Research map / Federated learning and privacy / Personalized and heterogeneous FL
Unlocking the Potential of Federated Learning: The Symphony of Dataset Distillation via Deep Generative Latents
Yuqi Jia, Saeed Vahidian, Jingwei Sun, Jianyi Zhang, Vyacheslav Kungurtsev, Neil Zhenqiang Gong, Yiran Chen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace aggregation-based FL (FedAvg) with server-side dataset distillation (FedDGM) under extreme heterogeneity | FedDGM · CIFAR-10 (Dir 0.01, average) | accuracy: 20.4 -> 63.4 | Improved |
| add generative-latent distillation on the server on top of FedDM-style distillation | FedDM · CIFAR-10 (Dir 0.01, average) | accuracy: 45.3 -> 63.4 | Improved |
| use FedDGM at near-iid partition (alpha 0.9) instead of FedDM | FedDGM · CIFAR-10 (Dir 0.9, average) | accuracy: 66.0 -> 67.5 | Improved |
| use FedDGM with a ViT global model at alpha 0.9 versus FedProx | FedDGM · CIFAR-10 (Dir 0.9, ViT) | accuracy: 55.0 -> 55.3 | No clear effect |
| use FedDGM on high-resolution ImageWoof with ResNet34 global model | FedDGM · ImageWoof (Dir 0.5, ResNet34) | accuracy: 60.3 -> 72.7 | Improved |
| use FedDGM instead of FedDM on ImageFood (ResNet34) | FedDGM · ImageFood (Dir 0.5, ResNet34) | accuracy: 59.8 -> 60.0 | No clear effect |
| distill in F9 latent space instead of F5 (later layer) (ablation) | FedDGM · CIFAR-10 | accuracy | Hurt |
| choose different StyleGAN-XL latent spaces among F0, F3, F4, F6 instead of F5 (ablation) | FedDGM · CIFAR-10 | accuracy | No clear effect |
| increase images per class (IPC) from 1 to 20 (ablation) | FedDGM · CIFAR-10 | accuracy | Improved |
| use simpler (2-layer) local surrogate model (ablation) | FedDM · CIFAR-10 | accuracy | Hurt |
| use simpler (2-layer) local surrogate model (ablation) | FedDGM · CIFAR-10 | accuracy | No clear effect |
| vary local epochs away from 20 (more or fewer) (ablation) | FedDGM · CIFAR-10 | accuracy | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FedDM — FedDM: Iterative Distribution Matching for Communication-Efficient Federated Learning (CVPR 2023) · better on 13 of 13 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- One-shot Federated Learning via Synthetic Distiller-Distillate Communication (NeurIPS 2024 · similarity 0.91)
- Fake It Till Make It: Federated Learning with Consensus-Oriented Generation (ICLR 2024 · similarity 0.91)
- Feature Distillation is the Better Choice for Model-Heterogeneous Federated Learning (NeurIPS 2025 · similarity 0.91)
- Federated Learning via Input-Output Collaborative Distillation (AAAI 2024 · similarity 0.90)
- Federated Learning with Heterogeneous LLMs: Integrating Small Student Client Models with a Large Hungry Model (AAAI 2025 · similarity 0.90)
- Data-Free Black-Box Federated Learning via Zeroth-Order Gradient Estimation (AAAI 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
