Dynamic threshold neural network for modeling interference in a sequence of associative memories

Modeling disorders of Cognitive Systems using Neural Network (NN) is a field of research now being developed. The goal of this work is to model some basic features of thought processes by a Dynamic Threshold Neural Networks (DTNN). The definition of 'thought process' is restricted to an orderly transition from one pattern to another, where each pattern can represent a word or a mental concept. This sequence of transitions is initiated by an external stimulus pattern. Global organization of information processing is governed by the dynamic threshold parameters. Deviate values of these parameters can cause some major sequencing breakdowns. Damage to stimulus-dependent memory retrieval may stimulate loosening of associations and delusions, while convergence into fixed memory states may simulate constriction of thought content and poverty of thought content. From the engineering point of view, this system can serve as a controllable 'smart' buffering and delay in NN architectures that involve temporal analysis, without needing an external clock. From the neurophysiological point of view, this system can suggest a possible framework in the effort to understand 'normal' and 'pathological' computations carried out by the neural system.