Complexity pursuit: Separating interesting components from time series

A generalization of projection pursuit for time series, that is, signals wi th time structure, is introduced. The goal is to find projections of time s eries that have interesting structure, defined using criteria related to Ko lmogoroff complexity or coding length. Interesting signals are those that c an be coded with a short code length. We derive a simple approximation of c oding length that takes into account both the nongaussianity and the autoco rrelations of the time series. Also, we derive a simple algorithm for its a pproximative optimization. The resulting method is closely related to blind separation of nongaussian, time-dependent source signals.