Energy based evolving mean shift algorithm for neural spike classification

This paper presents a novel nonparametric clustering algorithm, called energy based evolving mean shift (EMS) clustering. It defines an energy function to characterize the compactness of the underlying data set and proves the clustering procedure converges. Through iterations, the data points collapse into well formed clusters and the associated energy approaches zero. Although as a general algorithm, the EMS is designed for resolving neural spikes to individual sources which is usually called ldquospike sortingrdquo.