Book Image

TensorFlow 1.x Deep Learning Cookbook

Book Image

TensorFlow 1.x Deep Learning Cookbook

Overview of this book

Deep neural networks (DNNs) have achieved a lot of success in the field of computer vision, speech recognition, and natural language processing. This exciting recipe-based guide will take you from the realm of DNN theory to implementing them practically to solve real-life problems in the artificial intelligence domain. In this book, you will learn how to efficiently use TensorFlow, Google’s open source framework for deep learning. You will implement different deep learning networks, such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Deep Q-learning Networks (DQNs), and Generative Adversarial Networks (GANs), with easy-to-follow standalone recipes. You will learn how to use TensorFlow with Keras as the backend. You will learn how different DNNs perform on some popularly used datasets, such as MNIST, CIFAR-10, and Youtube8m. You will not only learn about the different mobile and embedded platforms supported by TensorFlow, but also how to set up cloud platforms for deep learning applications. You will also get a sneak peek at TPU architecture and how it will affect the future of DNNs. By using crisp, no-nonsense recipes, you will become an expert in implementing deep learning techniques in growing real-world applications and research areas such as reinforcement learning, GANs, and autoencoders.
Table of Contents (15 chapters)
14
TensorFlow Processing Units

Sparse autoencoder

The autoencoder that we saw in the previous recipe worked more like an identity network--they simply reconstruct the input. The emphasis is to reconstruct the image at the pixel level, and the only constraint is the number of units in the bottleneck layer; while it is interesting, pixel-level reconstruction does not ensure that the network will learn abstract features from the dataset. We can ensure that the network learns abstract features from the dataset by adding further constraints.

In sparse autoencoders, a sparse penalty term is added to the reconstruction error, which tries to ensure that fewer units in the bottleneck layer will fire at any given time. If m is the total number of input patterns, then we can define a quantity ρ_hat (you can check the mathematical details in Andrew Ng's Lecture at https://web.stanford.edu/class/cs294a/sparseAutoencoder_2011new...