We design the first game-theoretic efficient machine unlearning evaluation metric with provable properties.
Exploring a novel approach to exactly solve an NP-hard combinatorial optimization problem by using imitation learning.