ICML 2018oral58 citations

The Multilinear Structure of ReLU Networks

Thomas Laurent, James Brecht

Abstract

We study the loss surface of neural networks equipped with a hinge loss criterion and ReLU or leaky ReLU nonlinearities. Any such network defines a piecewise multilinear form in parameter space. By appealing to harmonic analysis we show that all local minima of such network are non-differentiable, except for those minima that occur in a region of parameter space where the loss surface is perfectly flat. Non-differentiable minima are therefore not technicalities or pathologies; they are heart of the problem when investigating the loss of ReLU networks. As a consequence, we must employ techniques from nonsmooth analysis to study these loss surfaces. We show how to apply these techniques in some illustrative cases.

BibTeX
@InProceedings{pmlr-v80-laurent18b,
  title = 	 {The Multilinear Structure of {R}e{LU} Networks},
  author =       {Laurent, Thomas and von Brecht, James},
  booktitle = 	 {Proceedings of the 35th International Conference on Machine Learning},
  pages = 	 {2908--2916},
  year = 	 {2018},
  editor = 	 {Dy, Jennifer and Krause, Andreas},
  volume = 	 {80},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {10--15 Jul},
  publisher =    {PMLR},
  pdf = 	 {http://proceedings.mlr.press/v80/laurent18b/laurent18b.pdf},
  url = 	 {https://proceedings.mlr.press/v80/laurent18b.html},
  abstract = 	 {We study the loss surface of neural networks equipped with a hinge loss criterion and ReLU or leaky ReLU nonlinearities. Any such network defines a piecewise multilinear form in parameter space. By appealing to harmonic analysis we show that all local minima of such network are non-differentiable, except for those minima that occur in a region of parameter space where the loss surface is perfectly flat. Non-differentiable minima are therefore not technicalities or pathologies; they are heart of the problem when investigating the loss of ReLU networks. As a consequence, we must employ techniques from nonsmooth analysis to study these loss surfaces. We show how to apply these techniques in some illustrative cases.}
}
The Multilinear Structure of ReLU Networks · ICML 2018