REVIEW 1 cited by
Asymptotics of Wide Networks from Feynman Diagrams
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Asymptotics of Wide Networks from Feynman Diagrams
read the original abstract
Understanding the asymptotic behavior of wide networks is of considerable interest. In this work, we present a general method for analyzing this large width behavior. The method is an adaptation of Feynman diagrams, a standard tool for computing multivariate Gaussian integrals. We apply our method to study training dynamics, improving existing bounds and deriving new results on wide network evolution during stochastic gradient descent. Going beyond the strict large width limit, we present closed-form expressions for higher-order terms governing wide network training, and test these predictions empirically.
Forward citations
Cited by 1 Pith paper
-
Bulk-boundary decomposition of neural networks
The paper reframes SGD training of deep networks as a local Lagrangian with data confined to the boundaries, but the advertised energy continuity equation is absent from the body.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.