Hi There. Do we plan to develop the GPU support for survival analysis? I am running a time to event analysis with large amount of obs and a very rare event, so it would be helpful to have GPU support. Or is there any suggestions of how to do survival analysis for this case?
Thank you!
XGBoost does not implement CoxPH, but you should try H2O if you have a big dataset and can scale horizontally https://github.com/h2oai/h2o-3
Actually I believe it implement coxPH. In the objective function https://xgboost.readthedocs.io/en/latest/gpu/index.html, there is survival:cox. And I can the first/second order gradient implemented in the src/objective/regression_obj.cu. In which the prediction is the non-parameter estimate of the effect of covariances instead of linear combination of covariate X’beta. It does not have the GPU support for coxPH now. Thank you for your suggestion, I will look into H2O.