DYNAMICAL SYMMETRIES AND TEMPORAL SEGMENTATION

Segmentation of a mixed input into recognizable patterns is a task tha t is common to many perceptual functions. It can be realized in neural models through temporal segmentation: formation of staggered oscillat ions such that within each period every nonlinear oscillator peaks onc e and is dominant for a short while. We investigate such behavior in a symmetric dynamical system. The fully segmented mode is one type of l imit cycle that this system can exhibit. We discuss its symmetry class ification and its dynamical characterization. We observe that it can b e sustained for only a small number of segments and relate this fact t o a limitation on the appearance of narrow subharmonic oscillations in our system.