Optimizing ML Training with Metagradient Descent

Logan Engstrom, Andrew Ilyas, Benjamin Chen +3

A major challenge in training large-scale machine learning models is configuring the training process to maximize model performance, i.e., finding the best training setup from a va…