Wide neural networks with cross-entropy loss remain in the lazy training regime under parameter-space regularization or non-degenerate targets, allowing explicit NTK-based solution characterization and uncertainty analysis.
Diagonalizing the softmax: Hadamard initialization for tractable cross-entropy dynamics.arXiv preprint arXiv:2512.04006
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
The Neural Tangent Kernel for Classification
Wide neural networks with cross-entropy loss remain in the lazy training regime under parameter-space regularization or non-degenerate targets, allowing explicit NTK-based solution characterization and uncertainty analysis.