[1912.02184] Towards Robust Image Classification Using Sequential Attention Modelscontact arXivarXiv Twitter

In this paper we propose to augment a modern neural-network architecture with an attention model inspired by human perception. Specifically, we adversarially train and analyze a neural model incorporating a human inspired, visual attention component that is guided by a recurrent top-down sequential process. Our experimental evaluation uncovers several notable findings about the robustness and behavior of this new model. First, introducing attention to the model significantly improves adversarial robustness resulting in state-of-the-art ImageNet accuracies under a wide range of random targeted attack strengths. Second, we show that by varying the number of attention steps (glances/fixations) for which the model is unrolled, we are able to make its defense capabilities stronger, even in light of stronger attacks --- resulting in a "computational race" between the attacker and the defender. Finally, we show that some of the adversarial examples generated by attacking our model are quite d

1 mentions: @mikechrzano
Keywords: attention
Date: 2020/02/26 00:51

Referring Tweets

@mikechrzano Our paper, Towards Robust Image Classification Using Sequential Attention Models (t.co/YfC0gFp1aZ), has been accepted to CVPR 2020. Yay!

Related Entries

Read more BPMF(Bayesian Probabilistic Matrix Factorization)によるレコメンド - LIVESENSE Data Analytics Blog
0 users, 0 mentions 2018/04/23 11:41
Read more 3D物体変形予測コンテスト | TC3株式会社
0 users, 1 mentions 2020/01/30 07:35
Read more [2001.11621] Constants of Motion of the Harmonic Oscillatorcontact arXivarXiv Twitter
0 users, 1 mentions 2020/02/06 12:54
Read more [2002.01725] Random Schrödinger Operators and Anderson localization in aperiodic mediacontact arXiva...
0 users, 1 mentions 2020/02/06 12:54
Read more Call for papers – The 10th International Conference on Probabilistic Graphical Models
0 users, 1 mentions 2020/03/25 15:51