Causes of Stability in Dynamic Coalition Formation
Authors: Niclas Boehmer, Martin Bullinger, Anna Maria Kerkmann
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In our simulations, we observe that our model of dynamic utilities leads to the (quick) convergence of Nash dynamics. Moreover, we analyze the structure and expressiveness of the produced outcomes. Finally, we outline results for other perception models and for computational questions concerned with finding shortest converging sequences. We analyze by means of simulations how resent and appreciation influence NS dynamics in ASHGs by examining the speed of convergence and the composition of the reached stable states. This gives insights in the actual process that leads to convergence beyond the convergence guarantees and counterexamples presented before. We only provide a brief overview of some of our results; see our full version for details (Boehmer, Bullinger, and Kerkmann 2022). We focus on ASHGs with n = 50 agents and sample 100 games for each of the following utility models: Uniform For two agents a, b N with a = b, we sample ua(b) by sampling an integer from [ 100, 100]. Gaussian For each agent a N, we sample her base qualification µa by sampling an integer from [ 100, 100]. For two agents a, b N with a = b, we sample ua(b) by drawing an integer from the Gaussian distribution with mean µb and standard deviation 10. Our dynamics start with the singleton partition. Subsequently, we perform an NS deviation selected uniformly at random until the dynamics converges. In addition to the concepts considered in our theoretical analysis, we also consider resentful-appreciative agents, i.e., agents that are both resentful and appreciative. Table 2 shows parts of our results. |
| Researcher Affiliation | Academia | Niclas Boehmer1, Martin Bullinger2, Anna Maria Kerkmann3 1 Algorithmics and Computational Complexity, Technische Universit at Berlin 2 School of Computation, Information and Technology, Technische Universit at M unchen 3 Institut f ur Informatik, Heinrich-Heine-Universit at D usseldorf |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described. |
| Open Datasets | No | The paper describes generating synthetic data ("sample 100 games") based on uniform and Gaussian utility models, but does not provide concrete access information (link, DOI, repository, or formal citation to an established public dataset) for a publicly available or open dataset. |
| Dataset Splits | No | The paper describes running simulations on sampled games but does not specify dataset splits (training, validation, test) typically used for model evaluation. |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types, or memory amounts) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details (library or solver names with version numbers) needed to replicate the experiment. |
| Experiment Setup | Yes | Our dynamics start with the singleton partition. Subsequently, we perform an NS deviation selected uniformly at random until the dynamics converges. We focus on ASHGs with n = 50 agents and sample 100 games for each of the following utility models: Uniform For two agents a, b N with a = b, we sample ua(b) by sampling an integer from [ 100, 100]. Gaussian For each agent a N, we sample her base qualification µa by sampling an integer from [ 100, 100]. |