ISOMAP based metrics for clustering
Ariel E. Bayá, Pablo M. Granitto · INTELIGENCIA ARTIFICIAL · 2008
"Many successful clustering techniques fail to handle data with a manifold structure, i.e. data that is notshaped in the form of compact point clouds, forming arbitrary shapes or paths through a high-dimensionalspace. In this paper we present a new method to evaluate distances in such spaces that naturally extend theapplication of many clustering algorithms to these cases. Our algorithm has two stages. Following ISOMAP,it searches for sets of locally-uniform manifolds, which could be disjoint. These manifolds are then connectedusing two slightly different strategies. We compare these strategies between them and with a state of theart algorithm using three artificial problems, obtaining encouraging result. Both new metrics allow diversealgorithms to easily find clusters of arbitrary shape"