AI RESEARCH
On the Expressive Power of Permutation-Equivariant Weight-Space Networks
arXiv CS.LG
•
ArXi:2602.01083v2 Announce Type: replace Weight-space learning studies neural architectures that operate directly on the parameters of other neural networks. Motivated by the growing availability of pretrained models, recent work has nstrated the effectiveness of weight-space networks across a wide range of tasks. SOTA weight-space networks rely on permutation-equivariant designs to improve generalization. However, this may negatively affect expressive power, warranting theoretical investigation.