SurvITE: Learning Heterogeneous Treatment Effects from Time-to-Event Data
Authors: Alicia Curth, Changhee Lee, Mihaela van der Schaar
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We investigate performance across a range of experimental settings and empirically confirm that our method outperforms baselines by addressing covariate shifts from various sources. |
| Researcher Affiliation | Academia | Alicia Curth University of Cambridge amc253@cam.ac.uk Changhee Lee Chung-Ang University changheelee@cau.ac.kr Mihaela van der Schaar University of Cambridge University of California, Los Angeles The Alan Turing Institute mv472@cam.ac.uk |
| Pseudocode | Yes | The pseudo-code of Surv ITE, the details of how to obtain Wass( , ) and how we set β can be found in Appendix D. |
| Open Source Code | Yes | The source code for Surv ITE is available in https://github.com/chl8856/surv ITE. |
| Open Datasets | Yes | Finally, we use the real-world dataset Twins [58] which has uncensored survival outcomes for twins (where the treatment is being born heavier ), and is hence free of Shifts 1 & 2. ... We consider the Twins benchmark dataset, containing survival times (in days, administratively censored at t=365) of 11400 pairs of twins, which is used in [58, 35] to measure HTEs of birthweight on infant mortality. |
| Dataset Splits | No | For synthetic experiments, the paper states: 'We use 5000 independently generated samples each for training and testing.' For the Twins dataset, it states: 'We split the data 50/50 for training and testing (by twin pairs)'. There is no explicit mention of a validation set split. |
| Hardware Specification | No | The paper does not explicitly describe the specific hardware (e.g., GPU/CPU models, memory amounts) used to run its experiments. |
| Software Dependencies | No | The paper mentions deep learning methods and refers to |
| Experiment Setup | Yes | Ltarget(θφ, θh) = Lrisk(θφ, θh) + βLipm(θφ) (6) where θh = {θha,τ }a {0,1},τ T , and β > 0 is a hyper-parameter. The pseudo-code of Surv ITE, the details of how to obtain Wass( , ) and how we set β can be found in Appendix D. |