Deterministic Policies for Constrained Reinforcement Learning in Polynomial Time

Authors: Jeremy McMahan

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical Our paper is purely theoretical and does not include any experiments. We present a novel algorithm that efficiently computes near-optimal deterministic policies for constrained reinforcement learning (CRL) problems. Our approach combines three key ideas: (1) value-demand augmentation, (2) action-space approximate dynamic programming, and (3) time-space rounding. Our algorithm constitutes a fully polynomial-time approximation scheme (FPTAS) for any time-space recursive (TSR) cost criteria.
Researcher Affiliation Academia Jeremy Mc Mahan University of Wisconsin-Madison jmcmahan@wisc.edu
Pseudocode Yes Algorithm 1 Reduction to RL, Algorithm 2 Augmented interaction, Algorithm 3 Approx Bellman Update, Algorithm 4 Approx Solve, Algorithm 5 Approximation Scheme, Algorithm 6 Approx Solve.
Open Source Code No Our paper is purely theoretical and does not include any experiments. (Neur IPS Paper Checklist, Q5)
Open Datasets No Our paper is purely theoretical and does not include any experiments. (Neur IPS Paper Checklist, Q4)
Dataset Splits No Our paper is purely theoretical and does not include any experiments. (Neur IPS Paper Checklist, Q4)
Hardware Specification No Our paper is purely theoretical and does not include any experiments. (Neur IPS Paper Checklist, Q4)
Software Dependencies No Our paper is purely theoretical and does not include any experiments. (Neur IPS Paper Checklist, Q4)
Experiment Setup No Our paper is purely theoretical and does not include any experiments. (Neur IPS Paper Checklist, Q4)