notesum.ai

Published at December 5

HyperMARL: Adaptive Hypernetworks for Multi-Agent RL

cs.LG

cs.AI

cs.MA

Released Date: December 5, 2024

Authors: Kale-ab Abebe Tessera¹, Arrasy Rahman², Stefano V. Albrecht¹

Aff.: ¹University of Edinburgh; ²University of Texas at Austin

Arxiv: http://arxiv.org/pdf/2412.04233v1

Refer to caption

	Specialisation Game
Method	2	4	8
NoPS	.9996 $\pm$ .0004	.9988 $\pm$ .0007	.9989 $\pm$ .0008
FuPS	.7493 $\pm$ .0062	.6826 $\pm$ .0039	.6545 $\pm$ .0025
FuPS+ID	.9999 $\pm$ .0002	.9995 $\pm$ .0003	.7275 $\pm$ .0293
	Synchronisation Game
Method	2	4	8
NoPS	.9993 $\pm$ .0004	.9976 $\pm$ .0022	.1250 $\pm$ 0
FuPS	.9999 $\pm$ .0002	1.0 $\pm$ 0	.3000 $\pm$ .3500
FuPS+ID	.9999 $\pm$ 0	.9999 $\pm$ .0003	.1250 $\pm$ 0