Mean-field analysis of piecewise linear solutions for wide ReLU networks
Shevchenko, Aleksandr
Kungurtsev, Vyacheslav
Mondelli, Marco
ddc:000
Understanding the properties of neural networks trained via stochastic gradient descent (SGD) is at the heart of the theory of deep learning. In this work, we take a mean-field view, and consider a two-layer ReLU network trained via noisy-SGD for a univariate regularized regression problem. Our main result is that SGD with vanishingly small noise injected in the gradients is biased towards a simple solution: at convergence, the ReLU network implements a piecewise linear map of the inputs, and the number of “knot” points -- i.e., points where the tangent of the ReLU network estimator changes -- between two consecutive training inputs is at most three. In particular, as the number of neurons of the network grows, the SGD dynamics is captured by the solution of a gradient flow and, at convergence, the distribution of the weights approaches the unique minimizer of a related free energy, which has a Gibbs form. Our key technical contribution consists in the analysis of the estimator resulting from this minimizer: we show that its second derivative vanishes everywhere, except at some specific locations which represent the “knot” points. We also provide empirical evidence that knots at locations distinct from the data points might occur, as predicted by our theory.
Journal of Machine Learning Research
2022
info:eu-repo/semantics/article
doc-type:article
text
http://purl.org/coar/resource_type/c_6501
https://research-explorer.ista.ac.at/record/11420
https://research-explorer.ista.ac.at/download/11420/11422
Shevchenko A, Kungurtsev V, Mondelli M. Mean-field analysis of piecewise linear solutions for wide ReLU networks. <i>Journal of Machine Learning Research</i>. 2022;23(130):1-55.
eng
info:eu-repo/semantics/altIdentifier/issn/1532-4435
info:eu-repo/semantics/altIdentifier/issn/1533-7928
info:eu-repo/semantics/altIdentifier/arxiv/2111.02278
info:eu-repo/semantics/openAccess