Publications
-
Node Classification with Integrated Reject Option [arXiv] [Code]
Uday Bhaskar, Jayadratha Gayen, Charu Sharma, and Naresh Manwani.
Accepted for Oral presentation at Deployable AI Workshop, AAAI 2025.
arXiv preprint. Under Review.
-
Efficient Training Strategies for Local Differentially Private Federated Networks [arXiv]
Uday Bhaskar, Varul Srivastava, Avyukta Manjunatha Vummintala, Naresh Manwani, Sujit Gujar, et al.
arXiv preprint. Under Review.
Projects
-
Physics of Language Models 3.1 —
[Code]
- Generated synthetic datasets bioR and bioS as mentioned in the papers.
- Forked NanoGPT by Karpathy Sensei and extended it for instruction tuning.
- Compared performance of these models for various training strategies like Mixed Training, Augmentation, etc.
-
DP-SGD —
[Code]
- Implement Vanilla DP-SGD on standard image classification datasets using PyTorch and Opacus.
- Added extentions like Tempered Sigmoid Activation, Extracting ScatterNet features of images and training linear classifier and CNNs.