Decision-making over complex model spaces
Abstract
The PI, Dr. Venkat Chandrasekaran, is exploring methods by which to identify features, the data, that bias how to learn-reason. In machine learning and data analytic approaches to scientific discovery, one is commonly faced with the problem of model (the data) selection. This task is usually approached by leveraging the framework of statistical hypothesis testing in which the scientist seeks to maximize the amount of discovery by selecting the most complex model that is supported by the data, while retaining control over the amount of false discovery. A consequence of model complexity can be conveniently specified as the number of variables-edges in a model and the amount of false discovery in a selected model corresponds to the number of variables-edges that are incorrectly included in the model. The PI seeks new methods to select models that represent a far richer range of phenomena.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 29, 2024
- Source ID
- FA95502310204
Entities
People
- Venkat Chandrasekaran
Organizations
- Air Force Office of Scientific Research
- California Institute of Technology
- United States Air Force