Clustering with Instance-Level Constraints

Kiri Wagstaff and Claire Cardie, Cornell University

We posit that problem-specific constraints can be incorporated into clustering algorithms to increase accuracy and decrease runtime. In experiments with a partitioning variant of COBWEB, we show marked improvements with surprisingly few constraints on three of four data sets. We also identify different types of constraints as appropriate in different settings.


This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.