Multiple spike-train analysis using mutual interval matrix.

The principal-component approach is applied to the analysis of sequences of neuronal action potentials (spike trains). Multiple spike trains are represented as a sequence of vectors of mutual interspike intervals and are considered to be part of the trajectory of a dynamic system. The trajectory matrix is decomposed into a number of 'basic spike patterns' and their relative magnitudes by singular-value decomposition. The representation provides a convenient framework for analysis of dynamic relations and cooperation between neurons in an observed network. Examples of applications to simulated and cerebellar data are presented.