Study Federated learning litterature and libraries

[RomainGoussault]

https://arxiv.org/pdf/1602.05629.pdf
https://arxiv.org/pdf/1805.09767.pdf

https://arxiv.org/pdf/1908.07873.pdf
https://arxiv.org/pdf/1912.04977.pdf

[jeromechambost]

Federated averaging is the state of the art method for fed learning. New framework FedProx theoretically provides more robust convergence and improves performance especially when there is systems heterogeneity
https://arxiv.org/pdf/1812.06127.pdf ==> could be tested with important number of partners

Increasing number of federated learning tools and frameworks:

• TensorFlow Federated : https://www.tensorflow.org/federated
• Federated AI Technology Enabler/ FATE : https://github.com/FederatedAI/FATE
• Leaf : https://leaf.cmu.edu (benchmark on impact of number of partners on federated avg performance on various datasets)
• PadleFL: https://github.com/PaddlePaddle/PaddleFL
• Clara Training Framework (focus on medical images): https://docs.nvidia.com/clara/tlt-mi/clara-train-sdk-v2.0/nvmidl/index.html

[bowni]

@jeromechambost this seems to be the perfect issue for you to update with your research 😄

[bowni]

@jeromechambost it would be interesting to complete the bibliography file @HeytemBou and @arthurPignet initiated in PR #348