev = np.dot(dZ,W)


After fixing this error when I ran Gradient Check I get

## 1.1c Gradient Check – Softmax Activation – Python (Corrected!!)

import numpy as np
N = 100 # number of points per class
D = 2 # dimensionality
K = 3 # number of classes
X = np.zeros((N*K,D)) # data matrix (each row = single example)
y = np.zeros(N*K, dtype='uint8') # class labels
for j in range(K):
ix = range(N*j,N*(j+1))
t = np.linspace(j*4,(j+1)*4,N) + np.random.randn(N)*0.2 # theta
X[ix] = np.c_[r*np.sin(t), r*np.cos(t)]
y[ix] = j

# Plot the data
#plt.scatter(X[:, 0], X[:, 1], c=y, s=40, cmap=plt.cm.Spectral)
layersDimensions = [2,3,3]
y1=y.reshape(-1,1).T
train_X=X.T
train_Y=y1
#Set layer dimensions
parameters = initializeDeepModel(layersDimensions)
#Perform forward prop
AL, caches, dropoutMat = forwardPropagationDeep(train_X, parameters, keep_prob=1,
hiddenActivationFunc="relu",outputActivationFunc="softmax")
#Compute cost
cost = computeCost(AL, train_Y, outputActivationFunc="softmax")
print("cost=",cost)
gradients = backwardPropagationDeep(AL, train_Y, caches, dropoutMat, lambd=0, keep_prob=1,
hiddenActivationFunc="relu",outputActivationFunc="softmax")
# Note the transpose of the gradients for Softmax has to be taken
L= len(parameters)//2
print(L)
gradient_check_n(parameters, gradients, train_X, train_Y, epsilon = 1e-7,outputActivationFunc="softmax")
## cost= 1.0986193170234435
## 2
## [92mYour backward propagation works perfectly fine! difference = 5.268804859613151e-07[0m
## 5.268804859613151e-07
##
##
## {'dW1': array([[ 0.00053206,  0.00038987],
##        [ 0.00093941,  0.00038077],
##        [-0.00012177, -0.0004692 ]]), 'db1': array([[ 0.00072662],
##        [-0.00210198],
##        [ 0.00046741]]), 'dW2': array([[-7.83441270e-05, -9.70179498e-04, -1.08715815e-04],
##        [-7.70175008e-04,  9.54478237e-04,  2.27690198e-04],
##        [ 8.48519135e-04,  1.57012608e-05, -1.18974383e-04]]), 'db2': array([[-8.52190476e-06],
##        [-1.69954294e-05],
##        [ 2.55173342e-05]])}
##
##
## {'dW1': array([[ 0.00053206,  0.00038987],
##        [ 0.00093941,  0.00038077],
##        [-0.00012177, -0.0004692 ]]), 'db1': array([[ 0.00072662],
##        [-0.00210198],
##        [ 0.00046741]]), 'dW2': array([[-7.83439980e-05, -9.70180603e-04, -1.08716369e-04],
##        [-7.70173925e-04,  9.54478718e-04,  2.27690089e-04],
##        [ 8.48520143e-04,  1.57018842e-05, -1.18973720e-04]]), 'db2': array([[-8.52096171e-06],
##        [-1.69964043e-05],
##        [ 2.55162558e-05]])}

## 1.2a Gradient Check – Sigmoid Activation – R

source("DLfunctions8.R")

x <- z[,1:2]
y <- z[,3]
X <- t(x)
Y <- t(y)
#Set layer dimensions
layersDimensions = c(2,5,1)
parameters = initializeDeepModel(layersDimensions)
#Perform forward prop
retvals = forwardPropagationDeep(X, parameters,keep_prob=1, hiddenActivationFunc="relu",
outputActivationFunc="sigmoid")
AL <- retvals[['AL']]
caches <- retvals[['caches']]
dropoutMat <- retvals[['dropoutMat']]
#Compute cost
cost <- computeCost(AL, Y,outputActivationFunc="sigmoid",
numClasses=layersDimensions[length(layersDimensions)])
print(cost)
## [1] 0.6931447
# Backward propagation.
gradients = backwardPropagationDeep(AL, Y, caches, dropoutMat, lambd=0, keep_prob=1, hiddenActivationFunc="relu",
outputActivationFunc="sigmoid",numClasses=layersDimensions[length(layersDimensions)])
epsilon = 1e-07
outputActivationFunc="sigmoid"
#Convert parameter list to vector
parameters_values = list_to_vector(parameters)
num_parameters = dim(parameters_values)[1]
#Initialize
J_plus = matrix(rep(0,num_parameters),
nrow=num_parameters,ncol=1)
J_minus = matrix(rep(0,num_parameters),
nrow=num_parameters,ncol=1)
nrow=num_parameters,ncol=1)

for(i in 1:num_parameters){
# Compute J_plus[i].
thetaplus = parameters_values
thetaplus[i][1] = thetaplus[i][1] + epsilon
retvals = forwardPropagationDeep(X, vector_to_list(parameters,thetaplus), keep_prob=1,
hiddenActivationFunc="relu",outputActivationFunc=outputActivationFunc)

AL <- retvals[['AL']]
J_plus[i] = computeCost(AL, Y, outputActivationFunc=outputActivationFunc)

# Compute J_minus[i].
thetaminus = parameters_values
thetaminus[i][1] = thetaminus[i][1] - epsilon
retvals  = forwardPropagationDeep(X, vector_to_list(parameters,thetaminus), keep_prob=1,
hiddenActivationFunc="relu",outputActivationFunc=outputActivationFunc)
AL <- retvals[['AL']]
J_minus[i] = computeCost(AL, Y, outputActivationFunc=outputActivationFunc)

}
#Compute L2Norm
difference =  numerator/denominator
if(difference > 1e-5){
cat("There is a mistake, the difference is too high",difference)
} else{
cat("The implementations works perfectly", difference)
}
## The implementations works perfectly 1.279911e-06
# This can be used to check
print("Gradients from backprop")
## [1] "Gradients from backprop"
vector_to_list2(parameters,grad)
## $dW1 ## [,1] [,2] ## [1,] -7.641588e-05 -3.427989e-07 ## [2,] -9.049683e-06 6.906304e-05 ## [3,] 3.401039e-06 -1.503914e-04 ## [4,] 1.535226e-04 -1.686402e-04 ## [5,] -6.029292e-05 -2.715648e-04 ## ##$db1
##               [,1]
## [1,]  6.930318e-06
## [2,] -3.283117e-05
## [3,]  1.310647e-05
## [4,] -3.454308e-05
## [5,] -2.331729e-08
##
## $dW2 ## [,1] [,2] [,3] [,4] [,5] ## [1,] 0.0001612356 0.0001113475 0.0002435824 0.000362149 2.874116e-05 ## ##$db2
##              [,1]
## [1,] -1.16364e-05
print("Grad approx from gradient check")
## [1] "Grad approx from gradient check"
vector_to_list2(parameters,gradapprox)
## $dW1 ## [,1] [,2] ## [1,] -7.641554e-05 -3.430589e-07 ## [2,] -9.049428e-06 6.906253e-05 ## [3,] 3.401168e-06 -1.503919e-04 ## [4,] 1.535228e-04 -1.686401e-04 ## [5,] -6.029288e-05 -2.715650e-04 ## ##$db1
##               [,1]
## [1,]  6.930012e-06
## [2,] -3.283096e-05
## [3,]  1.310618e-05
## [4,] -3.454237e-05
## [5,] -2.275957e-08
##
## $dW2 ## [,1] [,2] [,3] [,4] [,5] ## [1,] 0.0001612355 0.0001113476 0.0002435829 0.0003621486 2.87409e-05 ## ##$db2
##              [,1]
## [1,] -1.16368e-05

## 1.2b Gradient Check – Softmax Activation – R

source("DLfunctions8.R")

# Setup the data
X <- Z[,1:2]
y <- Z[,3]
X <- t(X)
Y <- t(y)
layersDimensions = c(2, 3, 3)
parameters = initializeDeepModel(layersDimensions)
#Perform forward prop
retvals = forwardPropagationDeep(X, parameters,keep_prob=1, hiddenActivationFunc="relu",
outputActivationFunc="softmax")
AL <- retvals[['AL']]
caches <- retvals[['caches']]
dropoutMat <- retvals[['dropoutMat']]
#Compute cost
cost <- computeCost(AL, Y,outputActivationFunc="softmax",
numClasses=layersDimensions[length(layersDimensions)])
print(cost)
## [1] 1.098618
# Backward propagation.
gradients = backwardPropagationDeep(AL, Y, caches, dropoutMat, lambd=0, keep_prob=1, hiddenActivationFunc="relu",
outputActivationFunc="softmax",numClasses=layersDimensions[length(layersDimensions)])
# Need to take transpose of the last layer for Softmax
L=length(parameters)/2
epsilon = 1e-7,outputActivationFunc="softmax")
## The implementations works perfectly 3.903011e-07[1] "Gradients from backprop"
## $dW1 ## [,1] [,2] ## [1,] 0.0007962367 -0.0001907606 ## [2,] 0.0004444254 0.0010354412 ## [3,] 0.0003078611 0.0007591255 ## ##$db1
##               [,1]
## [1,] -0.0017305136
## [2,]  0.0005393734
## [3,]  0.0012484550
##
## $dW2 ## [,1] [,2] [,3] ## [1,] -3.515627e-04 7.487283e-04 -3.971656e-04 ## [2,] -6.381521e-05 -1.257328e-06 6.507254e-05 ## [3,] -1.719479e-04 -4.857264e-04 6.576743e-04 ## ##$db2
##               [,1]
## [1,] -5.536383e-06
## [2,] -1.824656e-05
## [3,]  2.378295e-05
##
## $dW1 ## [,1] [,2] ## [1,] 0.0007962364 -0.0001907607 ## [2,] 0.0004444256 0.0010354406 ## [3,] 0.0003078615 0.0007591250 ## ##$db1
##               [,1]
## [1,] -0.0017305135
## [2,]  0.0005393741
## [3,]  0.0012484547
##
## $dW2 ## [,1] [,2] [,3] ## [1,] -3.515632e-04 7.487277e-04 -3.971656e-04 ## [2,] -6.381451e-05 -1.257883e-06 6.507239e-05 ## [3,] -1.719469e-04 -4.857270e-04 6.576739e-04 ## ##$db2
##               [,1]
## [1,] -5.536682e-06
## [2,] -1.824652e-05
## [3,]  2.378209e-05

## 1.3a Gradient Check – Sigmoid Activation – Octave

source("DL8functions.m")
################## Circles

X=data(:,1:2);
Y=data(:,3);
#Set layer dimensions
layersDimensions = [2 5  1]; #tanh=-0.5(ok), #relu=0.1 best!
[weights biases] = initializeDeepModel(layersDimensions);
#Perform forward prop
[AL forward_caches activation_caches droputMat] = forwardPropagationDeep(X', weights, biases,keep_prob=1,
hiddenActivationFunc="relu", outputActivationFunc="sigmoid");
#Compute cost
cost = computeCost(AL, Y',outputActivationFunc=outputActivationFunc,numClasses=layersDimensions(size(layersDimensions)(2)));
disp(cost);
hiddenActivationFunc="relu", outputActivationFunc="sigmoid",
numClasses=layersDimensions(size(layersDimensions)(2)));
epsilon = 1e-07;
outputActivationFunc="sigmoid";
# Convert paramter cell array to vector
parameters_values = cellArray_to_vector(weights, biases);
#Convert gradient cell array to vector
num_parameters = size(parameters_values)(1);
#Initialize
J_plus = zeros(num_parameters, 1);
J_minus = zeros(num_parameters, 1);
for i = 1:num_parameters
# Compute J_plus[i].
thetaplus = parameters_values;
thetaplus(i,1) = thetaplus(i,1) + epsilon;
[weights1 biases1] =vector_to_cellArray(weights, biases,thetaplus);
[AL forward_caches activation_caches droputMat] = forwardPropagationDeep(X', weights1, biases1, keep_prob=1,
hiddenActivationFunc="relu",outputActivationFunc=outputActivationFunc);
J_plus(i) = computeCost(AL, Y', outputActivationFunc=outputActivationFunc);

# Compute J_minus[i].
thetaminus = parameters_values;
thetaminus(i,1) = thetaminus(i,1) - epsilon ;
[weights1 biases1] = vector_to_cellArray(weights, biases,thetaminus);
[AL forward_caches activation_caches droputMat]  = forwardPropagationDeep(X',weights1, biases1, keep_prob=1,
hiddenActivationFunc="relu",outputActivationFunc=outputActivationFunc);
J_minus(i) = computeCost(AL, Y', outputActivationFunc=outputActivationFunc);

endfor

#Compute L2Norm
difference =  numerator/denominator;
disp(difference);
#Check difference
if difference > 1e-04
printf("There is a mistake in the implementation ");
disp(difference);
else
printf("The implementation works perfectly");
disp(difference);
endif
disp(weights1);
disp(biases1);
disp(weights2);
disp(biases2);

0.69315
1.4893e-005
The implementation works perfectly 1.4893e-005
{
[1,1] =
5.0349e-005 2.1323e-005
8.8632e-007 1.8231e-006
9.3784e-005 1.0057e-004
1.0875e-004 -1.9529e-007
5.4502e-005 3.2721e-005
[1,2] =
1.0567e-005 6.0615e-005 4.6004e-005 1.3977e-004 1.0405e-004
}
{
[1,1] =
-1.8716e-005
1.1309e-009
4.7686e-005
1.2051e-005
-1.4612e-005
[1,2] = 9.5808e-006
}
{
[1,1] =
5.0348e-005 2.1320e-005
8.8485e-007 1.8219e-006
9.3784e-005 1.0057e-004
1.0875e-004 -1.9762e-007
5.4502e-005 3.2723e-005
[1,2] =
[1,2] =
1.0565e-005 6.0614e-005 4.6007e-005 1.3977e-004 1.0405e-004
}
{
[1,1] =
-1.8713e-005
1.1102e-009
4.7687e-005
1.2048e-005
-1.4609e-005
[1,2] = 9.5790e-006
}


## 1.3b Gradient Check – Softmax Activation – Octave

source("DL8functions.m")

# Setup the data
X=data(:,1:2);
Y=data(:,3);
# Set the layer dimensions
layersDimensions = [2 3  3];
[weights biases] = initializeDeepModel(layersDimensions);
# Run forward prop
[AL forward_caches activation_caches droputMat] = forwardPropagationDeep(X', weights, biases,keep_prob=1,
hiddenActivationFunc="relu", outputActivationFunc="softmax");
# Compute cost
cost = computeCost(AL, Y',outputActivationFunc=outputActivationFunc,numClasses=layersDimensions(size(layersDimensions)(2)));
disp(cost);
# Perform backward prop
hiddenActivationFunc="relu", outputActivationFunc="softmax",
numClasses=layersDimensions(size(layersDimensions)(2)));

#Take transpose of last layer for Softmax
L=size(weights)(2);
outputActivationFunc="softmax",numClasses=layersDimensions(size(layersDimensions)(2)));

 1.0986
The implementation works perfectly  2.0021e-005
{
[1,1] =
-7.1590e-005  4.1375e-005
-1.9494e-004  -5.2014e-005
-1.4554e-004  5.1699e-005
[1,2] =
3.3129e-004  1.9806e-004  -1.5662e-005
-4.9692e-004  -3.7756e-004  -8.2318e-005
1.6562e-004  1.7950e-004  9.7980e-005
}
{
[1,1] =
-3.0856e-005
-3.3321e-004
-3.8197e-004
[1,2] =
1.2046e-006
2.9259e-007
-1.4972e-006
}
{
[1,1] =
-7.1586e-005  4.1377e-005
-1.9494e-004  -5.2013e-005
-1.4554e-004  5.1695e-005
3.3129e-004  1.9806e-004  -1.5664e-005
-4.9692e-004  -3.7756e-004  -8.2316e-005
1.6562e-004  1.7950e-004  9.7979e-005
}
{
[1,1] =
-3.0852e-005
-3.3321e-004
-3.8197e-004
[1,2] =
1.1902e-006
2.8200e-007
-1.4644e-006
}


## 2.1 Tip for tuning hyperparameters

Deep Learning Networks come with a large number of hyper parameters which require tuning. The hyper parameters are

1. $\alpha$ -learning rate
2. Number of layers
3. Number of hidden units
4. Number of iterations
5. Momentum – $\beta$ – 0.9
6. RMSProp – $\beta_{1}$ – 0.9
7. Adam – $\beta_{1}$,$\beta_{2}$ and $\epsilon$
8. learning rate decay
9. mini batch size
10. Initialization method – He, Xavier
11. Regularization

– Among the above the most critical is learning rate $\alpha$ . Rather than just trying out random values, it may help to try out values on a logarithmic scale. So we could try out values -0.01,0.1,1.0,10 etc. If we find that the cost is between 0.01 and 0.1 we could use a technique similar to binary search or bisection, so we can try 0.01, 0.05. If we need to be bigger than 0.01 and 0.05 we could try 0.25  and then keep halving the distance etc.
– The performance of Momentum and RMSProp are very good and work well with values 0.9. Even with this, it is better to try out values of 1-$\beta$ in the logarithmic range. So 1-$\beta$ could 0.001,0.01,0.1 and hence $\beta$ would be 0.999,0.99 or 0.9
– Increasing the number of hidden units or number of hidden layers need to be done gradually. I have noticed that increasing number of hidden layers heavily does not improve performance and sometimes degrades it.
– Sometimes, I tend to increase the number of iterations if I think I see a steady decrease in the cost for a certain learning rate
– It may also help to add learning rate decay if you see there is an oscillation while it decreases.
– Xavier and He initializations also help in a fast convergence and are worth trying out.

## 3.1 Final thoughts

As I come to a close in this Deep Learning Series from first principles in Python, R and Octave, I must admit that I learnt a lot in the process.

* Building a L-layer, vectorized Deep Learning Network in Python, R and Octave was extremely challenging but very rewarding
* One benefit of building vectorized versions in Python, R and Octave was that I was looking at each function that I was implementing thrice, and hence I was able to fix any bugs in any of the languages
* In addition since I built the generic L-Layer DL network with all the bells and whistles, layer by layer I further had an opportunity to look at all the functions in each successive post.
* Each language has its advantages and disadvantages. From the performance perspective I think Python is the best, followed by Octave and then R
* Interesting, I noticed that even if small bugs creep into your implementation, the DL network does learn and does generate a valid set of weights and biases, however this may not be an optimum solution. In one case of an inadvertent bug, I was not updating the weights in the final layer of the DL network. Yet, using all the other layers, the DL network was able to come with a reasonable solution (maybe like random dropout, remaining units can still learn the data!)
* Having said that, the Gradient Check method discussed and implemented in this post can be very useful in ironing out bugs.

## Conclusion

These last couple of months when I was writing the posts and the also churning up the code in Python, R and Octave were  very hectic. There have been times when I found that implementations of some function to be extremely demanding and I almost felt like giving up. Other times, I have spent quite some time on an intractable DL network which would not respond to changes in hyper-parameters. All in all, it was a great learning experience. I would suggest that you start from my first post Deep Learning from first principles in Python, R and Octave-Part 1 and work your way up. Feel free to take the code apart and try out things. That is the only way you will learn.

Hope you had as much fun as I had. Stay tuned. I will be back!!!

To see all post click Index of Posts

# Deep Learning from first principles in Python, R and Octave – Part 5

## Introduction

a. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
b. A robot must obey orders given it by human beings except where such orders would conflict with the First Law.
c. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

      Isaac Asimov's Three Laws of Robotics 

Any sufficiently advanced technology is indistinguishable from magic.

      Arthur C Clarke.   

In this 5th part on Deep Learning from first Principles in Python, R and Octave, I solve the MNIST data set of handwritten digits (shown below), from the basics. To do this, I construct a L-Layer, vectorized Deep Learning implementation in Python, R and Octave from scratch and classify the  MNIST data set. The MNIST training data set  contains 60000 handwritten digits from 0-9, and a test set of 10000 digits. MNIST, is a popular dataset for running Deep Learning tests, and has been rightfully termed as the ‘drosophila’ of Deep Learning, by none other than the venerable Prof Geoffrey Hinton.

The ‘Deep Learning from first principles in Python, R and Octave’ series, so far included  Part 1 , where I had implemented logistic regression as a simple Neural Network. Part 2 implemented the most elementary neural network with 1 hidden layer, but  with any number of activation units in that layer, and a sigmoid activation at the output layer.

This post, ‘Deep Learning from first principles in Python, R and Octave – Part 5’ largely builds upon Part3. in which I implemented a multi-layer Deep Learning network, with an arbitrary number of hidden layers and activation units per hidden layer and with the output layer was based on the sigmoid unit, for binary classification. In Part 4, I derive the Jacobian of a Softmax, the Cross entropy loss and the gradient equations for a multi-class Softmax classifier. I also  implement a simple Neural Network using Softmax classifications in Python, R and Octave.

In this post I combine Part 3 and Part 4 to to build a L-layer Deep Learning network, with arbitrary number of hidden layers and hidden units, which can do both binary (sigmoid) and multi-class (softmax) classification.

Note: A detailed discussion of the derivation for multi-class clasification can be seen in my video presentation Neural Networks 5

The generic, vectorized L-Layer Deep Learning Network implementations in Python, R and Octave can be cloned/downloaded from GitHub at DeepLearning-Part5. This implementation allows for arbitrary number of hidden layers and hidden layer units. The activation function at the hidden layers can be one of sigmoid, relu and tanh (will be adding leaky relu soon). The output activation can be used for binary classification with the ‘sigmoid’, or multi-class classification with ‘softmax’. Feel free to download and play around with the code!

I thought the exercise of combining the two parts(Part 3, & Part 4)  would be a breeze. But it was anything but. Incorporating a Softmax classifier into the generic L-Layer Deep Learning model was a challenge. Moreover I found that I could not use the gradient descent on 60,000 training samples as my laptop ran out of memory. So I had to implement Stochastic Gradient Descent (SGD) for Python, R and Octave. In addition, I had to also implement the numerically stable version of Softmax, as the softmax and its derivative would result in NaNs.

### Numerically stable Softmax

The Softmax function $S_{j} =\frac{e^{Z_{j}}}{\sum_{i}^{k}e^{Z_{i}}}$ can be numerically unstable because of the division of large exponentials.  To handle this problem we have to implement stable Softmax function as below

$S_{j} =\frac{e^{Z_{j}}}{\sum_{i}^{k}e^{Z_{i}}}$
$S_{j} =\frac{e^{Z_{j}}}{\sum_{i}^{k}e^{Z_{i}}} = \frac{Ce^{Z_{j}}}{C\sum_{i}^{k}e^{Z_{i}}} = \frac{e^{Z_{j}+log(C)}}{\sum_{i}^{k}e^{Z_{i}+log(C)}}$
Therefore $S_{j} = \frac{e^{Z_{j}+ D}}{\sum_{i}^{k}e^{Z_{i}+ D}}$
Here ‘D’ can be anything. A common choice is
$D=-max(Z_{1},Z_{2},... Z_{k})$

Here is the stable Softmax implementation in Python

# A numerically stable Softmax implementation
def stableSoftmax(Z):
#Compute the softmax of vector x in a numerically stable way.
shiftZ = Z.T - np.max(Z.T,axis=1).reshape(-1,1)
exp_scores = np.exp(shiftZ)
# normalize them for each example
A = exp_scores / np.sum(exp_scores, axis=1, keepdims=True)
cache=Z
return A,cache


While trying to create a L-Layer generic Deep Learning network in the 3 languages, I found it useful to ensure that the model executed correctly on smaller datasets.  You can run into numerous problems while setting up the matrices, which becomes extremely difficult to debug. So in this post, I run the model on 2 smaller data for sets used in my earlier posts(Part 3 & Part4) , in each of the languages, before running the generic model on MNIST.

Here is a fair warning. if you think you can dive directly into Deep Learning, with just some basic knowledge of Machine Learning, you are bound to run into serious issues. Moreover, your knowledge will be incomplete. It is essential that you have a good grasp of Machine and Statistical Learning, the different algorithms, the measures and metrics for selecting the models etc.It would help to be conversant with all the ML models, ML concepts, validation techniques, classification measures  etc. Check out the internet/books for background.

Checkout my book ‘Deep Learning from first principles: Second Edition – In vectorized Python, R and Octave’. My book starts with the implementation of a simple 2-layer Neural Network and works its way to a generic L-Layer Deep Learning Network, with all the bells and whistles. The derivations have been discussed in detail. The code has been extensively commented and included in its entirety in the Appendix sections. My book is available on Amazon as paperback ($18.99) and in kindle version($9.99/Rs449).

You may also like my companion book “Practical Machine Learning with R and Python:Second Edition- Machine Learning in stereo” available in Amazon in paperback($10.99) and Kindle($7.99/Rs449) versions. This book is ideal for a quick reference of the various ML functions and associated measurements in both R and Python which are essential to delve deep into Deep Learning.

### 1. Random dataset with Sigmoid activation – Python

This random data with 9 clusters, was used in my post Deep Learning from first principles in Python, R and Octave – Part 3 , and was used to test the complete L-layer Deep Learning network with Sigmoid activation.

import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import pandas as pd
from sklearn.datasets import make_classification, make_blobs
exec(open("DLfunctions51.py").read()) # Cannot import in Rmd.
# Create a random data set with 9 centeres
X1, Y1 = make_blobs(n_samples = 400, n_features = 2, centers = 9,cluster_std = 1.3, random_state =4)

#Create 2 classes
Y1=Y1.reshape(400,1)
Y1 = Y1 % 2
X2=X1.T
Y2=Y1.T
# Set the dimensions of L -layer DL network
layersDimensions = [2, 9, 9,1] #  4-layer model
# Execute DL network with hidden activation=relu and sigmoid output function
parameters = L_Layer_DeepModel(X2, Y2, layersDimensions, hiddenActivationFunc='relu', outputActivationFunc="sigmoid",learningRate = 0.3,num_iterations = 2500, print_cost = True)

### 2. Spiral dataset with Softmax activation – Python

The Spiral data was used in my post Deep Learning from first principles in Python, R and Octave – Part 4 and was used to test the complete L-layer Deep Learning network with multi-class Softmax activation at the output layer

import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import pandas as pd
from sklearn.datasets import make_classification, make_blobs

# Create an input data set - Taken from CS231n Convolutional Neural networks
# http://cs231n.github.io/neural-networks-case-study/
N = 100 # number of points per class
D = 2 # dimensionality
K = 3 # number of classes
X = np.zeros((N*K,D)) # data matrix (each row = single example)
y = np.zeros(N*K, dtype='uint8') # class labels
for j in range(K):
ix = range(N*j,N*(j+1))
t = np.linspace(j*4,(j+1)*4,N) + np.random.randn(N)*0.2 # theta
X[ix] = np.c_[r*np.sin(t), r*np.cos(t)]
y[ix] = j

X1=X.T
Y1=y.reshape(-1,1).T
numHidden=100 # No of hidden units in hidden layer
numFeats= 2 # dimensionality
numOutput = 3 # number of classes
# Set the dimensions of the layers
layersDimensions=[numFeats,numHidden,numOutput]
parameters = L_Layer_DeepModel(X1, Y1, layersDimensions, hiddenActivationFunc='relu', outputActivationFunc="softmax",learningRate = 0.6,num_iterations = 9000, print_cost = True)
## Cost after iteration 0: 1.098759
## Cost after iteration 1000: 0.112666
## Cost after iteration 2000: 0.044351
## Cost after iteration 3000: 0.027491
## Cost after iteration 4000: 0.021898
## Cost after iteration 5000: 0.019181
## Cost after iteration 6000: 0.017832
## Cost after iteration 7000: 0.017452
## Cost after iteration 8000: 0.017161

### 3. MNIST dataset with Softmax activation – Python

In the code below, I execute Stochastic Gradient Descent on the MNIST training data of 60000. I used a mini-batch size of 1000. Python takes about 40 minutes to crunch the data. In addition I also compute the Confusion Matrix and other metrics like Accuracy, Precision and Recall for the MNIST data set. I get an accuracy of 0.93 on the MNIST test set. This accuracy can be improved by choosing more hidden layers or more hidden units and possibly also tweaking the learning rate and the number of epochs.

import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import pandas as pd
import math
from sklearn.datasets import make_classification, make_blobs
from sklearn.metrics import confusion_matrix
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
# Read the MNIST training and test sets
# Create labels and pixel arrays
lbls=[]
pxls=[]
print(len(training))
#for i in range(len(training)):
for i in range(60000):
l,p=training[i]
lbls.append(l)
pxls.append(p)
labels= np.array(lbls)
pixels=np.array(pxls)
y=labels.reshape(-1,1)
X=pixels.reshape(pixels.shape[0],-1)
X1=X.T
Y1=y.T
# Set the dimensions of the layers. The MNIST data is 28x28 pixels= 784
# Hence input layer is 784. For the 10 digits the Softmax classifier
# has to handle 10 outputs
layersDimensions=[784, 15,9,10] # Works very well,lr=0.01,mini_batch =1000, total=20000
np.random.seed(1)
costs = []
# Run Stochastic Gradient Descent with Learning Rate=0.01, mini batch size=1000
# number of epochs=3000
parameters = L_Layer_DeepModel_SGD(X1, Y1, layersDimensions, hiddenActivationFunc='relu', outputActivationFunc="softmax",learningRate = 0.01 ,mini_batch_size =1000, num_epochs = 3000, print_cost = True)

# Compute the Confusion Matrix on Training set
# Compute the training accuracy, precision and recall
proba=predict_proba(parameters, X1,outputActivationFunc="softmax")
#A2, cache = forwardPropagationDeep(X1, parameters)
#proba=np.argmax(A2, axis=0).reshape(-1,1)
a=confusion_matrix(Y1.T,proba)
print(a)
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
print('Accuracy: {:.2f}'.format(accuracy_score(Y1.T, proba)))
print('Precision: {:.2f}'.format(precision_score(Y1.T, proba,average="micro")))
print('Recall: {:.2f}'.format(recall_score(Y1.T, proba,average="micro")))

lbls=[]
pxls=[]
print(len(test))
for i in range(10000):
l,p=test[i]
lbls.append(l)
pxls.append(p)
testLabels= np.array(lbls)
testPixels=np.array(pxls)
ytest=testLabels.reshape(-1,1)
Xtest=testPixels.reshape(testPixels.shape[0],-1)
X1test=Xtest.T
Y1test=ytest.T

# Compute the Confusion Matrix on Test set
# Compute the test accuracy, precision and recall
probaTest=predict_proba(parameters, X1test,outputActivationFunc="softmax")
#A2, cache = forwardPropagationDeep(X1, parameters)
#proba=np.argmax(A2, axis=0).reshape(-1,1)
a=confusion_matrix(Y1test.T,probaTest)
print(a)
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
print('Accuracy: {:.2f}'.format(accuracy_score(Y1test.T, probaTest)))
print('Precision: {:.2f}'.format(precision_score(Y1test.T, probaTest,average="micro")))
print('Recall: {:.2f}'.format(recall_score(Y1test.T, probaTest,average="micro")))

##1.  Confusion Matrix of Training set
0     1    2    3    4    5    6    7    8    9
## [[5854    0   19    2   10    7    0    1   24    6]
##  [   1 6659   30   10    5    3    0   14   20    0]
##  [  20   24 5805   18    6   11    2   32   37    3]
##  [   5    4  175 5783    1   27    1   58   60   17]
##  [   1   21    9    0 5780    0    5    2   12   12]
##  [  29    9   21  224    6 4824   18   17  245   28]
##  [   5    4   22    1   32   12 5799    0   43    0]
##  [   3   13  148  154   18    3    0 5883    4   39]
##  [  11   34   30   21   13   16    4    7 5703   12]
##  [  10    4    1   32  135   14    1   92  134 5526]]

##2. Accuracy, Precision, Recall of  Training set
## Accuracy: 0.96
## Precision: 0.96
## Recall: 0.96

##3. Confusion Matrix of Test set
0     1    2    3    4    5    6    7    8    9
## [[ 954    1    8    0    3    3    2    4    4    1]
##  [   0 1107    6    5    0    0    1    2   14    0]
##  [  11    7  957   10    5    0    5   20   16    1]
##  [   2    3   37  925    3   13    0    8   18    1]
##  [   2    6    1    1  944    0    7    3    4   14]
##  [  12    5    4   45    2  740   24    8   42   10]
##  [   8    4    4    2   16    9  903    0   12    0]
##  [   4   10   27   18    5    1    0  940    1   22]
##  [  11   13    6   13    9   10    7    2  900    3]
##  [   8    5    1    7   50    7    0   20   29  882]]
##4. Accuracy, Precision, Recall of  Training set
## Accuracy: 0.93
## Precision: 0.93
## Recall: 0.93

### 4. Random dataset with Sigmoid activation – R code

This is the random data set used in the Python code above which was saved as a CSV. The code is used to test a L -Layer DL network with Sigmoid Activation in R.

source("DLfunctions5.R")
# Read the random data set
x <- z[,1:2]
y <- z[,3]
X <- t(x)
Y <- t(y)
# Set the dimensions of the  layer
layersDimensions = c(2, 9, 9,1)

# Run Gradient Descent on the data set with relu hidden unit activation
# sigmoid activation unit in the output layer
retvals = L_Layer_DeepModel(X, Y, layersDimensions,
hiddenActivationFunc='relu',
outputActivationFunc="sigmoid",
learningRate = 0.3,
numIterations = 5000,
print_cost = True)
#Plot the cost vs iterations
iterations <- seq(0,5000,1000)
costs=retvals$costs df=data.frame(iterations,costs) ggplot(df,aes(x=iterations,y=costs)) + geom_point() + geom_line(color="blue") + ggtitle("Costs vs iterations") + xlab("Iterations") + ylab("Loss") ### 5. Spiral dataset with Softmax activation – R The spiral data set used in the Python code above, is reused to test multi-class classification with Softmax. source("DLfunctions5.R") Z <- as.matrix(read.csv("spiral.csv",header=FALSE)) # Setup the data X <- Z[,1:2] y <- Z[,3] X <- t(X) Y <- t(y) # Initialize number of features, number of hidden units in hidden layer and # number of classes numFeats<-2 # No features numHidden<-100 # No of hidden units numOutput<-3 # No of classes # Set the layer dimensions layersDimensions = c(numFeats,numHidden,numOutput) # Perform gradient descent with relu activation unit for hidden layer # and softmax activation in the output retvals = L_Layer_DeepModel(X, Y, layersDimensions, hiddenActivationFunc='relu', outputActivationFunc="softmax", learningRate = 0.5, numIterations = 9000, print_cost = True) #Plot cost vs iterations iterations <- seq(0,9000,1000) costs=retvals$costs
df=data.frame(iterations,costs)
ggplot(df,aes(x=iterations,y=costs)) + geom_point() + geom_line(color="blue") +
ggtitle("Costs vs iterations") + xlab("Iterations") + ylab("Costs")

### 6. MNIST dataset with Softmax activation – R

The code below executes a L – Layer Deep Learning network with Softmax output activation, to classify the 10 handwritten digits from MNIST with Stochastic Gradient Descent. The entire 60000 data set was used to train the data. R takes almost 8 hours to process this data set with a mini-batch size of 1000.  The use of ‘for’ loops is limited to iterating through epochs, mini batches and for creating the mini batches itself. All other code is vectorized. Yet, it seems to crawl. Most likely the use of ‘lists’ in R, to return multiple values is performance intensive. Some day, I will try to profile the code, and see where the issue is. However the code works!

Having said that, the Confusion Matrix in R dumps a lot of interesting statistics! There is a bunch of statistical measures for each class. For e.g. the Balanced Accuracy for the digits ‘6’ and ‘9’ is around 50%. Looks like, the classifier is confused by the fact that 6 is inverted 9 and vice-versa. The accuracy on the Test data set is just around 75%. I could have played around with the number of layers, number of hidden units, learning rates, epochs etc to get a much higher accuracy. But since each test took about 8+ hours, I may work on this, some other day!

source("DLfunctions5.R")
source("mnist.R")
show_digit(train$x[2,]) #Set the layer dimensions layersDimensions=c(784, 15,9, 10) # Works at 1500 x <- t(train$x)
X <- x[,1:60000]

#### 2.2 Multi-class classification with Softmax – R code

The spiral data set created with Python was saved, and is used as the input with R code. The R Neural Network seems to perform much,much slower than both Python and Octave. Not sure why! Incidentally the computation of loss and the softmax derivative are identical for both R and Octave. yet R is much slower. To compute the softmax derivative I create matrices for the One Hot Encoded yi and then stack them before subtracting pi-yi. I am sure there is a more elegant and more efficient way to do this, much like Python. Any suggestions?

library(ggplot2)
library(dplyr)
library(RColorBrewer)
source("DLfunctions41.R")
Z1=data.frame(Z)
#Plot the dataset
ggplot(Z1,aes(x=V1,y=V2,col=V3)) +geom_point() +
scale_colour_gradientn(colours = brewer.pal(10, "Spectral"))

# Setup the data
X <- Z[,1:2]
y <- Z[,3]
X1 <- t(X)
Y1 <- t(y)

# Initialize number of features, number of hidden units in hidden layer and
# number of classes
numFeats<-2 # No features
numHidden<-100 # No of hidden units
numOutput<-3 # No of classes

# Initialize model
parameters <-initializeModel(numFeats, numHidden,numOutput)

W1 <-parameters[['W1']]
b1 <-parameters[['b1']]
W2 <-parameters[['W2']]
b2 <-parameters[['b2']]

# Set the learning rate
learningRate <- 0.5
# Initialize losses
losses <- NULL
for(i in 0:9000){

# Forward propagation through hidden layer with Relu units
retvals <- layerActivationForward(X1,W1,b1,'relu')
A1 <- retvals[['A']]
cache1 <- retvals[['cache']]
forward_cache1 <- cache1[['forward_cache1']]
activation_cache <- cache1[['activation_cache']]

# Forward propagation through output layer with Softmax units
retvals = layerActivationForward(A1,W2,b2,'softmax')
A2 <- retvals[['A']]
cache2 <- retvals[['cache']]
forward_cache2 <- cache2[['forward_cache1']]
activation_cache2 <- cache2[['activation_cache']]

# No oftraining examples
numTraining <- dim(X)[1]
dA <-0

# Select the elements where the y values are 0, 1 or 2 and make a vector
a=c(A2[y==0,1],A2[y==1,2],A2[y==2,3])
# Take log
correct_probs = -log(a)
# Compute loss
loss= sum(correct_probs)/numTraining

if(i %% 1000 == 0){
sprintf("iteration %d: loss %f",i, loss)
print(loss)
}
# Backward propagation through output layer with Softmax units
retvals = layerActivationBackward(dA, cache2, y, activationFunc='softmax')
dA1 = retvals[['dA_prev']]
dW2= retvals[['dW']]
db2= retvals[['db']]
# Backward propagation through hidden layer with Relu units
retvals = layerActivationBackward(t(dA1), cache1, y, activationFunc='relu')
dA0 = retvals[['dA_prev']]
dW1= retvals[['dW']]
db1= retvals[['db']]

# Update parameters
W1 <- W1 - learningRate * dW1
b1 <- b1 - learningRate * db1
W2 <- W2 - learningRate * t(dW2)
b2 <- b2 - learningRate * t(db2)
}
## [1] 1.212487
## [1] 0.5740867
## [1] 0.4048824
## [1] 0.3561941
## [1] 0.2509576
## [1] 0.7351063
## [1] 0.2066114
## [1] 0.2065875
## [1] 0.2151943
## [1] 0.1318807

#Create iterations
iterations <- seq(0,10)
#df=data.frame(iterations,losses)
ggplot(df,aes(x=iterations,y=losses)) + geom_point() + geom_line(color="blue") +
ggtitle("Losses vs iterations") + xlab("Iterations") + ylab("Loss")

plotDecisionBoundary(Z,W1,b1,W2,b2)

Multi-class Confusion Matrix

library(caret)
library(e1071)

# Forward propagation through hidden layer with Relu units
retvals <- layerActivationForward(X1,W1,b1,'relu')
A1 <- retvals[['A']]

# Forward propagation through output layer with Softmax units
retvals = layerActivationForward(A1,W2,b2,'softmax')
A2 <- retvals[['A']]
yhat <- apply(A2, 1,which.max) -1
Confusion Matrix and Statistics
Reference
Prediction  0  1  2
0 97  0  1
1  2 96  4
2  1  4 95

Overall Statistics
Accuracy : 0.96
95% CI : (0.9312, 0.9792)
No Information Rate : 0.3333
P-Value [Acc > NIR] : <2e-16

Kappa : 0.94
Mcnemar's Test P-Value : 0.5724
Statistics by Class:

Class: 0 Class: 1 Class: 2
Sensitivity            0.9700   0.9600   0.9500
Specificity            0.9950   0.9700   0.9750
Pos Pred Value         0.9898   0.9412   0.9500
Neg Pred Value         0.9851   0.9798   0.9750
Prevalence             0.3333   0.3333   0.3333
Detection Rate         0.3233   0.3200   0.3167
Detection Prevalence   0.3267   0.3400   0.3333
Balanced Accuracy      0.9825   0.9650   0.9625


My book “Practical Machine Learning with R and Python” includes the implementation for many Machine Learning algorithms and associated metrics. Pick up your copy today!

#### 2.3 Multi-class classification with Softmax – Octave code

A 2 layer Neural network with the Softmax activation unit in the output layer is constructed in Octave. The same spiral data set is used for Octave also
 source("DL41functions.m") # Read the spiral data data=csvread("spiral.csv"); # Setup the data X=data(:,1:2); Y=data(:,3); # Set the number of features, number of hidden units in hidden layer and number of classes numFeats=2; #No features numHidden=100; # No of hidden units numOutput=3; # No of classes # Initialize model [W1 b1 W2 b2] = initializeModel(numFeats,numHidden,numOutput); # Initialize losses losses=[] #Initialize learningRate learningRate=0.5; for k =1:10000 # Forward propagation through hidden layer with Relu units [A1,cache1 activation_cache1]= layerActivationForward(X',W1,b1,activationFunc ='relu'); # Forward propagation through output layer with Softmax units [A2,cache2 activation_cache2] = layerActivationForward(A1,W2,b2,activationFunc='softmax'); # No of training examples numTraining = size(X)(1); # Select rows where Y=0,1,and 2 and concatenate to a long vector a=[A2(Y==0,1) ;A2(Y==1,2) ;A2(Y==2,3)]; #Select the correct column for log prob correct_probs = -log(a); #Compute log loss loss= sum(correct_probs)/numTraining; if(mod(k,1000) == 0) disp(loss); losses=[losses loss]; endif dA=0; # Backward propagation through output layer with Softmax units [dA1 dW2 db2] = layerActivationBackward(dA, cache2, activation_cache2,Y,activationFunc='softmax'); # Backward propagation through hidden layer with Relu units [dA0,dW1,db1] = layerActivationBackward(dA1', cache1, activation_cache1, Y, activationFunc='relu'); #Update parameters W1 += -learningRate * dW1; b1 += -learningRate * db1; W2 += -learningRate * dW2'; b2 += -learningRate * db2'; endfor # Plot Losses vs Iterations iterations=0:1000:9000 plotCostVsIterations(iterations,losses) # Plot the decision boundary plotDecisionBoundary( X,Y,W1,b1,W2,b2)

The code for the Python, R and Octave implementations can be downloaded from Github at Deep Learning – Part 4

#### Conclusion

In this post I have implemented a 2 layer Neural Network with the Softmax classifier. In Part 3, I implemented a multi-layer Deep Learning Network. I intend to include the Softmax activation unit into the generalized multi-layer Deep Network along with the other activation units of sigmoid,tanh and relu.

Stick around, I’ll be back!!
Watch this space!

To see all post click Index of posts