• 鲁棒性：对抗性攻击、定向攻击和非定向攻击、最小距离攻击、最大允许攻击、基于规则的攻击。通过纳微扰。支持向量机的鲁棒性。

   

• 学习理论：偏差和方差，训练和测试，泛化，PAC框架，Hoeffding不等式，VC维。

 

 

Part 3: Learning Theory
Objective: Understand the theoretical limits of machine learning algorithms
Week 8: Lecture 22 (PDF) 2020-03-06 Is Learning Feasible? (Video)
Week 9: Lecture 23 (PDF) 2020-03-09 Probability Inequality (Video)
Week 9: Special Announcement about COVID-19. (PDF) 2020-03-11 (Video)
Week 9: Midterm (take home) See Instructions
Week 10: Lecture 24 (PDF) 2020-03-23 Probably Approximately Correct (Video 1) (Video 2)
Week 10: Lecture 25 (PDF) 2020-03-25 Generalization (Video 1) (Video 2)
Week 10: Lecture 26 (PDF) 2020-03-27 Growth Function (Video 1) (Video 2)
Week 11: Lecture 27 (PDF) 2020-03-30 VC Dimension (Video 1) (Video 2)
Week 11: Lecture 28 (PDF) 2020-04-01 Sample and Model Complexity (Video 1) (Video 2)
Week 11: Lecture 29 (PDF) 2020-04-03 Bias and Variance (Video 1) (Video 2)
Week 12: Lecture 30 (PDF) 2020-04-06 Overfit (Video 1)(Video 2)
Week 12: Lecture 31 (PDF) 2020-04-08 Regularization (Video 1) (Video 2)(Video 3)
Week 12: Lecture 32 (PDF) 2020-04-10 Validation (Video 1) (Video 2)


Part 4: Robust Machine Learning
Objective: Understand the robustness of machine learning algorithms
Week 13: Lecture 33 (PDF) 2020-04-13 Overview of Adversarial Attacks (Video 1) (Video 2)
Week 13: Lecture 34 (PDF) 2020-04-15 Minimum Distance Attack (Video 1) (Video 2)
Week 13: Lecture 35 (PDF) 2020-04-17 Maximum Loss Attack and Regularized Attack (Video 1) (Video 2)
Week 14: Lecture 36 (PDF) 2020-04-20 Defending Adversarial Attacks (Video 1) (Video 2)
Week 14: Lecture 37 (PDF) 2020-04-22 Robustness and Accuracy Trade-Off (Video 1) (Video 2)


Week 14: Lecture 38 (PDF) 2020-04-24 Conclusion: Practical Advices (Video 1) (Video 2)