Evolutionary Reinforcement Learning for Sample-Efficient Multiagent Coordination

Authors: Somdeb Majumdar, Shauharda Khadka, Santiago Miret, Stephen Mcaleer, Kagan Tumer

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Results demonstrate that MERL significantly outperforms state-of-the-art methods, such as MADDPG, on a number of difficult coordination benchmarks.
Researcher Affiliation Collaboration 1Intel Labs 2University of California, Irvine 3Oregon State University.
Pseudocode Yes Algorithm 1 provides a detailed pseudo-code of the MERL algorithm.
Open Source Code Yes Additionally, our source code 1 is available online. Footnote 1 provides the URL: https://tinyurl.com/y6erclts
Open Datasets Yes We adopt environments from (Lowe et al., 2017) and (Rahmattalabi et al., 2016) to perform our experiments. Each environment consists of multiple agents and landmarks in a two-dimensional world.
Dataset Splits No The paper does not provide specific train/validation/test dataset splits. It describes periodic testing on task instances but does not define a separate validation set split from the data.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., CPU, GPU models, memory, cloud instance types) used to run the experiments.
Software Dependencies No The paper mentions various algorithms like DDPG, TD3, and MADDPG, and environments, but it does not specify software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x).
Experiment Setup Yes The choice of hyperparameters is explained in the Appendix.