The Model of the Corrupted Clique Graph

The clustering problem can be modeled by a corrupted clique graph. A clique graph is a graph consisting of disjoint cliques. The true clustering is represented by a clique graph H (vertices are genes and cliques are clusters). Contamination errors introduced into gene expression data result in similarity graph a C(H) which is not a clique graph. Under this model the problem of clustering is as follows: given C(H), restore the original clique graph H and thus the true clustering.