Fast and Scalable Structural SVM with Slack Rescaling

TL;DR

This paper introduces an efficient method for training slack-rescaled structural SVMs by approximating the most-violating-label, significantly reducing training time and enabling scalable, accurate learning.

Contribution

It presents a novel approach to efficiently find the most-violating-label in slack-rescaled SVMs using an oracle from margin-rescaled formulations, improving scalability.

Findings

Reduces training runtime by an order of magnitude

Enables scalable and accurate slack-rescaled SVM training

Provides a practical method for complex structured prediction tasks

Abstract

We present an efficient method for training slack-rescaled structural SVM. Although finding the most violating label in a margin-rescaled formulation is often easy since the target function decomposes with respect to the structure, this is not the case for a slack-rescaled formulation, and finding the most violated label might be very difficult. Our core contribution is an efficient method for finding the most-violating-label in a slack-rescaled formulation, given an oracle that returns the most-violating-label in a (slightly modified) margin-rescaled formulation. We show that our method enables accurate and scalable training for slack-rescaled SVMs, reducing runtime by an order of magnitude compared to previous approaches to slack-rescaled SVMs.