Path Regularization: A Convexity and Sparsity Inducing Regularization for Parallel ReLU Networks

Understanding the fundamental principles behind the success of deep neural networks is one of the most important open questions in the current literature.