Efficient Contextual Bandits with Continuous Actions

TL;DR

This paper introduces a computationally efficient algorithm for contextual bandits with continuous actions, capable of handling unknown structures and compatible with various supervised learning models, validated through theoretical analysis and large-scale experiments.

Contribution

It presents a novel reduction-style algorithm for continuous action contextual bandits that is both computationally tractable and broadly applicable.

Findings

Proves the algorithm's effectiveness in a general setting

Demonstrates scalability with large-scale experiments

Shows compatibility with most supervised learning representations

Abstract

We create a computationally tractable algorithm for contextual bandits with continuous actions having unknown structure. Our reduction-style algorithm composes with most supervised learning representations. We prove that it works in a general sense and verify the new functionality with large-scale experiments.