https://github.com/cobrbra/counterfactual_survival_analysis
ACM CHIL 2021: "Enabling Counterfactual Survival Analysis with Balanced Representations"
Science Score: 10.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
○CITATION.cff file
-
○codemeta.json file
-
○.zenodo.json file
-
○DOI references
-
✓Academic publication links
Links to: arxiv.org, ieee.org -
○Academic email domains
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (10.3%) to scientific vocabulary
Repository
ACM CHIL 2021: "Enabling Counterfactual Survival Analysis with Balanced Representations"
Basic Info
- Host: GitHub
- Owner: cobrbra
- License: gpl-3.0
- Default Branch: main
- Homepage: https://dl.acm.org/doi/10.1145/3450439.3451875
- Size: 1.62 MB
Statistics
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
- Releases: 0
Metadata Files
README.md
Enabling Counterfactual Survival Analysis with Balanced Representations (ACM CHIL 2021)
This repository contains the Pytorch code to replicate experiments in our paper Enabling Counterfactual Survival Analysis with Balanced Representations accepted at ACM Conference on Health, Inference, and Learning (ACM CHIL) 2021:
latex
@inproceedings{chapfuwa2021enabling,
title={Enabling Counterfactual Survival Analysis with Balanced Representations},
author={Chapfuwa, Paidamoyo and Assaad, Serge and Zeng, Shuxi and Pencina, Michael J and Carin, Lawrence and Henao, Ricardo},
booktitle={ACM Conference on Health, Inference, and Learning},
year={2021}
}
Model
Data
- ACTG: A longitudinal RCT study comparing monotherapy with Zidovudine or Didanosine with combination therapy in HIV patients
- Framingham: A subset (Framingham Offspring) of the longitudinal study of heart disease dataset, for predicting the effects of statins on survival time
- See actg_synthetic.ipynb to modify the generated ACTG-Synthetic data
Prerequisites
The code is implemented with the following dependencies:
- Python 3.6.4
- Torch 0.4.1
- Additional python packages can be installed by running:
pip install -r requirements.txt
Model Training
- To train all models (CSA, CSA-INFO, AFT, AFT-Weibull, SR) run param_search.py
python param_search.py
- The hyper-parameters settings can be found at .config/configs.txt
Metrics and Visualizations
Once the networks are trained and the results are saved, we extract the following key results:
- Training and evaluation metrics are logged in model_*.log
- Epoch based cost function plots can be found in the plots directory
- Numpy files to generate calibration and cluster plots are saved in matrix directory
- To select best validation alpha compute likelihoods according to evaluationalphaselection.ipynb
- Run the evaluation_ITE.ipynb to generate factual and causal metrics and evaluationstratifyHR.ipynb for HR risk stratification
Acknowledgments
This work leverages the calibration framework from SFM and the accuracy objective from DATE. Contact Paidamoyo for issues relevant to this project.
Owner
- Login: cobrbra
- Kind: user
- Repositories: 3
- Profile: https://github.com/cobrbra