Federated Learning with attention (A working title)

다기관에서, 각 기관별 모델에 대한 예측변수의 기여도가 이질적인 상황을 가정
전체 기관의 Attention Aggregation으로 종합된 예측 변수의 변수 기여도를 추출하고 평가

Setup

Clone this repo
"python -m venv .venv"
"source .venv/bin/activate"
"pip install -r requirements.txt"

Data

Results

To reproduce main results:

split terminal

cd 4_FL

terminal 1 : create server

python manage.py runserver

terminal 2 : create edge 1

python client.py --edge 0

terminal 3 : create edge 2

python client.py --edge 1

terminal 4 : create edge 3

python client.py --edge 2

result will be saved to /result/eicu

run plot.ipynb To plot a feather file as a heatmap.

Figures will be saved to /results/eicu.

Comparison of the results of learning with the entire data and the results of federated learning

Learning with the entire data (central).

python client.py --edge -1 --local true

Citation

If you use any of this code in your work, please reference us:

@misc{ ,
      title={Federated Learning with attention}, 
      author={},
      year={},
      eprint={},
      archivePrefix={},
      primaryClass={}
}

NeighborHeo/FLwithAttention