NSF CRII: RI: A Study of Rank-based Decomposable Losses for Machine Learning

Principal Investigator: Shu Hu

This project will be conducted in two interrelated thrusts. The first thrust explores a novel and general rank-based aggregate loss for supervised learning. The focus will encompass efficient algorithms that can optimize this loss with guaranteed convergence, along with streamlined techniques to determine relevant hyperparameters. The developed loss will be connected with distributionally robust optimization to gain insights into its sample-level robustness and the development of new types of rank-based aggregate losses. Additionally, theoretical guarantees will be established for rank-based aggregate losses, including classification calibration, classification consistency, and generalization properties. The second thrust aims to study a general formulation of rank-based individual loss with theoretical analysis, bolstering label-level robustness in multi-class and multi-label learning scenarios. Furthermore, the use of rank-based individual loss will be expanded to tackle fairness learning challenges and investigate the resilience of models trained with this loss against adversarial threats, including verification and defense mechanisms.

Personnel

Students: Li Lin, Santosh

Representative Publications

• @article{JMLR:v25:23-0888,
    author  = {Shu Hu and George H. Chen},
    title   = {Fairness in Survival Analysis with Distributionally Robust Optimization},
    journal = {Journal of Machine Learning Research},
    year    = {2024},
    volume  = {25},
    number  = {246},
    pages   = {1--85},
    url     = {http://jmlr.org/papers/v25/23-0888.html}
  }

• @INPROCEEDINGS{10672612,
    author={Santosh and Lin, Li and Amerini, Irene and Wang, Xin and Hu, Shu},
    booktitle={2024 IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)}, 
    title={Robust CLIP-Based Detector for Exposing Diffusion Model-Generated Images}, 
    year={2024},
    volume={},
    number={},
    pages={1-7},
    keywords={Training;Deepfakes;Codes;Image synthesis;Surveillance;Detectors;Multilayer perceptrons;Diffusion models;CLIP;Robust;AI images},
    doi={10.1109/AVSS61716.2024.10672612}}

• @inproceedings{lin2024preserving,
    title={Preserving fairness generalization in deepfake detection},
    author={Lin, Li and He, Xinan and Ju, Yan and Wang, Xin and Ding, Feng and Hu, Shu},
    booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
    pages={16815--16825},
    year={2024}
  }