ANALYSIS OF SWITCHING DYNAMICS WITH COMPETING NEURAL NETWORKS

We present a framework for the unsupervised segmentation of time serie s. It applies to non-stationary signals originating from different dyn amical systems which alternate in time, a phenomenon which appears in many natural systems. In our approach, predictors compete for data poi nts of a given time series. We combine competition and evolutionary in ertia to a learning rule. Under this learning rule the system evolves such that the predictors, which finally survive, unambiguously identif y the underlying processes. The segmentation achieved by this method i s very precise and transients are included, a fact, which makes our ap proach promising for future applications.