DBIS

With the introduction of data privacy protection laws, such as General Data Protection Regulation (GDPR) in the EU, Health Insurance Portability and Accountability Act (HIPAA) in the USA, and Data Protection Act in the UK, data privacy is becoming an integral part of every organization and traditional approaches for training machine learning models in a centralized fashion face significant challenges due to regulatory and privacy concerns. In recent years, federated learning, as a promising privacy-preserving machine learning paradigm to collaboratively learn a model from dispersed data without directly sharing the data, has gained enormous attention from both academia and industry. 

However, federated learning is often confronted with the statistical diversity of the data among clients, so-called non-iid problem. In practice, there is also another form of heterogeneity among clients – clients under the same federation may have related but different tasks. For example, some hospitals intend to classify prostate tissue biopsies into Gleason patterns, while some consider ISUP grade. In such a scenario, the clients’ ultimate goals are different, but all demand learning compact and generalizable representations for tissue images.

The objective of this thesis is to adapt Personalized Federated Learning to tackle the challenges of statistical diversity of distributed data and hetero-task learning.

If you are interested in this thesis, do not hesitate to contact us via mou@dbis.rwth-aachen.de.

Please find the seed literature as followings:

1. Goan E, Fookes C. Bayesian neural networks: An introduction and survey. InCase Studies in Applied Bayesian Data Science 2020 (pp. 45-87). Springer, Cham.
2. Zhang X, Li Y, Li W, Guo K, Shao Y. Personalized Federated Learning via Variational Bayesian Inference. In International Conference on Machine Learning 2022 Jun 28 (pp. 26293-26310). PMLR.
3. Achituve I, Shamsian A, Navon A, Chechik G, Fetaya E. Personalized Federated Learning with Gaussian Processes. Advances in Neural Information Processing Systems. 2021 Dec 6;34:8392-406.
4. Chen HY, Chao WL. Fedbe: Making bayesian model ensemble applicable to federated learning. arXiv preprint arXiv:2009.01974. 2020 Sep 4.
5. Liu L, Zheng F, Chen H, Qi GJ, Huang H, Shao L. A Bayesian Federated Learning Framework with Online Laplace Approximation. arXiv preprint arXiv:2102.01936. 2021 Feb 3.