Kernel method for clustering based on optimal target vector

We introduce the notion of optimal target vector, and describe how it creates a link between supervised and unsupervised learning. We exploit this notion to construct Ising models, for dichotomic clustering, whose couplings are (i) both ferro- and anti-ferromagnetic (ii) depending on the whole data-set and not only on pairs of samples. The effectiveness of the method is shown in the case of the well known iris data-set and in benchmarks of gene expression levels.