Global Minimizers of ℓp-Regularized Objectives Yield the Sparsest ReLU Neural Networks

Open in new window