### Abstract

We consider the optimization problem associated with training simple ReLUneural networks of the form $\mathbf{x}\mapsto\sum_{i=1}^{k}\max\{0,\mathbf{w}_i^\top \mathbf{x}\}$ with respect to thesquared loss. We provide a computer-assisted proof that even if the inputdistribution is standard Gaussian, even if the dimension is arbitrarily large,and even if the target values are generated by such a network, with orthonormalparameter vectors, the problem can still have spurious local minima once $6\lek\le 20$. By a concentration of measure argument, this implies that in highinput dimensions, \emph{nearly all} target networks of the relevant sizes leadto spurious local minima. Moreover, we conduct experiments which show that theprobability of hitting such local minima is quite high, and increasing with thenetwork size. On the positive side, mild over-parameterization appears todrastically reduce such local minima, indicating that an over-parameterizationassumption is necessary to get a positive result in this setting.

### Introduction (beta)

None

### Conclusion (beta)

None