Exploring One-Shot Semi-supervised Federated Learning with Pre-trained Diffusion Models
Authors: Mingzhao Yang, Shangchao Su, Bin Li, Xiangyang Xue
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on three large-scale datasets demonstrate that Fed DISC effectively addresses the semi-FL problem on non-IID clients and outperforms the compared SOTA methods. |
| Researcher Affiliation | Academia | Mingzhao Yang *, Shangchao Su *, Bin Li , Xiangyang Xue Shanghai Key Laboratory of Intelligent Information Processing School of Computer Science, Fudan University {mzyang20,scsu20,libin,xyxue}@fudan.edu.cn |
| Pseudocode | No | The paper describes the steps of the Fed DISC method in prose but does not include a formal pseudocode block or algorithm listing. |
| Open Source Code | No | The paper does not provide an explicit statement or link to the open-source code for the described methodology. |
| Open Datasets | Yes | Datasets. We adopt three datasets to evaluate the performance of Fed DISC: Domain Net (Peng et al. 2019), Open Image (Kuznetsova et al. 2020), and NICO++ (Zhang et al. 2022b). |
| Dataset Splits | No | The paper states 'We divide each dataset into six clients based on the inherent domain division of the dataset itself' but does not provide specific percentages or counts for training, validation, and test splits for the overall model or individual clients. |
| Hardware Specification | No | The paper does not provide specific details regarding the hardware used for running experiments, such as GPU models, CPU types, or cloud computing specifications. |
| Software Dependencies | No | The paper mentions using 'pre-trained CLIP image encoder' and 'Stable Diffusion' but does not specify version numbers for these or any other software dependencies, libraries, or programming languages. |
| Experiment Setup | No | The paper describes some aspects of the experimental setup, such as the use of CLIP and the structure of the classification model, and mentions hyperparameters like 'n' for noise intensity and 'wf', 'wg' for weights, and `L` and `R` in ablation studies, but does not provide their specific numerical values for the main experiments, nor does it specify training details like learning rate, batch size, or epochs. |