An Efficient Expectation-Maximisation Algorithm for Spike Classification
Pedro Tomás, Leonel A. Sousa · 2007
This paper presents a new Expectation-Maximisation algorithm for classifying a data set originated from an unknown number of sources. The proposed algorithm is based on the Kullback-Leibler divergence and uses a minimum message length criteria to penalise adding extra data sources. It is able to estimate the parameters of the model for each data source and to determine the total number of sources producing the data. We apply our algorithm to the classification of spikes originated from multiple neurons but recorded by a single microelectrode. The obtained experimental results show the effectiveness of the proposed algorithm.