Global Minimizers of \ell p -Regularized Objectives Yield the Sparsest ReLU Neural Networks

Open in new window