Monte Carlo Tree Search With Iteratively Refining State Abstractions
Authors: Samuel Sokota, Caleb Y Ho, Zaheen Ahmad, J. Zico Kolter
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To support this claim, we present a series of experimental examples in which abstraction refining outperforms progressive widening, given equal simulation budgets. |
| Researcher Affiliation | Collaboration | Samuel Sokota Carnegie Mellon University ssokota@andrew.cmu.edu Caleb Ho Independent Researcher caleb.yh.ho@gmail.com Zaheen Ahmad University of Alberta zfahmad@ualberta.ca J. Zico Kolter Carnegie Mellon University zkolter@cs.cmu.edu |
| Pseudocode | Yes | Algorithm 1 Progressive Widening procedure SAMPLE(s, a) ... Algorithm 2 Abstraction Refining procedure SAMPLE(s, a) |
| Open Source Code | Yes | We included each of the codebases used for the experiments in the main body of the paper in the supplementary material and included descriptions of our experiments in the appendix. |
| Open Datasets | No | The paper describes using a 'continuous variant of blackjack', 'a variant of the trap problem', and 'five by five Go' with Open Spiel, but does not provide concrete access information (link, DOI, specific repository, or formal citation with authors/year for a public dataset) for any of these. |
| Dataset Splits | No | The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning. |
| Hardware Specification | No | We did not keep track of the total amount of compute used. We used an internal cluster. |
| Software Dependencies | No | The paper mentions 'Open Spiel' but does not provide specific version numbers for any software dependencies. |
| Experiment Setup | Yes | For k = 1, setting α to a small value... abstraction refining with ϵn = n 0.1. ... Abstraction refining with ϵn = 2n 0.1 is shown in brown... Abstraction refining with ϵn = 0.1n 0.1 is shown in brown... |