Learning Distributions Generated by One-Layer ReLU Networks

Neural Information Processing Systems 

We consider the problem of estimating the parameters of a $d$-dimensional rectified Gaussian distribution from i.i.d.