* Corresponding author. Tel.: #41-21-692-5534; fax: #41-21-692-5505. E-mail addresses: avilla@lnh.unil.ch, lnh@lnh.unil.ch (A.E.P. Villa). Neurocomputing 38} 40 (2001) 1709 }1714 Pattern grouping algorithm and de-convolution "ltering of non-stationary correlated Poisson processes Igor V. Tetko, Alessandro E.P. Villa* Laboratory of Neuroheuristic, Institute of Physiology, University of Lausanne, Rue du Bugnon 7, CH-1005 Lausanne, Switzerland Department of Physiology of Brain, Bogomoletz Institute of Physiology and Department of Biomedical Applications, IBPC, Ukrainian Academy of Sciences, Kyiv, Ukraine Abstract The existence of precise temporal relations in sequences of spike intervals, referred to as `spatiotemporal patternsa, is suggested by brain theories that emphasize the role of temporal coding. A pattern grouping algorithm was designed to identify and to evaluate the statistical signi"cance of such patterns, particularly for data generated according to stationary Poisson processes. The experimental time series, however, can be characterized by considerable devi- ations from independent stationary Poisson processes. This article describes a "ltering method that de-convolute time series according to their correlation functions and makes possible an application of the pattern grouping algorithm for such data too.  2001 Elsevier Science B.V. All rights reserved. Keywords: Spatiotemporal "ring pattern; Burst "ltering; Temporal coding; Correlation functions 1. Introduction Multiple dimensions of behaviorally relevant stimuli are processed by thousands of neurons distributed over many areas of the brain. This processing is re#ected by changes in "ring rate and in the temporal structure of the spike trains*the time series formed by the sequences of time intervals between spikes. It has been hypothesized 0925-2312/01/$- see front matter  2001 Elsevier Science B.V. All rights reserved. PII:S0925-2312(01)00536-7