Overlapping between clusters is a major issue in clustering. In this cluster configuration, an object can belong to one or many clusters without any membership coefficient. Based on the assumption that an object really belongs to many clusters, overlapping clustering is different from both crisp and fuzzy clustering. Detecting overlapping structures and identifying clusters with complex shapes and forms are two major issues in this data mining task.

We present a kernel overlapping clustering algorithm called Kernel Overlapping k-Means (KOKM) to produce clusters in a high, possibly infinite, feature space. A non linear mapping of original data to a higher feature space is implicitly realized using Mercer kernels. The clusters prototypes and objects images are computed in input space and only distance between objects are computed in feature space. The proposed KOKM algorithm combines advantages of kernel k-means algorithm which allows detection of non linearly separable clusters and advantages of OKM algorithm (Cleuziou, 2008) which produces overlapping clusters.