Book Image

Hands-On Deep Learning with Go

By : Gareth Seneque, Darrell Chua
Book Image

Hands-On Deep Learning with Go

By: Gareth Seneque, Darrell Chua

Overview of this book

Go is an open source programming language designed by Google for handling large-scale projects efficiently. The Go ecosystem comprises some really powerful deep learning tools such as DQN and CUDA. With this book, you'll be able to use these tools to train and deploy scalable deep learning models from scratch. This deep learning book begins by introducing you to a variety of tools and libraries available in Go. It then takes you through building neural networks, including activation functions and the learning algorithms that make neural networks tick. In addition to this, you'll learn how to build advanced architectures such as autoencoders, restricted Boltzmann machines (RBMs), convolutional neural networks (CNNs), recurrent neural networks (RNNs), and more. You'll also understand how you can scale model deployments on the AWS cloud infrastructure for training and inference. By the end of this book, you'll have mastered the art of building, training, and deploying deep learning models in Go to solve real-world problems.
Table of Contents (15 chapters)
Free Chapter
1
Section 1: Deep Learning in Go, Neural Networks, and How to Train Them
6
Section 2: Implementing Deep Neural Network Architectures
11
Section 3: Pipeline, Deployment, and Beyond!

Advanced gradient descent algorithms

Now that we have an understanding of SGD and backpropagation, let's look at a number of advanced optimization methods (building on SGD) that offer us some kind of advantage, usually an improvement in training time (or the time it takes to minimize the cost function to the point where our network converges).

These improved methods include a general notion of velocity as an optimization parameter. Quoting from Wibisono and Wilson, in the opening to their paper on Accelerated Methods in Optimization:

"In convex optimization, there is an acceleration phenomenon in which we can boost the convergence rate of certain gradient-based algorithms."

In brief, a number of these advanced algorithms all rely on a similar principle—that they can pass through local optima quickly, carried by their momentum—essentially, a moving...