futianfan / pearl

deep prototype learning for EHR data

PEARL: Prototype Learning via Rule Learning

🚀 Installation

conda create -n pearl python=3.7 
conda activate pearl 
pip install torch 
pip install scikit-learn 
pip install matplotlib 

📊 Data

data/: data folder

src/stream.py: process data

⚙️ Configuration

src/config.py: configuration

🤖 Model

src/decision_tree.py: decision-tree for interpretable rule

src/model_torch.py: NN model (Pytorch)

train

python src/train_torch.py

📞 Contact

Tianfan Fu ([email protected])

pearl's Issues

Something about Prototype learning

Hello, I am reading your paper and code, it's a wonderful work. I noticed that the paper describes the pseudo representation of one rule as follows:

which h(x) is the embedded events, but only embedded events read from file are used in the code, it has nothing to do with the h(x) encoder.
Looking forward to your reply! Thank you!

Data to test PEARL

Thank you for making the code for this paper available. I was just wondering if you could supply the data that you used to train this model somehow. I would like to try out PEARL for myself.