On Local Regret
Abstract
Online learning aims to perform nearly as well as the best hypothesis in hindsight. For some hypothesis classes, though, even finding the best hypothesis offline is challenging. In such offline cases, local search techniques are often employed and only local optimality guaranteed. For online decision-making with such hypothesis classes, we introduce local regret, a generalization of regret that aims to perform nearly as well as only nearby hypotheses. We then present a general algorithm to minimize local regret with arbitrary locality graphs. We also show how the graph structure can be exploited to drastically speed learning. These algorithms are then demonstrated on a diverse set of online problems: online disjunct learning, online Max-SAT, and online decision tree learning.
Cite
@article{arxiv.1206.3318,
title = {On Local Regret},
author = {Michael Bowling and Martin Zinkevich},
journal= {arXiv preprint arXiv:1206.3318},
year = {2012}
}
Comments
This is the longer version of the same-titled paper appearing in the Proceedings of the Twenty-Ninth International Conference on Machine Learning (ICML), 2012