Machine Learning Study
미시간 대학의 EECS 498 강의를 기반으로 기계 학습의 기초를 학습하는 스터디
참여자:
김도형: Lec 01 (Introduction to Deep Learning), Lec 09 (Hardwares and Softwares), Lec 14 (Visualizing and Understanding)
김동현: Lec 04 (Optimization), Lec 11 (Training Neural Networks II), Lec 16 (Detection and Segmentation)
김진영: Lec 03 (Linear Classifiers), Lec 15 (Object Detection), Lec 20 (Generative Models II)
남유찬: Lec 06 (Backpropagation), Lec 13 (Attention), Lec 18 (Videos)
오영민: Lec 02 (Image Classification), Lec 10 (Training Neural Networks I), Lec 19 (Generative Models I)
이건희: Lec 05 (Neural Networks), Lec 12 (Recurrent Networks), Lec 21 (Reinforcement Learning)
이승헌: Lec 08 (CNN Architectures)
유재혁: Lec 07 (Convolutional Networks), Lec 17 (3D Vision), Lec 22 (Conclusion)
Trustworthy AI Study
인공지능과 신뢰성 인공지능의 대표적인 논문들을 읽고 설명 가능성, 공정성, 프라이버시 등의 개념을 공부하고 최신 연구 결과들을 공유하는 스터디
발표 논문:
AI 기초
1. Supervised Contrastive Learning — 발표자: 김동현
2. Barlow Twins: Self-Supervised Learning via Redundancy Reduction — 발표자: 남유찬
3. Attention is All You Need — 발표자: 이건희
4. Learning Transferable Visual Models From Natural Language Supervision — 발표자: 김도형
5. Pre-training of Deep Bidirectional Transformers for Language Understanding — 발표자: 김도형
6. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale — 발표자: 유재혁
7. mixup: Beyond Empirical Risk Minimization — 발표자: 유재혁
8. Sharpness-Aware Minimization for Efficiently Improving Generalization — 발표자: 오영민
9. Generative Adversarial Networks — 발표자: 오영민
10. Diffusion models beat gans on image synthesis — 발표자: 김진영
설명가능성
11. Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization — 발표자: 송은서
12. Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks — 발표자: 송은서
13. Concept Bottleneck Models — 발표자: 송은서
프라이버시
14. Training-Free Safe Denoisers for Safe Use of Diffusion Models — 발표자: 김도형
15. Membership Inference Attacks Against Machine Learning Models — 발표자: 김진영
16. Extracting Training Data from Diffusion Models — 발표자: 김동현
17. Machine Unlearning — 발표자: 이건희
보안
18. Stealing Machine Learning Models via Prediction APIs — 발표자: 남유찬
19. How to Backdoor Federated Learning — 발표자: 유재혁
20. Towards Deep Learning Models Resistant to Adversarial Attacks — 발표자: 오영민
21. A Watermark for Large Language Models — 발표자: 송은서
22. A Recipe for Watermarking Diffusion Models — 발표자: 남유찬
공정성
23. Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization — 발표자: 김동현
24. Learning Adversarially Fair and Transferable Representations — 발표자: 송은서
25. Learning from Failure: De-biasing Classifier from Biased Classifier — 발표자: 이건희