Summary of Federa:efficient Fine-tuning Of Language Models in Federated Learning Leveraging Weight Decomposition, by Yuxuan Yan et al.
FeDeRA:Efficient Fine-tuning of Language Models in Federated Learning Leveraging Weight Decomposition
by Yuxuan Yan, Qianqian Yang, Shunpu Tang, Zhiguo Shi
First submitted to arxiv on: 29 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
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 paper proposes a novel federated learning (FL) approach called FeDeRA that improves upon existing parameter-efficient fine-tuning (PEFT) methods. The authors recognize the limitations of pre-trained language models (PLMs) in FL due to their large number of parameters and explore ways to address this issue. Specifically, they extend a widely used PEFT method called low-rank adaption (LoRA) by initializing low-rank matrices using singular value decomposition (SVD) on the pre-trained weight matrices. This approach requires only 1% trainable parameters compared to full parameter fine-tuning (FFT), reducing training time costs by more than 90%. The authors demonstrate the effectiveness of FeDeRA through extensive experiments across various tasks and datasets, showing that it outperforms PEFT baselines and is comparable to or even surpasses FFT in terms of task performance. Additionally, FeDeRA exhibits robustness against data heterogeneity. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper introduces a new way to fine-tune pre-trained language models (PLMs) for federated learning. Right now, PLMs are really good at doing certain tasks after being trained on lots of data, but they have a big problem: they need a lot of data and computation power. The authors propose a solution called FeDeRA that makes it more efficient to fine-tune these models by using only the most important parts of the model. This helps reduce the amount of data and computation needed, making it more practical for real-world use. The authors test this approach on different tasks and show that it works well and is even better than some other methods. |
Keywords
» Artificial intelligence » Federated learning » Fine tuning » Lora » Parameter efficient
