Online Least Squares Estimation with Self-Normalized Processes: An Application to Bandit Problems

Abstract

The analysis of online least squares estimation is at the heart of many stochastic sequential decision making problems. We employ tools from the self-normalized processes to provide a simple and self-contained proof of a tail bound of a vector-valued martingale. We use the bound to construct a new tighter confidence sets for the least squares estimate. We apply the confidence sets to several online decision problems, such as the multi-armed and the linearly parametrized bandit problems. The confidence sets are potentially applicable to other problems such as sleeping bandits, generalized linear bandits, and other linear control problems. We improve the regret bound of the Upper Confidence Bound (UCB) algorithm of Auer et al. (2002) and show that its regret is with high-probability a problem dependent constant. In the case of linear bandits (Dani et al., 2008), we improve the problem dependent bound in the dimension and number of time steps. Furthermore, as opposed to the previous result, we prove that our bound holds for small sample sizes, and at the same time the worst case bound is improved by a logarithmic factor and the constant is improved.

Submitted to the 24th Annual Conference on Learning Theory (COLT 2011)

References in corpus (1)

• On Upper-Confidence Bound Policies for Non-Stationary Bandit Problems

Cited by in corpus (19)

• On the Complexity of Best Arm Identification in Multi-Armed Bandit Models
• On the Sample Complexity of the Linear Quadratic Regulator
• Informational Confidence Bounds for Self-Normalized Averages and Applications
• Improper Learning for Non-Stochastic Control
• Logarithmic Regret Bound in Partially Observable Linear Dynamical Systems
• Learning Linear Dynamical Systems with Semi-Parametric Least Squares
• Active Learning for Nonlinear System Identification with Guarantees
• Finite-time Analysis of Approximate Policy Iteration for the Linear Quadratic Regulator
• Martingale Methods for Sequential Estimation of Convex Functionals and Divergences
• Structured Linear Contextual Bandits: A Sharp and Geometric Smoothed Analysis
• Accurate Inference for Adaptive Linear Models
• Learning Stabilizing Controllers for Unstable Linear Quadratic Regulators from a Single Trajectory
• Finite Time Identification in Unstable Linear Systems
• Near-optimal Offline and Streaming Algorithms for Learning Non-Linear Dynamical Systems
• Exact Asymptotics for Linear Quadratic Adaptive Control
• Regret Bounds for LQ Adaptive Control Under Database Attacks (Extended Version)
• Statistical Consistency of Set-Membership Estimator for Linear Systems
• Learning in Networked Control Systems
• Efficient Episodic Learning of Nonstationary and Unknown Zero-Sum Games Using Expert Game Ensembles