Co-classification sous contraintes par la somme des résidus quadratiques

In many applications, the expert interpretation of co-clustering is easier than for monodimensional clustering. Co-clustering aims at computing a bi-partition or a collection of coclusters: each co-cluster is a group of objects associated to a group of attributes and these associations can support interpretations. Many constrained clustering algorithms have been proposed to exploit the domain knowledge and improve partition relevancy in the monodimensional case, e.g., by using "must-link" and "cannot-link" constraints. Here, we consider constrained co-clustering for these constraints extended to both dimensions of objects and attributes, but also for interval constraints that enforce properties of co-clusters when considering ordered domains. We propose an iterative co-clustering algorithm which exploit user-defined constraints while minimizing the sum-squared residues. We show the added value of our approach in applications in transcriptomics.