Convex Relaxations of ReLU Neural Networks Approximate Global Optima in Polynomial Time.

Sungyoon Kim, Mert Pilanci

Browse the full ICML paper archive.