Reliable Variable Selection for Biomedical Data Science: From Shrinkage Estimation to Interpretable Learning
IDWSDS 2026, Session S115, sponsored by the Bernoulli SocietyAbstract
In high-dimensional biomedical data, choosing variables reliably matters as much as predicting well. This talk moves from shrinkage and penalized estimation to correlation-informed adaptive regularization, and shows how these ideas lead to stable, interpretable variable selection, illustrated with glioma subtype prediction. Joint work with Marta Belchior Lopes and Tomás da Rosa Bandeira (MSc thesis, NOVA FCT).
Date
Oct 6, 2026 1:30 PM — 2:00 PM
Event
International Day of Women in Statistics and Data Science (IDWSDS 2026), Plenary Session S115
Location
Online
Many thanks to the Bernoulli Society for Mathematical Statistics and Probability for selecting and sponsoring this session, to the session chair Chiara Orsini, and to the IDWSDS organisers.
The slides are interactive: use the arrow keys to move between slides, N to show speaker notes and F for full screen.
