Model selection for support vector machines using an asynchronous parallel evolution strategy

Published 1 January 2003

Thomas Philip Rúnarsson, Sven Sigurðsson

DOI10.1109/icnnsp.2003.1279319

Citations3

TL;DR

A new non-blocking asynchronous ES is developed for model selection for support vector machines optimizing a number of heuristic bounds on the expected generalization error.

Abstract

The application of a parallel evolutionary algorithm (ES) to model selection for support vector machines is examined. The problem of model selection is a computationally intense non-convex optimization problem. For this reason a parallel search strategy is desirable. A new non-blocking asynchronous ES is developed for this task. The algorithm is tested on five standard test sets optimizing a number of heuristic bounds on the expected generalization error.

Keywords

Computer Science

ACM Transactions on Intelligent Systems and TechnologyLIBSVM

41,340 Citations2011Chih-Chung Chang, Chih‐Jen Lin

Issues such as solving SVM optimization problems theoretical convergence multiclass classification probability estimates and parameter selection are discussed in detail.

Evolution and Optimum Seeking

2,200 Citations1995Hans–Paul Schwefel

Problems and Methods of Optimization Hill Climbing Strategies Random Strategies Evolution Strategies for Numerical Optimization Comparison of Direct Search Strategies for Parameter Optimization.

Machine LearningChoosing Multiple Parameters for Support Vector Machines

2,182 Citations2002Olivier Chapelle, Vladimir Vapnik +2 more

The problem of automatically tuning multiple parameters for pattern recognition Support Vector Machines (SVMs) is considered by minimizing some estimates of the generalization error of SVMs using a gradient descent algorithm over the set of parameters.

View PDF

Natural computing seriesThe Theory of Evolution Strategies

863 Citations2001Hans-Georg Beyer

Lecture notes in computer scienceEvolutionary Programming VII

429 Citations1998Porto, V. W, Eiben, A. E +2 more

Neural ComputationRadius Margin Bounds for Support Vector Machines with the RBF Kernel

219 Citations2003Kai-Min Chung, Wei‐Chun Kao +3 more

Foundations of genetic algorithmsGeneration Gaps Revisited

136 Citations1993Kenneth De Jong, Jayshree Sarma

Recent theoretical and empirical results provide the background for a much clearer understanding of the pros/cons of overlapping generations, and a review, combine, and extend these results in a way that significantly sharpens insight.

Evolutionary programming

40 Citations2002Yuanwei Cao, Qian Wu

The MVEP provides an improvement in global search and convergence performance in a mixed-variable space and an approach to handle various kinds of variables and constraints is discussed.

Continuous selection and self-adaptive evolution strategies

12 Citations2003Thomas Philip Rúnarsson, Xin Yao

The intention of this work is to eliminate the need for a synchronous generation scheme in the (/spl mu//sup /spl plusmn///spl lambda/) evolution strategy by considering how this gain in efficiency may influence the overall effectiveness of the evolution strategy.