상세정보
학술대회
Adversarial Training With Stochastic Weight Average
- 발행일
- 202109
- 출처
- International Conference on Image Processing (ICIP) 2021, pp.814-818
- 협약과제
- 20HS5100, (딥뷰-1세부) 실시간 대규모 영상 데이터 이해·예측을 위한 고성능 비주얼 디스커버리 플랫폼 개발, 배유석
- 초록
- Although adversarial training is the most reliable method to train robust deep neural networks so far, adversarially trained networks still show large gap between their accuracies on clean images and those on adversarial images. In conventional classification problem, one can gain higher accuracy by ensembling multiple networks. However, in adversarial training, there are obstacles to adopt such ensemble method. First, as inner maximization is expensive, training multiple networks adversarially becomes overburden. Moreover, the naive ensemble faces dilemma on choosing target model to generate adversarial examples with. Training adversarial examples of the members causes covariate shift, while training those of ensemble diminishes the benefit of ensembling. With these insights, we adopt stochastic weight average methods and improve it by considering overfitting nature of adversarial training. Our method take the benefit of ensemble while avoiding the described problems. Experiments on CIFAR10 and CIFAR100 shows our method improves the robustness effectively.
- KSP 제안 키워드
- Adversarial Training, Classification problems, Covariate shift, Deep neural network(DNN), Ensemble method, Large gap, Multiple network, target model