pith. sign in

arxiv: 1810.03974 · v1 · pith:BY2BWENDnew · submitted 2018-10-09 · 💻 cs.NE · cs.NA· math.NA

Collective evolution of weights in wide neural networks

classification 💻 cs.NE cs.NAmath.NA
keywords evolutioncollectiveequationpointsstationarytransportweightsagreement
0
0 comments X
read the original abstract

We derive a nonlinear integro-differential transport equation describing collective evolution of weights under gradient descent in large-width neural-network-like models. We characterize stationary points of the evolution and analyze several scenarios where the transport equation can be solved approximately. We test our general method in the special case of linear free-knot splines, and find good agreement between theory and experiment in observations of global optima, stability of stationary points, and convergence rates.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.