Non-Stationary Off-Policy Optimization

TL;DR

This paper introduces a novel method for off-policy optimization in non-stationary environments, combining offline data partitioning and adaptive online policy switching, with theoretical guarantees and empirical validation.

Contribution

It proposes a two-phase approach for off-policy optimization in piecewise-stationary settings, with guarantees on policy quality and regret, advancing the handling of non-stationarity in off-policy learning.

Findings

Outperforms state-of-the-art baselines on synthetic datasets.

Effective in real-world non-stationary environments.

Provides theoretical guarantees on policy quality and regret.

Abstract

Off-policy learning is a framework for evaluating and optimizing policies without deploying them, from data collected by another policy. Real-world environments are typically non-stationary and the offline learned policies should adapt to these changes. To address this challenge, we study the novel problem of off-policy optimization in piecewise-stationary contextual bandits. Our proposed solution has two phases. In the offline learning phase, we partition logged data into categorical latent states and learn a near-optimal sub-policy for each state. In the online deployment phase, we adaptively switch between the learned sub-policies based on their performance. This approach is practical and analyzable, and we provide guarantees on both the quality of off-policy optimization and the regret during online deployment. To show the effectiveness of our approach, we compare it to state-of-the-art baselines on both synthetic and real-world datasets. Our approach outperforms methods that act only on observed context.