We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ML

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: SODEN: A Scalable Continuous-Time Survival Model through Ordinary Differential Equation Networks

Abstract: In this paper, we propose a flexible model for survival analysis using neural networks along with scalable optimization algorithms. One key technical challenge for directly applying maximum likelihood estimation (MLE) to censored data is that evaluating the objective function and its gradients with respect to model parameters requires the calculation of integrals. To address this challenge, we recognize that the MLE for censored data can be viewed as a differential-equation constrained optimization problem, a novel perspective. Following this connection, we model the distribution of event time through an ordinary differential equation and utilize efficient ODE solvers and adjoint sensitivity analysis to numerically evaluate the likelihood and the gradients. Using this approach, we are able to 1) provide a broad family of continuous-time survival distributions without strong structural assumptions, 2) obtain powerful feature representations using neural networks, and 3) allow efficient estimation of the model in large-scale applications using stochastic gradient descent. Through both simulation studies and real-world data examples, we demonstrate the effectiveness of the proposed method in comparison to existing state-of-the-art deep learning survival analysis models. The implementation of the proposed SODEN approach has been made publicly available at this https URL
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG)
Cite as: arXiv:2008.08637 [stat.ML]
  (or arXiv:2008.08637v2 [stat.ML] for this version)

Submission history

From: Weijing Tang [view email]
[v1] Wed, 19 Aug 2020 19:11:25 GMT (234kb,D)
[v2] Mon, 6 Dec 2021 03:49:01 GMT (226kb,D)

Link back to: arXiv, form interface, contact.