Tensorflow implementation of "Personalized Federated Graph Learning on Non-IID Electronic Health Records".
- python>=3.7
- tensorflow==2.3.0
- numpy
- scipy
- pandas
Write the following code in "Tf.keras.models" model:
def set_weights(self, weight):
"""set the weights of the model.
Returns:
A flat list of Numpy arrays.
"""
with self.distribute_strategy.scope():
return super(Model, self).set_weights(weight)Easily find the place: In the same class as "PEARL_pretrain.get_weights()" in "FL_train.py". The form of "set_weights" is the same as that of "get_weights".
The parameters are configured in each python file.
python run_preprocess.py
python run_hyperbolic_embedding.py
python FL_train.pyPlease cite our paper if you use this code in your own work:
@ARTICLE{10475550,
author={Tang, Tao and Han, Zhuoyang and Cai, Zhen and Yu, Shuo and Zhou, Xiaokang and Oseni, Taiwo and Das, Sajal K.},
journal={IEEE Transactions on Neural Networks and Learning Systems},
title={Personalized Federated Graph Learning on Non-IID Electronic Health Records},
year={2024},
volume={},
number={},
pages={1-14},
keywords={Diseases;Predictive models;Federated learning;Data models;Adaptation models;Training;Task analysis;Disease prediction;electronic health record (EHR);graph neural network (GNN);non-independent and identically distributed (Non-IID) data;personalized federated learning},
doi={10.1109/TNNLS.2024.3370297}
}