Preface

Deep learning is advancing at a rapid pace, both in terms of its constituent neural network architectures, as well as their applications in real-world scenarios. This book takes you from the basics of building a neural network to the development of multiple advanced architectures that are used in various applications. You will find this book is divided into five parts.

In the first part, you will learn about how a neural network functions by building its components from scratch in Python before building them in Keras. Further, you will learn about the impact of various hyperparameters on a network's accuracy, in addition to learning about the flexibility of leveraging neural networks for multiple applications in various domains.

In the second part, you will learn about building a Convolutional Neural Network (CNN) from scratch in Python before leveraging it for image classification, where you will be learning about building a model to detect gender of persons in image and also to identify facial key points on the image of a face. Furthermore, you will be learning about the power of transfer learning in object detection and localization exercises to classify objects in image and also to identify the location of a person in image. Additionally, you will also be learning about the various applications of image analysis in self-driving car applications, leveraging semantic segmentation and more.

In the third part, we will pivot from image to text analysis by learning about encoding input both for image and text data, so that we can group similar images and similar bodies of text together using Autoencoders and Word vectors respectively. Additionally, you will learn about the various modeling aspects of building a recommender system so that you can recommend relevant movies to a user. You will also learn about leveraging Generative Adversarial Networks (GANs) for generating new images, as well as generating artistic images while learning about adversarial attacks to fool a network.

In the fourth part, you will dive deep into text analysis, where you will be learning about Recurrent Neural Networks (RNNs) and long short-term memory (LSTM) networks by building them from scratch in Python and then progressing toward building multiple use cases that leverage text analysis, such as stock price prediction, sentiment classification, machine translation, and building a chatbot, using advanced neural network architectures such as bi-directional LSTMs and attention mechanisms.

In the final part, you will be learning about end-to-end learning, where you will be transcribing images and audio along with generating captions. You will also be learning about Deep Q-learning, where you will build agents to play various Atari games.

By the end of this book, you will have developed the skills necessary for being able to apply various deep learning architectures to a majority of the deep learning problems that you might come across.