SEEK: Sequential Knockoffs for Variable Selection in Reinforcement Learning

A 1.2 MB PDF authored by Tao Ma, last updated May 2026, describing the SEEK algorithm for identifying minimal sufficient states in Markov decision processes. The method is designed for systems with high-dimensional, complex nonlinear dynamics and includes a Python implementation. Empirical experiments compare its variable selection accuracy and policy suboptimality gap against competing methods.

Use Cases

• Estimating minimal sufficient states in Markov decision processes based on the described sequential knockoffs algorithm.
• Improving policy learning efficiency by reducing state dimensionality, as mentioned in the description.
• Benchmarking variable selection methods for RL against the accuracy and suboptimality results provided.

Strengths

• Includes a Python implementation of the SEEK algorithm, accessible via a provided GitHub link.
• The method is described as agnostic to the downstream reinforcement learning algorithm, potentially increasing its utility.

Limitations

• The dataset is a 1.2 MB PDF; the actual data used in the empirical experiments is not included.
• Row count and column-level documentation are absent, limiting suitability assessment for direct data analysis.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

figshare

Collection Method

Likely contains results from empirical experiments verifying theoretical results for the proposed algorithm.

Time Range

null

Freshness

Last updated 2026-05-01 15:00:06; freshness should be verified.

Geography

null