Machine Learning for Adaptive Many-Core Machines: A Practical Approach

The overwhelming data produced everyday and the increasing performance and cost requirements of applications are transversal to a wide range of activities in society, from science to industry. In particular, the magnitude and complexity of the tasks that Machine Learning (ML) algorithms have to solve are driving the need to devise adaptive many-core machines that scale well with the volume of data, or in other words, can handle Big Data.

This book gives a concise view on how to extend the applicability of well-known ML algorithms in Graphics Processing Unit (GPU) with data scalability in mind. It presents a series of new techniques to enhance, scale and distribute data in a Big Learning framework. It is not intended to be a comprehensive survey of the state of the art of the whole field of machine learning for Big Data. Its purpose is less ambitious and more practical: to explain and illustrate existing and novel GPU-based ML algorithms, not viewed as a universal solution for the Big Data challenges but rather as part of the answer, which may require the use of different strategies coupled together.

Table of Contents

Part I Introduction
Chapter 1 Motivation and Preliminaries
Chapter 2 GPU Machine Learning Library (GPUMLib)

Part II Supervised Learning
Chapter 3 Neural Networks
Chapter 4 Handling Missing Data
Chapter 5 Support Vector Machines (SVMs)
Chapter 6 Incremental Hypersphere Classifier (IHC)

Part III Unsupervised and Semi-supervised Learning
Chapter 7 Non-Negative Matrix Factorization (NMF)
Chapter 8 Deep Belief Networks (DBNs)

Part IV Large-Scale Machine Learning
Chapter 9 Adaptive Many-Core Machines

Appendix A Experimental Setup and Performance Evaluation