Summary of Federated Impression For Learning with Distributed Heterogeneous Data, by Atrin Arya et al.
Federated Impression for Learning with Distributed Heterogeneous Data
by Atrin Arya, Sana Ayromlou, Armin Saadat, Purang Abolmaesumi, Xiaoxiao Li
First submitted to arxiv on: 11 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); 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 The paper proposes a novel approach called FedImpres for mitigating catastrophic forgetting during federated learning (FL) when training models on distributed datasets. FL is a decentralized learning method that enables the aggregation of knowledge from multiple sources without requiring data sharing, addressing privacy concerns. However, local training may lead to catastrophic forgetting due to heterogeneity in data collection protocols and patient demographics across participating health centers. FedImpres alleviates this issue by creating synthetic data representing global information through model distillation, which is then used alongside local data for enhanced generalization. Experimental results demonstrate state-of-the-art performance on BloodMNIST and Retina datasets with up to 20% improvement in classification accuracy. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper explores a new way to make machine learning work better when different groups of people share their own data without sharing it all together. Right now, this kind of collaboration can be tricky because each group might collect data differently or have different patients. This makes it hard for the computers to learn from all that data at once. The researchers came up with a new idea called FedImpres that helps solve this problem by making fake data that represents what’s common across all groups. They tested their idea on two big datasets and found that it really works, improving accuracy by as much as 20%. |
Keywords
» Artificial intelligence » Classification » Distillation » Federated learning » Generalization » Machine learning » Synthetic data
