CoRE Kernels

TL;DR

This paper introduces correlation-resemblance (CoRE) kernels for high-dimensional, sparse, non-binary data, demonstrating their effectiveness in classification and proposing hashing algorithms to enable scalable linear kernel methods.

Contribution

The paper proposes new nonlinear CoRE kernels for non-binary sparse data and develops probabilistic hashing algorithms to make them practical for large-scale applications.

Findings

CoRE kernels improve classification accuracy on sparse data.

Hashing algorithms enable scalable linear kernel approximations.

No tuning parameters are required for the kernels.

Abstract

The term "CoRE kernel" stands for correlation-resemblance kernel. In many applications (e.g., vision), the data are often high-dimensional, sparse, and non-binary. We propose two types of (nonlinear) CoRE kernels for non-binary sparse data and demonstrate the effectiveness of the new kernels through a classification experiment. CoRE kernels are simple with no tuning parameters. However, training nonlinear kernel SVM can be (very) costly in time and memory and may not be suitable for truly large-scale industrial applications (e.g. search). In order to make the proposed CoRE kernels more practical, we develop basic probabilistic hashing algorithms which transform nonlinear kernels into linear kernels.