RegScore

Scoring Systems for Regression Tasks

Project setup

cd RegScrore
conda env create --file environment.yaml
conda activate regscore

RegScore

In this repository we provide the code for training RegScore models. To train RegScore on dataset with binary features run:

from regscore.wrapper import RegScore

rs = RegScore(k=5)
rs.fit(X_binary, y)
rs.print_risk_card(feature_names, X_binary, unit="mmHg")

Personalized Linear Regression (PLR)

To run self-supervised pretraining and training of PLR run:

CUDA_VISIBLE_DEVICES=0 python -u run.py --config-name config_ph_reg_REGSCORE scoring_strategy=personalized_linreg exp_name=plr

Personalized RegScore (PRS)

To run self-supervised pretraining and training of PRS run:

CUDA_VISIBLE_DEVICES=0 python -u run.py --config-name config_ph_reg_REGSCORE scoring_strategy=personalized_regscore exp_name=plr

We provide pretrained PLR and PRS weights here and here.

Acknowledgements

Our work is based on the following repositories:

• TIP
• FasterRisk
• MMCL