Abstract: Manifold distance was used as the basic measure of the sample similarity between samples. The pair-wise constrains prior information was introduced,then the measure matrix was obtained through affinity propagation. So the clustering problem was transformed as one optimal model. Clonal selection algorithm was employed to solve this model, and the clustering results were given. Experiments on artificial data sets and UCI benchmark data set show that the pro- posed method can give the better accuracy.

Key words: Manifold distance, Semi supervised clustering, Artificial immune algorithm