Summary of Evidential Federated Learning For Skin Lesion Image Classification, by Rutger Hendrix et al.
Evidential Federated Learning for Skin Lesion Image Classification
by Rutger Hendrix, Federica Proietto Salanitri, Concetto Spampinato, Simone Palazzo, Ulas Bagci
First submitted to arxiv on: 15 Nov 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); 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 The proposed Federated Evidence Prompt (FedEvPrompt) is a distributed learning framework that combines principles of evidential deep learning, prompt tuning, and knowledge distillation to classify skin lesions. The approach utilizes two sets of prompts: basic visual knowledge (b-prompts) and task-specific knowledge (t-prompts), which are prepended to frozen pre-trained Vision Transformer models. These models are trained in an evidential learning framework that maximizes class evidences. Knowledge sharing across clients is achieved through distillation on attention maps generated by local models, ensuring privacy preservation. The algorithm optimizes performance within a round-based paradigm, where each round involves training local models and sharing attention maps with all clients. Experimental results on the ISIC2019 dataset demonstrate FedEvPrompt’s superior performance compared to baseline methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated learning is a way for devices or computers to work together to improve their skills without sharing their data. The new approach, called FedEvPrompt, helps these devices learn from each other by giving them hints about what they should focus on. This approach works better than others because it only shares hints, not actual information. It also does a good job of learning even when the data is different or not as good quality. This can help doctors and researchers better diagnose skin lesions without seeing individual patient data. |
Keywords
» Artificial intelligence » Attention » Deep learning » Distillation » Federated learning » Knowledge distillation » Prompt » Vision transformer
