Summary of Diffusion-driven Data Replay: a Novel Approach to Combat Forgetting in Federated Class Continual Learning, by Jinglin Liang et al.
Diffusion-Driven Data Replay: A Novel Approach to Combat Forgetting in Federated Class Continual Learning
by Jinglin Liang, Jin Zhong, Hanlin Gu, Zhongqi Lu, Xingxing Tang, Gang Dai, Shuangping Huang, Lixin Fan, Qiang Yang
First submitted to arxiv on: 2 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
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 In Federated Class Continual Learning (FCCL), the challenge of adapting to new classes without forgetting old ones is merged with the need for distributed client learning. To address this, we propose a novel method using data replay based on diffusion models. This approach reduces computational resources and time consumption while generating effective guidance for the model. Additionally, contrastive learning enhances the classifier’s domain generalization ability on generated and real data. Our method significantly outperforms existing baselines in comprehensive experiments. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated Class Continual Learning (FCCL) is a way to learn new classes without forgetting old ones while sharing data with many clients. The challenge is to make sure the model doesn’t forget what it learned before, but still adapts to new information. We came up with a new method that uses special models called diffusion models to help the main model remember old classes and adapt to new ones. This makes the whole process faster and more efficient. It also helps the model learn from both real data and fake data we generate. Our method works better than other ways of doing this, according to our tests. |
Keywords
» Artificial intelligence » Continual learning » Domain generalization
