FedCIA
A federated recommendation framework for collaborative information aggregation
Install / Use
/learn @Mingzhe-Han/FedCIAREADME
FedCIA
This repository is for the paper FedCIA: Federated Collaborative Information Aggregation for Privacy-Preserving Recommendation.
The paper has been accepted in SIGIR 2025.
Running
A log file is needed in our file directory.
Run experiments on the baselines:
python -m entry.fedavg --config=config/ml100k/fedmf.ini.
Run experiments on our FedCIA:
python -m entry.fedcia_param --config=config/ml100k/fedmf.ini.
For GSP models:
python -m entry.fedcia_base --config=config/ml100k/fedgsp.ini.
For the ml100k dataset where only one user in a client:
python -m entry.fedcia_param --config=config/ml100k_all/fedmf.ini.
You can change self.fast_com in flalgorithm/fedcia_param.py to reduce data in uploading.
Repeat Experiments
If the output file in the config already exists, the experiment will not start.
You can change the output in config in [exp] to start a second experiment.
