Parallel Clustering Method based on Density Peaks

This work develops a clustering method which is based on the identi-fication of density peaks in the available data. The realization is characterized by its goal of carrying in parallel as many operations as it is possible, and to ex-ploit current commodity many/multi core machines with shared memory. The algorithm is prototyped in Java using parallel streams and lambda expressions. The paper first describes the rationale underlying the design and implementa-tion of the clustering method. Then the tool is practically applied to several and challenging benchmark datasets, for example admitting general not spherical clusters. The experimental results confirm efficiency and reliability of the pro-posed method. Finally, conclusions are drawn with an indication of on-going and future work.