Maximum Entropy Monte-Carlo Planning

Authors: Chenjun Xiao, Ruitong Huang, Jincheng Mei, Dale Schuurmans, Martin Müller

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experimental results also demonstrate that MENTS is more sample efficient than UCT in both synthetic problems and Atari 2600 games.
Researcher Affiliation Collaboration Chenjun Xiao1 Jincheng Mei1 Ruitong Huang2 Dale Schuurmans1 Martin M uller1 1University of Alberta 2Borealis AI
Pseudocode No Section 4.1 “Algorithmic Design” describes the steps of MENTS, but it does not include a clearly labeled pseudocode or algorithm block.
Open Source Code No The paper does not contain any explicit statements or links indicating that the source code for the described methodology is publicly available.
Open Datasets Yes We then test MENTS on five Atari games: Beam Rider, Breakout, Q*bert, Seaquest and Space Invaders.
Dataset Splits No The paper mentions training a DQN model, but it does not provide specific details about dataset splits (e.g., train/validation/test percentages or counts) for the experiments conducted with MENTS or UCT. It refers to an Appendix for setup details, which is not provided.
Hardware Specification No The paper does not provide specific hardware details (such as exact GPU/CPU models, processor types, or memory amounts) used for running the experiments.
Software Dependencies No The paper mentions using a “vanilla DQN” but does not specify any software dependencies (e.g., libraries or solvers) with version numbers that would be needed to replicate the experiment setup.
Experiment Setup Yes The temperature is set to 0.1. At each time step we use 500 simulations to generate a move. The UCT algorithm adopts the following tree-policy introduced in Alpha Go [13], PUCT(s, a) = Q(s, a) + c P(s, a) / (b N(s, b) + 1 + N(s, a)).