Book Image

Deep Learning for Beginners

By : Dr. Pablo Rivas
Book Image

Deep Learning for Beginners

By: Dr. Pablo Rivas

Overview of this book

With information on the web exponentially increasing, it has become more difficult than ever to navigate through everything to find reliable content that will help you get started with deep learning. This book is designed to help you if you're a beginner looking to work on deep learning and build deep learning models from scratch, and you already have the basic mathematical and programming knowledge required to get started. The book begins with a basic overview of machine learning, guiding you through setting up popular Python frameworks. You will also understand how to prepare data by cleaning and preprocessing it for deep learning, and gradually go on to explore neural networks. A dedicated section will give you insights into the working of neural networks by helping you get hands-on with training single and multiple layers of neurons. Later, you will cover popular neural network architectures such as CNNs, RNNs, AEs, VAEs, and GANs with the help of simple examples, and learn how to build models from scratch. At the end of each chapter, you will find a question and answer section to help you test what you've learned through the course of the book. By the end of this book, you'll be well-versed with deep learning concepts and have the knowledge you need to use specific algorithms with various tools for different tasks.
Table of Contents (20 chapters)
1
Section 1: Getting Up to Speed
8
Section 2: Unsupervised Deep Learning
13
Section 3: Supervised Deep Learning

Examining the VAE model

The VAE is a particular type of autoencoder (Kingma, D. P., & Welling, M. (2013)). It learns specific statistical properties of the dataset derived from a Bayesian approach. First, let's define as the prior probability density function of a random latent variable, . Then, we can describe a conditional probability density function, , which can be interpreted as a model that can produce data—say, . It follows that we can approximate the posterior probability density function in terms of the conditional and prior distributions, as follows:

It turns out that an exact posterior is intractable, but this problem can be solved, approximately, by making a few assumptions and using an interesting idea to compute gradients. To begin with, the prior can be assumed to follow an isotropic Gaussian distribution, . We can also assume that the conditional distribution, , can be parametrized and modeled using a neural network; that is, given a latent vector , we...