Fetching the paper…

Gradient Descent can Learn Less Over-parameterized Two-layer Neural Networks on Classification Problems · Around