CAREER: Ubiquitous, Large-Scale Machine Learning

University Of Washington, Seattle WA

Investigators

Abstract

This is the first year of funding of a 4 year continuing award. The goal of this project is to help machine learning make the leap from today's one shot, standalone applications to the ubiquitous, continuously functioning, seamlessly integrated web of learning systems of the future. The work will focus on four thrusts that are necessary to bring this about: (a) Scaling up machine learning algorithms so they can be applied to very large databases, and developing intrinsically scalable new algorithms; (b) Increasing the autonomy of learning systems through the use of automated data integration and improved overfitting avoidance; (c)Enabling learning systems to handle, data from heterogeneous sources and with different types of structure, such as the World Wide Web; (d) Educating a new generation of scientists to take advantage of the methodologies and tools provided by machine learning. If successful, the proposed work will lead to an order of magnitude or more increase in the size of databases to which machine learning algorithms can be usefully applied, to a significant reduction in the level of expertise required to develop and deploy a learning system, to learning systems that are able to function in environments of a greater variety and complexity than is currently the case, and to a new crop of scientists who are able to perform analyses that are currently beyond their reach with corresponding increase in the potential fornew discoveries and innovations.