Internal representations--a prelude for neurosemantics
Abstract
Following the concept of internal representations, signal processing in a neuronal system has to be evaluated exclusively on the basis of internal system characteristics. Thus, this approach omits the external observer as a control function for sensory integration. Instead, the configuration of the system and its computational performance are the effects of endogeneous factors. Such self-referential operation is due to a strictly local computation in a network. Thereby, computations follow a set of rules that constitutes the emergent behaviour of the system. Because these rules can be demonstrated to correspond to a "logic" intrinsic to the system, it can be shown that the concept of internal representation provides the basis for neurosemantics