Neural Collapse is Globally Optimal in Deep Regularized ResNets and Transformers

Devs

Neural Collapse is Globally Optimal in Deep Regularized ResNets and Transformers | Read Paper on Bytez