Intel Xeon Phi Processor High Performance Programming, 2nd Edition

Chapter 24

Machine learning

Abstract

Focuses on mapping and executing machine learning algorithms on to Knights Landing. Discusses two specific algorithms that exercise different computational and memory access patterns and showcase the wide range of algorithms that can run on Knights Landing: Convolutional Neural Networks (CNNs), and k-Nearest Neighbors (KNN).

Keywords

MCDRAM; k-Nearest neighbors; Convolutional neural networks; kd-tree

What is new with Knights Landing in this chapter?

AVX-512 and MCDRAM boost performance

Machine learning spans a broad set of algorithms that are used to extract useful models from raw data. These models in turn are used in a variety of mining tasks (where we extract one or more instances of a specific model ...