2 ms·Gradient descent for wide two-layer neural networks – I: Global convergence1 points by matt_d 6y ago