ELF: Federated Langevin Algorithms with Primal, Dual and Bidirectional Compression
Abstract
Federated sampling algorithms have recently gained great popularity in the community of machine learning and statistics. This paper studies variants of such algorithms called Error Feedback Langevin algorithms (ELF). In particular, we analyze the combinations of EF21 and EF21-P with the federated Langevin Monte-Carlo. We propose three algorithms: P-ELF, D-ELF, and B-ELF that use, respectively, primal, dual, and bidirectional compressors. We analyze the proposed methods under Log-Sobolev inequality and provide non-asymptotic convergence guarantees.
Cite
Text
Karagulyan and Richtárik. "ELF: Federated Langevin Algorithms with Primal, Dual and Bidirectional Compression." ICML 2023 Workshops: FL, 2023.Markdown
[Karagulyan and Richtárik. "ELF: Federated Langevin Algorithms with Primal, Dual and Bidirectional Compression." ICML 2023 Workshops: FL, 2023.](https://mlanthology.org/icmlw/2023/karagulyan2023icmlw-elf/)BibTeX
@inproceedings{karagulyan2023icmlw-elf,
title = {{ELF: Federated Langevin Algorithms with Primal, Dual and Bidirectional Compression}},
author = {Karagulyan, Avetik and Richtárik, Peter},
booktitle = {ICML 2023 Workshops: FL},
year = {2023},
url = {https://mlanthology.org/icmlw/2023/karagulyan2023icmlw-elf/}
}