Maximum Likelihood Estimation of Flexible Survival Densities with Importance Sampling

Mert Ketenci, Shreyas Bhave, Noemie Elhadad, Adler Perotte
Proceedings of the 8th Machine Learning for Healthcare Conference, PMLR 219:360-380, 2023.

Abstract

Survival analysis is a widely-used technique for analyzing time-to-event data in the presence of censoring. In recent years, numerous survival analysis methods have emerged which scale to large datasets and relax traditional assumptions such as proportional hazards. These models, while being performant, are very sensitive to model hyperparameters including: (1) number of bins and bin size for discrete models and (2) number of cluster assignments for mixture-based models. Each of these choices requires extensive tuning by practitioners to achieve optimal performance. In addition, we demonstrate in empirical studies that: (1) optimal bin size may drastically differ based on the metric of interest (e.g., concordance vs brier score), and (2) mixture models may suffer from mode collapse and numerical instability. We propose a survival analysis approach which eliminates the need to tune hyperparameters such as mixture assignments and bin sizes, reducing the burden on practitioners. We show that the proposed approach matches or outperforms baselines on several real-world datasets.

Cite this Paper

BibTeX
@InProceedings{pmlr-v219-ketenci23a, title = {Maximum Likelihood Estimation of Flexible Survival Densities with Importance Sampling}, author = {Ketenci, Mert and Bhave, Shreyas and Elhadad, Noemie and Perotte, Adler}, booktitle = {Proceedings of the 8th Machine Learning for Healthcare Conference}, pages = {360--380}, year = {2023}, editor = {Deshpande, Kaivalya and Fiterau, Madalina and Joshi, Shalmali and Lipton, Zachary and Ranganath, Rajesh and Urteaga, Iñigo and Yeung, Serene}, volume = {219}, series = {Proceedings of Machine Learning Research}, month = {11--12 Aug}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v219/ketenci23a/ketenci23a.pdf}, url = {https://proceedings.mlr.press/v219/ketenci23a.html}, abstract = {Survival analysis is a widely-used technique for analyzing time-to-event data in the presence of censoring. In recent years, numerous survival analysis methods have emerged which scale to large datasets and relax traditional assumptions such as proportional hazards. These models, while being performant, are very sensitive to model hyperparameters including: (1) number of bins and bin size for discrete models and (2) number of cluster assignments for mixture-based models. Each of these choices requires extensive tuning by practitioners to achieve optimal performance. In addition, we demonstrate in empirical studies that: (1) optimal bin size may drastically differ based on the metric of interest (e.g., concordance vs brier score), and (2) mixture models may suffer from mode collapse and numerical instability. We propose a survival analysis approach which eliminates the need to tune hyperparameters such as mixture assignments and bin sizes, reducing the burden on practitioners. We show that the proposed approach matches or outperforms baselines on several real-world datasets.} }
Endnote
%0 Conference Paper %T Maximum Likelihood Estimation of Flexible Survival Densities with Importance Sampling %A Mert Ketenci %A Shreyas Bhave %A Noemie Elhadad %A Adler Perotte %B Proceedings of the 8th Machine Learning for Healthcare Conference %C Proceedings of Machine Learning Research %D 2023 %E Kaivalya Deshpande %E Madalina Fiterau %E Shalmali Joshi %E Zachary Lipton %E Rajesh Ranganath %E Iñigo Urteaga %E Serene Yeung %F pmlr-v219-ketenci23a %I PMLR %P 360--380 %U https://proceedings.mlr.press/v219/ketenci23a.html %V 219 %X Survival analysis is a widely-used technique for analyzing time-to-event data in the presence of censoring. In recent years, numerous survival analysis methods have emerged which scale to large datasets and relax traditional assumptions such as proportional hazards. These models, while being performant, are very sensitive to model hyperparameters including: (1) number of bins and bin size for discrete models and (2) number of cluster assignments for mixture-based models. Each of these choices requires extensive tuning by practitioners to achieve optimal performance. In addition, we demonstrate in empirical studies that: (1) optimal bin size may drastically differ based on the metric of interest (e.g., concordance vs brier score), and (2) mixture models may suffer from mode collapse and numerical instability. We propose a survival analysis approach which eliminates the need to tune hyperparameters such as mixture assignments and bin sizes, reducing the burden on practitioners. We show that the proposed approach matches or outperforms baselines on several real-world datasets.
APA
Ketenci, M., Bhave, S., Elhadad, N. & Perotte, A.. (2023). Maximum Likelihood Estimation of Flexible Survival Densities with Importance Sampling. Proceedings of the 8th Machine Learning for Healthcare Conference, in Proceedings of Machine Learning Research 219:360-380 Available from https://proceedings.mlr.press/v219/ketenci23a.html.

Related Material