Towards Trustworthy AI | ICML 24 25 | Neurips 24 Spotlight | Purdue
This is the recording of the PhD final exam of Deepak Ravikumar at Purdue ECE. The research discussed in this work can be found below:
1. Unveiling Privacy, Memorization, and Input Curvature Links. ICML 2024. https://openreview.net/forum?id=4dxR7...
2. Curvature Clues: Decoding Deep Learning Privacy with Input Loss Curvature. NeurIPS 2024. https://openreview.net/forum?id=ZEVDM... (Spotlight 2% Acceptance).
3. Towards Memorization Estimation: Fast, Formal and Free. ICML 2025. https://openreview.net/forum?id=KZlQE...
4. TREND: Transferability based Robust ENsemble Design. IEEE TAI 2022. https://ieeexplore.ieee.org/abstract/...
5. Intra-class mixup for out-of-distribution detection. IEEE Access 2023. https://ieeexplore.ieee.org/iel7/6287...
6. Homogenizing Non-IID Datasets via In-Distribution Knowledge Distillation for Decentralized Learning. TMLR 2024. https://openreview.net/forum?id=CuyJk...
7. Norm-Scaling for Out-of-Distribution Detection. arXiv 2022. https://arxiv.org/pdf/2205.03493