Symmetry Breaking and Equivariant Neural Networks

arXiv:2312.09016v2 Announce Type: replace
Abstract: Using symmetry as an inductive bias in deep learning has been proven to be a principled approach for sample-efficient model design. However, the relationship between symmetry and the imperative for equivariance in neural networks is not always obvious. Here, we analyze a key limitation that arises in equivariant functions: their incapacity to break symmetry at the level of individual data samples. In response, we introduce a novel notion of 'relaxed equivariance' that circumvents this limitation. …

abstract analyze arxiv bias breaking cs.lg deep learning design functions however inductive key model design networks neural networks relationship sample stat.ml symmetry type