A GPU Implementation of a Bat Algorithm Trained Neural Network

In recent times, there has been an exponential growth in the viability of Neural Networks (NN) as a Machine Learning tool. Most standard training algorithms for NNs, like gradient descent and its variants fall prey to local optima. Metaheuristics have been found to be a viable alternative to traditi...

Full description

Saved in:
Bibliographic Details
Published inNeural Information Processing Vol. 9949; pp. 326 - 334
Main Authors Choudhury, Amit Roy, Jain, Rishabh, Sharma, Kapil
Format Book Chapter
LanguageEnglish
Published Switzerland Springer International Publishing AG 2016
Springer International Publishing
SeriesLecture Notes in Computer Science
Subjects
Online AccessGet full text
ISBN3319466747
9783319466743
ISSN0302-9743
1611-3349
DOI10.1007/978-3-319-46675-0_36

Cover

More Information
Summary:In recent times, there has been an exponential growth in the viability of Neural Networks (NN) as a Machine Learning tool. Most standard training algorithms for NNs, like gradient descent and its variants fall prey to local optima. Metaheuristics have been found to be a viable alternative to traditional training methods. Among these metaheuristics the Bat Algorithm (BA), has been shown to be superior. Even though BA promises better results, yet being a population based metaheuristic, it forces us to involve many Neural Networks and evaluate them on nearly every iteration. This makes the already computationally expensive task of training a NN even more so. To overcome this problem, we exploit the inherent concurrent characteristics of both NNs as well as BA to design a framework which utilizes the massively parallel architecture of Graphics Processing Units (GPUs). Our framework is able to offer speed-ups of upto 47 $$\times $$ depending on the architecture of the NN.
Bibliography:Original Abstract: In recent times, there has been an exponential growth in the viability of Neural Networks (NN) as a Machine Learning tool. Most standard training algorithms for NNs, like gradient descent and its variants fall prey to local optima. Metaheuristics have been found to be a viable alternative to traditional training methods. Among these metaheuristics the Bat Algorithm (BA), has been shown to be superior. Even though BA promises better results, yet being a population based metaheuristic, it forces us to involve many Neural Networks and evaluate them on nearly every iteration. This makes the already computationally expensive task of training a NN even more so. To overcome this problem, we exploit the inherent concurrent characteristics of both NNs as well as BA to design a framework which utilizes the massively parallel architecture of Graphics Processing Units (GPUs). Our framework is able to offer speed-ups of upto 47\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times $$\end{document} depending on the architecture of the NN.
ISBN:3319466747
9783319466743
ISSN:0302-9743
1611-3349
DOI:10.1007/978-3-319-46675-0_36