Federated Learning in Vision Tasks | Umberto Michieli, PhD@Uni of Padova, Intern@Samsung Research

Abstract :
1. Background on Federated Learning
2. How can we learn models in a distributed fashion if data is highly non- i.i.d.?
3. Can we leverage federated optimization on the basis of the learned representation in each client? We will do it from class prototype margins (FedProto).
4. Can we understand what a method has learned and why? Quantitative and qualitative analyses on both image classification and semantic segmentation.
Speaker :
- Ph.D. candidate on ‘Visual Understanding across Semantic Groups, Domains and Devices’ @UniPD
- M.Sc. Telecommunication Engineering, 2018 @UniPD
- Master thesis in Dresden (@TUD), 6 months
- Internship in London @Samsung Research UK, 8 months
Discussed Work :
[1] Michieli U. and Ozay M., ”Are all Users treated Fairly in Federated Learning Systms?." CVPR Workshop on Responsible Computer Vision 2021
[2] Michieli U. and Ozay M. . "Prototype Guided Federated Learning of Visual Feature Representations”, arXiv 2021.
Slides : drive.google.com/file/d/1AxrL...

[@akhildev312]

Can you please share the slides asap? The link is broken

[@user-pl6ur3mn7z]

please share the slides