The limiting eigenvalue distribution of the two-layer NTK in the quadratic scaling n/(dp) tends to a Marchenko-Pastur map applied to a deterministic measure depending on the activation and output weights.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.PR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Eigenvalue distribution of the Neural Tangent Kernel in the quadratic scaling
The limiting eigenvalue distribution of the two-layer NTK in the quadratic scaling n/(dp) tends to a Marchenko-Pastur map applied to a deterministic measure depending on the activation and output weights.