Generalization Guarantees of Gradient Descent for Shallow Neural Networks | MIT Press Journals & Magazine | IEEE Xplore