ABSTRACT
The systematic class of system dynamics-based and synapse-based architecture types includes those connectionistic models that can be subsumed under the classification of "Spiking Neural Networks" or "Spiking Neural Models". These neuroarchitectures are characterised by the fact that the dynamic time aspect is implemented as realistically as possible in neuronal network modelling, and thus the exact time of a spike's occurrence is of importance. These neuroarchitectures are characterized by the fact that algorithms which can be subject to "synaptic modulation" in different time scales are used, so that rapid changes in connection weights are possible. This chapter analyses the stable synchronization processes within the temporal dynamics of stochastic synapses. It discusses the special recurrent neuroarchitectures that have an internal, recurrent layer of neurons, the "reservoir". The chapter describes a dynamic binding mechanism by means of synfire waves to construct compositional representations. It presents a novel mechanism based on resonance frequencies.
