Learning Rates as a Function of Batch Size: A Random Matrix Theory Approach to Neural Network Training

Open in new window