@article{wang2021survtrace,
title={Surv{TRACE}: Transformers for Survival Analysis with Competing Events},
author={Zifeng Wang and Jimeng Sun},
year={2021},
eprint={2110.00855},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
https://github.com/RyanWangZf/SurvTRACE
easydict==1.9
scikit_survival==0.14.0
pandas==1.1.5
torchtuples==0.2.2
numpy==1.19.5
torch==1.9.1
matplotlib==3.3.4
pycox==0.2.2
scikit_learn
Import metabric and support from pycox.datasets could directly generates the dataframe of dataset. Detail can be seen in both the notebooks and the run.py file.
For baseline models, please refer to the jupyter notebooks. For SurvTRACE, simply run the run.py file. To change the dataset,you need to change the dataset in run.py. To change the lambda value and enable multi-task learning, please change the corresponding parameter in the survtrace/config file.
- The results on SUPPORT, with claimed results in bracket
- The results on METABRIC, with claimed results in bracket