Thank you for your interest in our work!

This is the code for the paper Federated Few-shot Learning, published in SIGKDD 2023.

torch==1.11.0+cu113
torchvision==0.12.0+cu113  

First download the data file from here and unzip it into the folder 'data'.

To run the command for image datasets, i.e., 'miniImageNet' and 'FC100':

python main_image.py --dataset dataset_name

To run the command for text datasets, i.e., '20newsgroup' and 'huffpost':

python main_text.py --dataset dataset_name

Note that the text model requires the GloVe embedding file named 'glove.42B.300d.zip', which should be put in the main folder. The download link is here.

Welcome to cite our work!

@inproceedings{wang2022faith,
title={Federated Few-shot Learning},
author={Wang, Song and Fu, Xingbo and Ding, Kaize and Chen, Chen and Chen, Huiyuan and Li, Jundong},
booktitle={SIGKDD},
year={2023}
}