Fishgrs Work Instant

"Fishr" is an ICML 2022 machine learning framework designed for out-of-distribution (OOD) generalization, focusing on regularizing the variance of gradients across different training domains. This approach improves model robustness by enforcing consistent sensitivity across domains to learn invariant features rather than domain-specific shortcuts. Read the full paper at proceedings.mlr.press .

(Placeholders for funding agencies, collaborators, and permitting authorities.) fishgrs work