Chirag Nagpal's dcotoral thesis defense: May 2 11am

[Artur Dubrawski]

Team,

Please mark your calendars for another joyful event.

This time, it will be a summary of Chirag's recent contributions to the
science of survival analysis, for which he aspires to receive a doctorate.

I am sure everyone here is curious if that is indeed possible.
No better way to find that out than to attend the even and see it with your
own eyes.

See details below.

Cheers
Artur

=================================================
The LTI is proud to announce the following PhD Thesis Defense:

Leveraging Heterogeneity in Time-to-Event Predictions

Chirag Nagpal

Tuesday, May 2, 2023
11:00am (est.) NSH 1305 & Zoom
<https://cmu.zoom.us/j/94172364391?pwd=RW1KLzQxQTZtbURTVkZ2emJjOUV2Zz09>

*Committee:*

Artur Dubrawski, Robotics Institute, (Chair)
Louis-Philippe Morency
Bhiksha Ramakrishnan
Russell Greiner, (University of Alberta)
Katherine Heller, (Google Research)

*Abstract:*

Time-to-Event Regression, often referred to as Survival Analysis involves
learning of statistical estimators of the survival distribution of an
individual given their covariates. Unlike standard regression, survival
analysis requires accounting for outcomes censored due to loss of follow
up. This  circumstance is common in, e.g., biostatistics, predictive
maintenance, or econometrics. With the recent advances in machine learning
methodology, especially deep learning, it is now possible to exploit
expressive deep representations to help model survival outcomes. My thesis
contributes to this new body of work by demonstrating that problems in
survival analysis often manifest inherent heterogeneity which can be
effectively discovered, characterized, and modeled, in order to learn
better estimators of survival.

Heterogeneity may arise in a multitude of settings in the context of
survival analysis. Some examples include heterogeneity in the form of input
features or covariates (for instance, static vs. streaming, time-varying
data), or multiple outcomes of simultaneous interest (more commonly
referred to as competing risks). Other sources of heterogeneity involve
latent subgroups that manifest different base survival rates or diverse
responses to an intervention.

In this thesis, I aim to demonstrate that carefully modeling the inherent
structure of heterogeneity can boost predictive power of survival models
while improving their specificity and precision of estimated survival at an
individual level. An overarching methodological framework of this work is
the application of graphical models to reflect inherent structure in
time-to-event problems that explicitly model heterogeneity, while employing
advances in deep learning to learn powerful representations of data.
Furthermore, through innovative probabilistic and numerical optimization
techniques, we explore how the learnt estimators can become actionable
tools for decision support. By enforcing constraints that improve model
interpretability, we also explore opportunities for enhancing the utility
of resulting models, a requirement that is paramount in key application
scenarios such as healthcare.

A copy of the defense thesis can be found here
<https://drive.google.com/file/d/1ccYhCOFdzh57p3gp96GRrkAFti2DYz2P/view?usp=sharing>
.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20230421/f58dbb46/attachment.html>