A Regularizer That Teaches Neural Networks to Be Compressible While They Train — For Less Than 1% Overhead

A Regularizer That Teaches Neural Networks to Be Compressible While They Train — For Less Than 1% Overhead | BedrockNews