Perspective reversal caused by chaotic switching in PDP schema model

A new dynamics of chaotic switching is proposed based on a particular type of network, i.e., the PDP (parallel distributed processing) scheme model. By applying a mean-field approximation, a one-dimensional map is derived, by which a variety of dynamics in the PDP schema model are reproduced. Introducing a dynamics of relative biases, successive chaotic switchings are realized. This new dynamics is applied to a well-known cognitive phenomenon of perspective reversal for ambiguous patterns such as a Necker cube. The chaotic states of network activations are interpreted as wondering states to search a new percept. New experimental results are given, which are categorized into three states: (1) one percept with confidence, (2) alternative percept, and (3) uncertain percept without confidence. The frequencies of persistent times in the three states are compared with simulations by logistic maps. It is shown that the chaotic theory could give new insights to cognitive science