ABSTRACT
This chapter focuses on the three architectures of oscillatory artificial neural networks which model the binding problem in the form of the problem of "feature binding". These architectures effectively model the binding problem by means of an integrative synchronization mechanism, as well as developing a "emulative neurosemantics". The systematic class of oscillator-based architecture types is based on the model of a neural oscillator, which is an instance of the more general "(nonlinear relaxation) oscillator". This represents a large class of nonlinear dynamic systems that occur in a variety of physical and biological systems. The chapter demonstrates that the neural structure provides a compositional semantics of the language. It shows that the elements of the neural structure are internal representations that reliably co-vary with external contents. These external contents are identical with the standard model-theoretical denotations for the language.
