On power-law covariance least squares problems, SignSVD (Muon) and SignSGD (Adam proxy) show three phases of relative performance depending on data exponent α and target exponent β.
Universality laws for high-dimensional learning with random features.IEEE Transactions on Information Theory, 69(3):1932–1964
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 2roles
background 1polarities
background 1representative citing papers
The covariance matrix of nonlinear recurrent neural networks equals that of a linear network with the same couplings, where DMFT order parameters determine the effective transfer function and noise.
citing papers explorer
-
Phases of Muon: When Muon Eclipses SignSGD
On power-law covariance least squares problems, SignSVD (Muon) and SignSGD (Adam proxy) show three phases of relative performance depending on data exponent α and target exponent β.
-
Linear equivalence of nonlinear recurrent neural networks
The covariance matrix of nonlinear recurrent neural networks equals that of a linear network with the same couplings, where DMFT order parameters determine the effective transfer function and noise.