Fetching the paper…

SGD and Weight Decay Secretly Minimize the Rank of Your Neural Network · Around