Regularizing Neural Networks by Penalizing Confident Output Distributions | Read Paper on Bytez