Online Algorithms for Multiclass Classification using Partial Labels

TL;DR

This paper introduces new online algorithms for multiclass classification with partial labels, extending Perceptron and Pegasos methods, and demonstrates their effectiveness through theoretical bounds and experiments.

Contribution

It presents novel variants of Perceptron and Pegasos algorithms tailored for partial label data, along with theoretical mistake and regret bounds.

Findings

Proposed Avg Perceptron and Max Perceptron outperform standard Perceptron.

Proposed Avg Pegasos and Max Pegasos show improved regret bounds.

Experimental results validate the effectiveness of the new algorithms.

Abstract

In this paper, we propose online algorithms for multiclass classification using partial labels. We propose two variants of Perceptron called Avg Perceptron and Max Perceptron to deal with the partial labeled data. We also propose Avg Pegasos and Max Pegasos, which are extensions of Pegasos algorithm. We also provide mistake bounds for Avg Perceptron and regret bound for Avg Pegasos. We show the effectiveness of the proposed approaches by experimenting on various datasets and comparing them with the standard Perceptron and Pegasos.