Learning Strategy-Aware Linear Classifiers

The question of repeatedly learning linear classifiers against agents who are trying to game the deployed classifiers, and the use of the Stackelberg regret to measure the performance of the algorithms, is addressed.