Summary of Stable Diffusion-based Data Augmentation For Federated Learning with Non-iid Data, by Mahdi Morafah et al.
Stable Diffusion-based Data Augmentation for Federated Learning with Non-IID Data
by Mahdi Morafah, Matthias Reisser, Bill Lin, Christos Louizos
First submitted to arxiv on: 13 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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 introduction of Federated Learning (FL) has brought decentralized and collaborative model training to the forefront, while preserving clients’ data privacy. However, FL faces challenges with performance reduction and poor convergence when dealing with Non-Independent and Identically Distributed (Non-IID) data distributions among participating clients. To address this issue, researchers have employed client drift mitigation and advanced server-side model fusion techniques, but these efforts often overlook the root cause of the performance reduction – the absence of identical data mirroring the global data distribution among clients. This paper introduces Gen-FedSD, a novel approach that leverages state-of-the-art text-to-image foundation models to bridge Non-IID performance gaps in FL. Each client constructs textual prompts for each class label and uses an off-the-shelf pre-trained Stable Diffusion model to synthesize high-quality data samples. The generated synthetic data is tailored to each client’s unique local data gaps, effectively making the final augmented local data IID. Experimental results demonstrate that Gen-FedSD achieves state-of-the-art performance and significant communication cost savings across various datasets and Non-IID settings. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated Learning helps devices train models together while keeping their data private. But it has a big problem: when devices have different types of data, the model doesn’t work well. To fix this, researchers are using special text-to-image models to create new, fake training data that makes the devices’ data match each other. This paper shows how this idea works and proves it’s better than previous methods. |
Keywords
» Artificial intelligence » Diffusion model » Federated learning » Synthetic data
