Scalable Kernel Learning via the Discriminant Information

TL;DR

This paper introduces a scalable supervised kernel learning approach that optimizes kernel feature maps using the Discriminant Information criterion, enhancing class separability and improving performance over existing methods.

Contribution

It generalizes the Discriminant Information measure for broader kernel learning applications, enabling scalable and discriminant-rich kernel feature optimization.

Findings

Improved optimization and generalization on multiple datasets.

Enhanced class separability with the proposed kernel learning method.

Outperforms state-of-the-art kernel learning techniques.

Abstract

Kernel approximation methods create explicit, low-dimensional kernel feature maps to deal with the high computational and memory complexity of standard techniques. This work studies a supervised kernel learning methodology to optimize such mappings. We utilize the Discriminant Information criterion, a measure of class separability with a strong connection to Discriminant Analysis. By generalizing this measure to cover a wider range of kernel maps and learning settings, we develop scalable methods to learn kernel features with high discriminant power. Experimental results on several datasets showcase that our techniques can improve optimization and generalization performances over state of the art kernel learning methods.