Learning Coalition Structures with Games

Authors: Yixuan Even Xu, Chun Kai Ling, Fei Fang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we conduct experiments to evaluate IG in the auction setting and the results align with our theoretical analysis. We conduct experiments to evaluate the performance of our algorithms in practice.
Researcher Affiliation Academia 1Tsinghua University 2Columbia University 3Carnegie Mellon University
Pseudocode Yes Algorithm 1: Iterative Grouping (IG) and Algorithm 2: IG with Auctions (Auction IG)
Open Source Code Yes The source codes can be found at https://github.com/Yixuan Even Xu/coalition-learning.
Open Datasets No We model this more realistic setting by assuming that the values are drawn from an item pool V, which is a distribution U[0, 1]n over Rn. For each setting, we fix n and either fix m or sample m from U[n]. Then, we synthesize a coalition structure S with exactly n agents and m coalitions at random. The paper describes a synthetic data generation process rather than using an existing publicly available dataset with concrete access information.
Dataset Splits No For each setting, we fix n and either fix m or sample m from U[n]. Then, we synthesize a coalition structure S with exactly n agents and m coalitions at random. We then run Auction IG, check the correctness of its output, and record the sample complexity (the total number of samples used). The paper describes a simulation-based evaluation without explicit train/validation/test splits commonly found in supervised learning.
Hardware Specification Yes We implement it in Python and evaluate it on a server with 56 cores and 504G RAM, running Ubuntu 20.04.6.
Software Dependencies No We implement it in Python and evaluate it on a server with 56 cores and 504G RAM, running Ubuntu 20.04.6. No specific library versions are provided.
Experiment Setup Yes Experiment setup. We evaluate Auction IG under different settings of n and m, where n is the number of agents and m is the number of coalitions. For each setting, we fix n and either fix m or sample m from U[n]. Then, we synthesize a coalition structure S with exactly n agents and m coalitions at random. We then run Auction IG, check the correctness of its output, and record the sample complexity (the total number of samples used).