Summary of Fl-guard: a Holistic Framework For Run-time Detection and Recovery Of Negative Federated Learning, by Hong Lin et al.
FL-GUARD: A Holistic Framework for Run-Time Detection and Recovery of Negative Federated Learning
by Hong Lin, Lidan Shou, Ke Chen, Gang Chen, Sai Wu
First submitted to arxiv on: 7 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
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 Medium Difficulty summary: Federated Learning (FL) enables models to be learned from distributed data without compromising privacy. While FL works well in ideal federations, it may fail when the federation is unhealthy, leading to Negative Federated Learning (NFL). Previous solutions either prevent NFL upfront or address it after numerous learning rounds, incurring additional costs or wasting learning time. This paper proposes FL-GUARD, a framework that dynamically detects NFL and recovers from it during runtime. It uses an estimation of performance gain on clients to detect NFL and activates recovery measures when necessary. Clients learn adapted models in parallel with the global model to recover from NFL. Experimental results demonstrate FL-GUARD’s effectiveness in detecting NFL and recovering to a healthy learning state, while being compatible with previous solutions and robust against reluctant clients. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty summary: Federated Learning is a way for many devices to work together to create a model without sharing sensitive information. Sometimes, this collaboration doesn’t go well, leading to problems. Previous attempts to fix these issues either try to prevent the problem from happening or wait until it’s already occurred. This paper proposes a new approach called FL-GUARD that can detect when things are going wrong and take steps to correct the issue while still learning. It works by checking how much each device is benefiting from working together, and if they’re not getting enough benefits, it takes steps to help them catch up. The results show that this approach is effective in fixing problems and keeping the collaboration going smoothly. |
Keywords
* Artificial intelligence * Federated learning
