Reinforcement Learning Model Fitting for Bandit Behavioral Data

A 2026 research paper by Hao Zhu introduces a convex optimization method for fitting reinforcement learning models to behavioral data in multi-armed bandit environments. The work includes an open-source Python package and is evaluated in simulated and real-world environments.

Use Cases

• Implement the provided convex relaxation method to fit RL models to behavioral data from multi-armed bandit experiments.
• Use the open-source Python package to analyze decision-making behavior in simulated bandit environments.
• Benchmark the proposed fitting method's computation time against state-of-the-art techniques from the literature.

Strengths

• Includes an open-source Python package for direct application by researchers.
• Theoretical analysis provides convexity properties for a wide range of RL models.
• Method evaluated in several simulated and real-world bandit environments.

Limitations

• The dataset is a 213.6 KB PDF document, containing a research paper rather than structured behavioral data.
• No raw behavioral data, column definitions, or sample records are provided for direct machine learning use.

Provenance

Source

figshare, author Hao Zhu.

Collection Method

Research paper detailing a novel computational method.

Time Range

null

Freshness

Last updated March 2026.

Geography

null