Loading Now

Summary of Korea-sfl: Knowledge Replay-based Split Federated Learning Against Catastrophic Forgetting, by Zeke Xia and Ming Hu and Dengke Yan and Ruixuan Liu and Anran Li and Xiaofei Xie and Mingsong Chen


KoReA-SFL: Knowledge Replay-based Split Federated Learning Against Catastrophic Forgetting

by Zeke Xia, Ming Hu, Dengke Yan, Ruixuan Liu, Anran Li, Xiaofei Xie, Mingsong Chen

First submitted to arxiv on: 19 Apr 2024

Categories

  • Main: Machine Learning (cs.LG)
  • Secondary: None

     Abstract of paper      PDF of paper


GrooveSquid.com Paper Summaries

GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!

Summary difficulty Written by Summary
High Paper authors High Difficulty Summary
Read the original abstract here
Medium GrooveSquid.com (original content) Medium Difficulty Summary
The proposed novel Split Federated Learning approach, called KoReA-SFL, addresses the issues of low training accuracy and catastrophic forgetting in traditional SFL methods by adopting a multi-model aggregation mechanism and knowledge replay strategy. This approach maintains multiple branch model portions on cloud servers for local training and aggregated master-model portion for knowledge sharing among branch portions. The main server selects assistant devices for knowledge replay based on data distribution, allowing for improved test accuracy up to 23.25%. This method demonstrates significant performance improvement in both non-IID and IID scenarios.
Low GrooveSquid.com (original content) Low Difficulty Summary
KoReA-SFL is a new way of doing Split Federated Learning that helps with sharing knowledge among devices. Right now, this type of learning has a problem where the training accuracy is low because it doesn’t take into account differences in data between devices. To solve this issue, KoReA-SFL uses multiple models and replaying old knowledge to help the devices learn better. It even chooses which devices get to share their knowledge based on how much they have learned. This new approach works really well, increasing test accuracy by up to 23.25%.

Keywords

» Artificial intelligence  » Federated learning