Proxy Convexity: A Unified Framework for the Analysis of Neural Networks Trained by Gradient Descent

Open in new window