Apple Researchers Propose Skipping Low-Influence Data Points to Make Machine Unlearning Cheaper

Apple researchers propose leaving out data points with negligible influence on a model, to cut the cost of machine 'unlearning' without hurting accuracy.