The Transfer Learning Exploratory Graphical Models

DATE: November 17, 2009

TIME: 11:00 AM – 12:00 PM

LOCATION: Executive classroom

FEES: none

EVENT CONTACT:

Anita Race, H. Milton Stewart School of Industrial and Systems Engineering
Contact Anita Race

TITLE: The Transfer Learning Exploratory Graphical Models with Application in Alzheimer’s Disease

SPEAKER: Dr. Jing Li

ABSTRACT:

Knowledge discovery in fields such as medical informatics, bioinformatics, and modern manufacturing systems/processes often requires building statistical models in which many random variables interact with each other in complex ways. Graphical models provide a general methodology for learning the complex relationships from observational data. Transfer learning of graphical models deal with the situations where graphical models need to be learned for multiple related groups of subjects, so that the knowledge/information gained during the learning of one group can be effectively transferred to the learning of another group. Such “groups” can be, for example, patients with similar but not identical neurological disorders, or manufacturing products belonging to several product varieties. The concept of “transfer learning” was originated in psychology and recently has been introduced into statistics and machine learning society. This talk will present our recent development in transfer learning of undirected graphical models, which integrates statistics and optimization. The developed methodology is applied to build brain connectivity networks from neuroimaging data for Alzheimer’s Disease studies.