ABSTRACT
The systematic class of information-based architecture types contains those connectionist models which strive for a solution to the binding problem by means of a coherent and integrated process organization of the neural information. This chapter defines three possible measures. First, an information-theoretical measure can be defined on the basis of a reentrant process organization that is the synchronization of neurophysiological activity between groups of neuronal populations. A second possible measure is an "optimally harmonic" information structure. Thirdly, the measure of the minimization of "free energy" can be defined. In order to approach the degree of cognitive flexibility present in human action and behavior, it is necessary to return to the description of those knowledge atoms that represent the elements of cognitive computation, and not the schemata and scripts. Thus, the problem of modeling schemata and scripts is subsumed under the problem of modeling vectorial information atoms by developing the harmony function as the measure of self-consistency of a neural network.
