Multi-source data clustering

In this paper, we consider a special multi-source data clustering problem for which the data-points from the same source cannot be grouped into the same cluster, namely cannot link (CL) constraint, and the sizes of the generated clusters are subject to maximum thresholds. No prior information is given about the level of clutter (namely noisy data) or the number of clusters. Particularly, the clusters might be closely distributed in the space (overlapping clusters) with one another and have to be carefully partitioned to meet the CL constraint. This particular CL constrained data mining problem corresponds to a significant problem of multi-sensor data fusion (MSDF) raised in the multi-target detection context. A novel clustering method as well as the online parameter learning procedure is proposed for this particular dataset model. Clustering results are provided to demonstrate the validity of the present approach.