Instance-optimal PAC Algorithms for Contextual Bandits

Authors: Zhaoqi Li, Lillian Ratliff, houssam nassif, Kevin G. Jamieson, Lalit Jain

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical We characterize the first instance-dependent PAC sample complexity of contextual bandits through a quantity , and provide matching upper and lower bounds in terms of for the agnostic and linear contextual best-arm identification settings. We show that no algorithm can be simultaneously minimax-optimal for regret minimization and instance-dependent PAC for best-arm identification. Our main result is a new instance-optimal and computationally efficient algorithm that relies on a polynomial number of calls to an argmax oracle.
Researcher Affiliation Collaboration Zhaoqi Li Department of Statistics University of Washington zli9@uw.edu Lillian Ratliff Department of Electrical and Computer Engineering University of Washington ratliffl@uw.edu Houssam Nassif Amazon houssamn@amazon.com Kevin Jamieson Allen School of Computer Science & Engineering University of Washington jamieson@cs.washington.edu Lalit Jain Foster School of Business University of Washington lalitj@uw.edu
Pseudocode Yes Algorithm 1 Elimination Contextual RAGE Algorithm 2 Non-elimination Contextual RAGE Algorithm 3 Contextual Oracle-efficient Dualized Algorithm (CODA)
Open Source Code No The paper does not provide a link to open-source code or explicitly state that code for the methodology is available. The checklist also states '[N/A]' for code reproduction.
Open Datasets No The paper is theoretical and does not conduct empirical experiments on a publicly available dataset. It makes an assumption about access to a large dataset of contexts D (Assumption 1) but does not specify a real-world public dataset with concrete access information. The checklist also states '[N/A]' for experiments.
Dataset Splits No The paper is theoretical and does not conduct empirical experiments, thus no training/validation/test splits are specified. The checklist also states '[N/A]' for experiments.
Hardware Specification No The paper is theoretical and does not describe any specific hardware used for experiments. The checklist also states '[N/A]' for experiments.
Software Dependencies No The paper is theoretical and does not describe any specific software dependencies with version numbers used for experiments. The checklist also states '[N/A]' for experiments.
Experiment Setup No The paper is theoretical and does not describe specific experimental setup details such as hyperparameters. The checklist also states '[N/A]' for experiments.