Certifiable Robustness to Discrete Adversarial Perturbations for Factorization Machines
Yang Liu, Xianzhuo Xia, Liang Chen, Xiangnan He, Carl Yang, Zibin Zheng
Abstract
Factorization machines (FMs) have been widely adopted to model the discrete feature interactions in recommender systems. Despite their great success, currently there is no study of their robustness to discrete adversarial perturbations. Whether modifying a certain number of the discrete input features has a dramatic effect on the FM's prediction? Although there exist robust training methods for FMs, they neglect the discrete property of input features and lack of an effective mechanism to verify the model robustness.
Topics & Concepts
Robustness (evolution)Computer scienceAdversarial systemFactorizationArtificial intelligenceProperty (philosophy)Control theory (sociology)AlgorithmControl (management)ChemistryPhilosophyGeneEpistemologyBiochemistryRecommender Systems and TechniquesAdvanced Graph Neural NetworksTopic Modeling