Exponentiated Gradient LINUCB for Contextual Multi-Armed Bandits

TL;DR

This paper introduces Exponentiated Gradient LINUCB, an algorithm for contextual multi-armed bandits that optimizes exploration using Exponentiated Gradient, showing superior performance over existing methods in offline simulations.

Contribution

The paper proposes a novel combination of Exponentiated Gradient with LINUCB for improved exploration in contextual bandits, validated through real data simulations.

Findings

Outperforms surveyed algorithms in offline evaluations

Demonstrates effective exploration optimization

Validated with real online event log data

Abstract

We present Exponentiated Gradient LINUCB, an algorithm for con-textual multi-armed bandits. This algorithm uses Exponentiated Gradient to find the optimal exploration of the LINUCB. Within a deliberately designed offline simulation framework we conduct evaluations with real online event log data. The experimental results demonstrate that our algorithm outperforms surveyed algorithms.