ICML 2018oral327 citations

Spurious Local Minima are Common in Two-Layer ReLU Neural Networks

Itay Safran, Ohad Shamir

Abstract

We consider the optimization problem associated with training simple ReLU neural networks of the form $\mathbf{x}\mapsto \sum_{i=1}^{k}\max\{0,\mathbf{w}_i^\top \mathbf{x}\}$ with respect to the squared loss. We provide a computer-assisted proof that even if the input distribution is standard Gaussian, even if the dimension is arbitrarily large, and even if the target values are generated by such a network, with orthonormal parameter vectors, the problem can still have spurious local minima once $6\le k\le 20$. By a concentration of measure argument, this implies that in high input dimensions,

BibTeX
@InProceedings{pmlr-v80-safran18a,
  title = 	 {Spurious Local Minima are Common in Two-Layer {R}e{LU} Neural Networks},
  author =       {Safran, Itay and Shamir, Ohad},
  booktitle = 	 {Proceedings of the 35th International Conference on Machine Learning},
  pages = 	 {4433--4441},
  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/safran18a/safran18a.pdf},
  url = 	 {https://proceedings.mlr.press/v80/safran18a.html},
  abstract = 	 {We consider the optimization problem associated with training simple ReLU neural networks of the form $\mathbf{x}\mapsto \sum_{i=1}^{k}\max\{0,\mathbf{w}_i^\top \mathbf{x}\}$ with respect to the squared loss. We provide a computer-assisted proof that even if the input distribution is standard Gaussian, even if the dimension is arbitrarily large, and even if the target values are generated by such a network, with orthonormal parameter vectors, the problem can still have spurious local minima once $6\le k\le 20$. By a concentration of measure argument, this implies that in high input dimensions,