Improving Multi-agent Coordination by Learning to Estimate Contention

Authors: Panayiotis Danassis, Florian Wiedemair, Boi Faltings

IJCAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We provide a thorough evaluation in a variety of synthetic benchmarks and a real-world meeting scheduling problem. In all of them ALMA-Learning is able to quickly (as little as 64 training steps) reach allocations of high social welfare (less than 5% loss) and fairness (up to almost 10% lower inequality compared to the best performing baseline).
Researcher Affiliation Academia Panayiotis Danassis , Florian Wiedemair and Boi Faltings Artificial Intelligence Laboratory, Ecole Polytechnique F ed erale de Lausanne (EPFL), Switzerland {panayiotis.danassis, florian.wiedemair, boi.faltings}@epfl.ch
Pseudocode Yes The pseudo-codes for ALMA and ALMA-Learning are presented in Algorithms 1 and 2, respectively.
Open Source Code No The paper refers to supplementary material in [Danassis et al., 2021] for implementation details, but does not provide a direct link to open-source code for the methodology within the paper itself or the supplementary material reference. The arXiv link is to the paper itself, not the code.
Open Datasets Yes We evaluate ALMA-Learning in a variety of synthetic benchmarks and a meeting scheduling problem based on real data from [Romano and Nunamaker, 2001].
Dataset Splits No The paper mentions 'training time-steps' and 'evaluation time-steps' related to the learning process, but does not provide specific train/validation/test dataset splits for reproducibility.
Hardware Specification No The paper mentions 'on-device deployment' for potential applications but does not specify the hardware (e.g., CPU, GPU models) used to conduct its experiments.
Software Dependencies No The paper mentions 'IBM ILOG CP optimizer' as a baseline solver but does not provide specific version numbers for it or any other software dependencies used in their own implementation.
Experiment Setup Yes ALMA-Learning was trained for 512 time-steps. The loss of rstart is then updated according to the following averaging process, where α is the learning rate: loss[rstart] (1 α)loss[rstart] + α (u(rstart) u(rwon))