DescriptionDevelop a metric that measures the real similarities and differences between machine learning algorithms (in this case classifiers) based on output behavior. Previous study included 17 representative algorithms and used 30 datasets from the UCI Machine Learning Repository. The main goal of the current effort is to extend the metric using semi-supervised learning techniques. I would also like, if possible, to experiment with more recent datasets beyond what is traditional from the UCI Repository; and to add more algorithms to the study.
OrganizationThe Citadel
DepartmentMathematics and Computer Science
Sponsor Virtual OrganizationOSG
Principal Investigator
George Rudolph
Field Of ScienceComputer and Information Science and Engineering