Timezone: »
We consider the problem of OOD generalization,where the goal is to train a model that performs well on test distributions that are different from the training distribution. Deep learning models are known to be fragile to such shifts and can suffer large accuracy drops even for slightly different test distributions (Hendrycks & Dietterich, 2019).We propose a new method –DAFT– based on the intuition that adversarially robust combination of a large number of rich features should provide OOD robustness. Our method carefully distills the model from a powerful teacher that learns several discriminative features using standard training while combining them using adversarial training. The standard adversarial training procedure is modified to produce teachers which can guide the student better. We evaluate DAFT on standard benchmarks in the DomainBed framework, and find that DAFT consistently out-performs well-tuned ERM and distillation baselines by up to 6%, with more pronounced gains for smaller networks
Author Information
Anshul Nasery (Google)
Sravanti Addepalli (Indian Institute of Science)
Praneeth Netrapalli (Microsoft Research)
Prateek Jain (Google Research)
More from the Same Authors
-
2021 : Towards Achieving Adversarial Robustness Beyond Perceptual Limits »
Sravanti Addepalli · Samyak Jain · Gaurang Sriramanan · Shivangi Khare · Venkatesh Babu Radhakrishnan -
2021 : Differentially Private Model Personalization »
Prateek Jain · J K Rush · Adam Smith · Shuang Song · Abhradeep Guha Thakurta -
2021 : Private Alternating Least Squares: Practical Private Matrix Completion with Tighter Rates »
Steve Chien · Prateek Jain · Walid Krichene · Steffen Rendle · Shuang Song · Abhradeep Guha Thakurta · Li Zhang -
2022 : Efficient and Effective Augmentation Strategy for Adversarial Training »
Sravanti Addepalli · Samyak Jain · Venkatesh Babu Radhakrishnan -
2022 : DAFT: Distilling Adversarially Fine-tuned teachers for OOD Robustness »
Anshul Nasery · Sravanti Addepalli · Praneeth Netrapalli · Prateek Jain -
2023 Poster: Multi-Task Differential Privacy Under Distribution Skew »
Walid Krichene · Prateek Jain · Shuang Song · Mukund Sundararajan · Abhradeep Guha Thakurta · Li Zhang -
2023 Poster: Multi-User Reinforcement Learning with Low Rank Rewards »
Dheeraj Nagaraj · Suhas Kowshik · Naman Agarwal · Praneeth Netrapalli · Prateek Jain -
2021 Poster: Private Alternating Least Squares: Practical Private Matrix Completion with Tighter Rates »
Steve Chien · Prateek Jain · Walid Krichene · Steffen Rendle · Shuang Song · Abhradeep Guha Thakurta · Li Zhang -
2021 Oral: Private Alternating Least Squares: Practical Private Matrix Completion with Tighter Rates »
Steve Chien · Prateek Jain · Walid Krichene · Steffen Rendle · Shuang Song · Abhradeep Guha Thakurta · Li Zhang -
2021 Poster: Optimal regret algorithm for Pseudo-1d Bandit Convex Optimization »
Aadirupa Saha · Nagarajan Natarajan · Praneeth Netrapalli · Prateek Jain -
2021 Spotlight: Optimal regret algorithm for Pseudo-1d Bandit Convex Optimization »
Aadirupa Saha · Nagarajan Natarajan · Praneeth Netrapalli · Prateek Jain