{ "id": "2008.06081", "version": "v1", "published": "2020-08-13T18:49:15.000Z", "updated": "2020-08-13T18:49:15.000Z", "title": "Adversarial Training and Provable Robustness: A Tale of Two Objectives", "authors": [ "Jiameng Fan", "Wenchao Li" ], "comment": "16 pages", "categories": [ "cs.LG", "stat.ML" ], "abstract": "We propose a principled framework that combines adversarial training and provable robustness verification for training certifiably robust neural networks. We formulate the training problem as a joint optimization problem with both empirical and provable robustness objectives and develop a novel gradient-descent technique that can eliminate bias in stochastic multi-gradients. We perform both theoretical analysis on the convergence of the proposed technique and experimental comparison with state-of-the-arts. Results on MNIST and CIFAR-10 show that our method can match or outperform prior approaches for provable l infinity robustness.", "revisions": [ { "version": "v1", "updated": "2020-08-13T18:49:15.000Z" } ], "analyses": { "keywords": [ "provable robustness", "adversarial training", "objectives", "outperform prior approaches", "training certifiably robust neural networks" ], "note": { "typesetting": "TeX", "pages": 16, "language": "en", "license": "arXiv", "status": "editable" } } }