A G-only collective kernel EFT for pre-activation ResNets yields accurate mean kernel predictions at all depths but accumulates O(1) errors in covariance and fails for the 1/n correction due to source closure breakdown.
Hanin,Random fully connected neural networks as perturbatively solvable hierarchies, Journal of Machine Learning Research25(2024) 1
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
-
Collective Kernel EFT for Pre-activation ResNets
A G-only collective kernel EFT for pre-activation ResNets yields accurate mean kernel predictions at all depths but accumulates O(1) errors in covariance and fails for the 1/n correction due to source closure breakdown.