Neural collapse is globally optimal in deep regularized ResNets and transformers, with the approximation improving as depth grows.
Intrinsic dimension of data representa- tions in deep neural networks
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Neural Collapse is Globally Optimal in Deep Regularized ResNets and Transformers
Neural collapse is globally optimal in deep regularized ResNets and transformers, with the approximation improving as depth grows.