FedNL. Making Newton-Type Methods Applicable to FL.

FedNL. Practical Implementation (Ongoing project) at VCC OPEN HOUSE 2023 event.

Federated Learning (FL) has emerged as a promising technique for edge devices to collaboratively learn a shared machine learning model while keeping training data locally on the device. A Federated Learning setting has several attributes which make it different from distributed optimization including communication complexity, data heterogeneity, device heterogeneity, and data privacy. The proposed algorithms from the work of M.Safaryan, R.Islamov, X.Qian, P.Richtárik FedNL: Making Newton-Type Methods Applicable to Federated Learning (arXiv:2106.02969, ICML 2021) bring a family of Federated Newton Learn methods and is a significant step in making second-order methods applicable to Federated Learning (FL) and Distributed Optimization.

We believe the settings in which FedNL is targeted are general enough to be applicable in various areas of science and engineering in which there are problems with using the classical Newton method. However, to make this happens we need to bring the practical implementation of these algorithms to the whole world. Our work is targeted to close the gap between theoretical algorithms and their practical implementation in modern computing and communication devices.

The current stage we achieved will be presented at VCC OPEN HOUSE 2023 event.

  • Registration: Registration for “VCC OPEN HOUSE 2023” is available from the website of the Visual Computing Center (VCC). Link for registration.
  • Location: Building 1, The Spine, 4700 King Abdullah University of Science and Technology, Thuwal 23955-6900, Saudi Arabia. Link in Google Maps..
  • Date and Time: Thursday, March 2, 12:00 - 14:00, UTC+3.

Written on February 19, 2023