Research > Faculty Projects

Learning from Heterogeneous Data

Principal Investigator
Maya Gupta

Sponsor(s)
Office of Naval Research

Award Period
08/01/2008 - 08/01/2011

Abstract
Current pattern recognition approaches are not well- developed to learn from a large set of disparate data sources that may include cell phone activity, descriptive features, hearsay, chemical sensors, estimates based on partial data, video data, etc. We propose to develop and analyze classification and regression algorithms given samples described by heterogeneous features, which may include: Euclidean features, descriptive features (categorical), similarities to other samples, and predictions based on partial data. We will use standard statistical learning theory as a framework to fuse the data in the context of learning. Our focus will be on optimal nearest-neighbor and local algorithms, and generative classifiers. As a secondary question, we will consider within this context the practical problem of learning given uncertainty in the test sample.

For more information, visit idl.ee.washington.edu

Updates or corrections to this page should be sent to gheaton@u.washington.edu.