A Comparison of Support Vector Machines Training GPU-Accelerated Open Source Implementations

TL;DR

This paper benchmarks various open-source GPU-accelerated SVM training implementations, providing a fair comparison across different hardware and datasets, and recommends the best options for dense and sparse data.

Contribution

It introduces a comprehensive benchmark for GPU-accelerated SVM implementations, including modifications for cross-platform compatibility and evaluation on popular datasets.

Findings

Identified top-performing implementations for dense datasets

Identified top-performing implementations for sparse datasets

Provided recommendations for practical use cases

Abstract

Last several years, GPUs are used to accelerate computations in many computer science domains. We focused on GPU accelerated Support Vector Machines (SVM) training with non-linear kernel functions. We had searched for all available GPU accelerated C++ open-source implementations and created an open-source C++ benchmark project. We modifed all the implementations to run on actual hardware and software and in both Windows and Linux operating systems. The benchmark project offers making a fair and direct comparison of the individual implementations under the same conditions, datasets, and hardware. In addition, we selected the most popular datasets in the community and tested them. Finally, based on the evaluation, we recommended the best-performing implementations for dense and sparse datasets.