Summary of Federated Active Learning Framework For Efficient Annotation Strategy in Skin-lesion Classification, by Zhipeng Deng et al.
Federated Active Learning Framework for Efficient Annotation Strategy in Skin-lesion Classification
by Zhipeng Deng, Yuqiao Yang, Kenji Suzuki
First submitted to arxiv on: 17 Jun 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
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 Federated learning enables multiple institutions to train models collaboratively without sharing private data, but current research assumes ideal data collection. In medical scenarios, data annotation requires expertise and labor, a critical problem in federated learning. Active learning has shown promise in reducing annotations in medical image analysis. We propose a federated active learning framework, executing AL periodically under FL. We use ensemble-entropy-based AL as an efficient data-annotation strategy in FL, decreasing the amount of annotated data while maintaining performance. Our framework applies FedAL to medical images and validates it on real-world dermoscopic datasets. Using 50% of samples, our framework achieves state-of-the-art performance on a skin-lesion classification task, outperforming several state-of-the-art AL methods under FL. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated learning lets many places train models together without sharing secrets. Right now, researchers are working on making this process more efficient and private. But in medical situations, people need to label data by hand, which is hard work. Active learning has been good at reducing the amount of labeling needed in medical image analysis. We’re proposing a new way to do this, called federated active learning (FedAL). It lets us use local models at each hospital and combine them with a global model from all the hospitals. We use entropy-based active learning to pick which data points are most important to label. This helps us reduce the amount of labeled data we need while keeping our results good. We tested this on real medical images and were able to get better results using only half as much data. |
Keywords
* Artificial intelligence * Active learning * Classification * Federated learning
