Machine learning is fast becoming the preferred way to solve data problems, thanks to

the huge variety of mathematical algorithms that find patterns otherwise invisible to us.

Applied Deep Learning with PyTorch takes your understanding of deep learning, its

algorithms, and its applications to a higher level. The book begins by helping you

browse through the basics of deep learning and PyTorch. Once you are well versed

with the PyTorch syntax and capable of building a single-layer neural network,

you will gradually learn to tackle more complex data problems by configuring and

training a convolutional neural network (CNN) to perform image classification. As you

progress through the chapters, you'll discover how you can solve an NLP problem by

implementing a recurrent neural network (RNN).

By the end of this book, you'll be able to apply the skills and confidence you've gathered

along your learning process to use PyTorch for building deep learning solutions that

can solve your business data problems.

Table of Contents

Preface i

Introduction to Deep Learning and PyTorch 1

Introduction  

Understanding Deep Learning

Why Is Deep Learning Important? Why Has It Become Popular?  

Applications of Deep Learning  

PyTorch Introduction

Advantages  

Disadvantages  

What Are Tensors?  

Exercise 1: Creating Tensors of Different Ranks Using PyTorch

Key Elements of PyTorch  

Activity 1: Creating a Single-Layer Neural Network

Summary  

Building Blocks of Neural Networks 21

Introduction  

Introduction to Neural Networks  

What Are Neural Networks?  

Exercise 2: Performing the Calculations of a Perceptron  

Multi-Layer Perceptron  

The Learning Process of a Neural Network

Advantages and Disadvantages  

Introduction to Artificial Neural Networks  

Introduction to Convolutional Neural Networks

Introduction to Recurrent Neural Networks

Data Preparation  

Dealing with Messy Data  

Exercise 3: Dealing with Messy Data  

Data Rescaling  

Exercise 4: Rescaling Data

Splitting the Data  

Exercise 5: Splitting a Dataset

Activity 2: Performing Data Preparation

Building a Deep Neural Network  

Exercise 6: Building a Deep Neural Network Using PyTorch  

Activity 3: Developing a Deep Learning Solution

for a Regression Problem

Summary  

A Classification Problem Using DNNs 59

Introduction  

Problem Definition  

Deep Learning in Banking  

Exploring the Dataset  

Data Preparation  

Building the Model  

ANNs for Classification Tasks  

A Good Architecture

PyTorch Custom Modules  

Activity 4: Building an ANN  

Dealing with an Underfitted or Overfitted Model  

Error Analysis  

Exercise 7: Performing Error Analysis

Activity 5: Improving a Model's Performance  

Deploying Your Model  

Saving and Loading Your Model  

PyTorch for Production in C++  

Activity 6: Making Use of Your Model  

Summary  

Convolutional Neural Networks 95

Introduction  

Building a CNN  

Why CNNs?

The Inputs

Applications of CNNs

Building Blocks of CNNs  

Exercise 8: Calculating the Output Shape of a Convolutional Layer  

Exercise 9: Calculating the Output Shape of a set of Convolutional

and Pooling Layers  

Side Note – Downloading Datasets from PyTorch  

Activity 7: Building a CNN for an Image Classification Problem  

Data Augmentation  

Data Augmentation with PyTorch  

Activity 8: Implementing Data Augmentation  

Batch Normalization  

Batch Normalization with PyTorch

Activity 9: Implementing Batch Normalization

Summary  

Style Transfer 127

Introduction  

Style Transfer  

How Does It Work?  

Implementation of Style Transfer Using the VGG-19

Network Architecture  

Inputs: Loading and Displaying  

Exercise 10: Loading and Displaying Images  

Loading the Model  

Exercise 11: Loading a Pretrained Model in PyTorch  

Extracting the Features .

Exercise 12: Setting up the Feature Extraction Process  

The Optimization Algorithm, Losses, and Parameter Updates .

Exercise 13: Creating the Target Image .

Activity 10: Performing Style Transfer .

Summary  

Analyzing the Sequence of Data with RNNs 151

Introduction .

Recurrent Neural Networks

Applications of RNNs  

How Do RNNs Work? .

RNNs in PyTorch  

Activity 11: Using a Simple RNN for a Time Series Prediction .

Long Short-Term Memory Networks (LSTMs)  

Applications .

How Do LSTM Networks Work?  

LSTM Networks in PyTorch  

Preprocessing the Input Data .

One-Hot Encoding  

Building the Architecture .

Training the Model

Performing Predictions  

Activity 12: Text Generation with LSTM Networks  

Natural Language Processing (NLP) .

Sentiment Analysis .

Sentiment Analysis in PyTorch  

Preprocessing the Input Data .

Building the Architecture .

Training the Model ..

Activity 13: Performing NLP for Sentiment Analysis  

Summary .