Improving Local Training in Federated Learning via Temperature Scaling

Kichang Lee, Songkuk Kim, JeongGil Ko
1/18/2024
cs.LGcs.AI

No official code found. Best implementation: temperaturescaling/temperature-scaling (0 stars, last commit 2024-10-22). 1 more below.

Abstract

Federated learning is inherently hampered by data heterogeneity: non-i.i.d. training data over local clients. We propose a novel model training approach for federated learning, FLex&Chill, which exploits the Logit Chilling method. Through extensive evaluations, we demonstrate that, in the presence of non-i.i.d. data characteristics inherent in federated learning systems, this approach can expedite model convergence and improve inference accuracy. Quantitatively, from our experiments, we observe up to 6X improvement in the global federated learning model convergence time, and up to 3.37% improvement in inference accuracy.

View on arXivView PDF

Code Implementations(2)

temperaturescaling/temperature-scalingCommunity65%

Repository for "Improving Local Training in Federated Learning via Temperature Scaling"

00Python, Jupyter NotebookOct 8, 20242 years ago
leekichang/temperature-scalingCommunity56%

Repository for "Improving Local Training in Federated Learning via Temperature Scaling"

00Python, Jupyter NotebookNov 25, 202511 months ago

Similar Papers

FedBug: A Bottom-Up Gradual Unfreezing Framework for Federated Learning

86% match2023

The Best of Both Worlds: Accurate Global and Personalized Models through Federated Learning with Data-Free Hyper-Knowledge Distillation

86% match1 repo242023

FedImpro: Measuring and Improving Client Update in Federated Learning

86% match2024

An Aggregation-Free Federated Learning for Tackling Data Heterogeneity

86% match2024

Unlocking the Potential of Model Calibration in Federated Learning

86% match2024

FedDPC : Handling Data Heterogeneity and Partial Client Participation in Federated Learning

85% match2025

Cite this paper

@article{lee2024improving,
  title  = {Improving Local Training in Federated Learning via Temperature Scaling},
  author = {Kichang Lee and Songkuk Kim and JeongGil Ko},
  year   = {2024},
  eprint = {2401.09986},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2401.09986v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2401.09986.svg)](https://codeofpaper.com/paper/2401.09986)

Discussion