$L\sb 1$ and $L\sb 2$ approximation clustering for mixed data: scatter decompositions and algorithms

Clustering is considered usually an art rather than a science because of lacking comprehensive mathematical theories in the discipline. The major issue raised in this paper is that L2 and L\ approximation bilinear clustering can provide a theoretical framework for an extensive part of partitioning and hierarchic clustering concerning its algorithmical and interpretational aspects, which is supported with a theoretical evidence.