The 37th Annual Conference of the Japanese Society for Artificial Intelligence, p. 3U1IS301, June 8, 2023
The 37th Annual Conference of the Japanese Society for Artificial Intelligence (JSAI2023)
Multi-agent simulations are useful in social sciences but they encounter an evaluation difficulty in that many social phenomena are qualitative, and it is difficult to evaluate quantitatively the realness of simulations. Therefore, we propose a new quantitative evaluation method for multi-agent simulation in social sciences using a generative adversarial network (GAN). In our proposed method, GAN's critic was used as a simulation evaluator. We implemented a GAN and a multi-agent simulation for financial markets in experiments to test the proposed method. Results showed that our proposed method achieved promising results as an alternative to the traditional qualitative evaluation; it enabled successful quantitative evaluation with good correspondence with the traditional qualitative evaluation. The realization of quantitative evaluation using GAN as an alternative to the traditional qualitative evaluation may expand the usage of multi-agent simulation.
Multi-agent simulation; Generative adversarial network; Evaluation; Financial Markets;
10.11517/pjsai.JSAI2023.0_3U1IS301
@inproceedings{Hirano2023-jsai37,
title={{Quantitative Evaluation of Multi-agent Simulation using Generative Adversarial Network -- An Alternative of Qualitative Evaluation for Artificial Market Simulation}},
author={Masanori HIRANO and Kiyoshi IZUMI},
booktitle={The 37th Annual Conference of the Japanese Society for Artificial Intelligence},
pages={3U1IS301},
doi={10.11517/pjsai.JSAI2023.0_3U1IS301},
year={2023}
}