Non-parametric Approach to Extract Information from Interspike Intervals

A nonparametric approach is developed to extract information from interspike interval data. In terms of Expectation-Maximization (EM) algorithm, interspike interval data from experiments and simulations are first approximated by a mixture of various probability distributions, including Gamma, inverse Gaussian, log-normal, and the interspike interval distribution of the leaky integrate-and-fire model. We demonstrate that our approach is successful when fitting benchmark data which failed to be fitted in the literature. Also, an approach to fit mixture distributions to censored data, collected naturally in trial-to-trial or multi-electrode array experiments, is presented. The software to perform above computations is available at http://www.informatics.sussex.ac.uk/users/er28/em/. Our results demonstrate how to efficiently and rapidly read out information from an ensemble of spike trains.