Efficient Transductive Online Learning via Randomized Rounding

TL;DR

This paper introduces a novel online learning algorithm for transductive settings that employs randomized rounding, enabling efficient solutions for collaborative filtering and bridging batch and online learning.

Contribution

It presents the first computationally efficient transductive online algorithm using randomized rounding, applicable to collaborative filtering and connecting batch and online learning.

Findings

Developed an efficient online algorithm for collaborative filtering with trace-norm constraints.

Provided the first solution linking batch learning and transductive online learning.

Demonstrated the effectiveness of the approach in theoretical and practical scenarios.

Abstract

Most traditional online learning algorithms are based on variants of mirror descent or follow-the-leader. In this paper, we present an online algorithm based on a completely different approach, tailored for transductive settings, which combines "random playout" and randomized rounding of loss subgradients. As an application of our approach, we present the first computationally efficient online algorithm for collaborative filtering with trace-norm constrained matrices. As a second application, we solve an open question linking batch learning and transductive online learning