As in the author's earlier papers, the model consists of a diffuse neuronal net, where simultaneous stimuli cause correlations between the columns in the matrix of synaptic weights resulting in a sort of statistical mapping of the outer environment. However, the present version is endowed with two new features. Firstly, instead of the deterministic relation between the input signals and the output signal of neurons, only a probabilistic dependence is assumed which leads to the occurrence of spontaneous activity, even in the absence of any input signals. Secondly, a simple feedback is provided which adds up the signals from the associative layer—properly delayed—to the actual input. In the new model, the momentary state of memory changes continually, the resulting sequence of signals being chaotic in an “inexperienced” system. In a system with a fair amount of associations derived from the observation of the environment and interaction with it the spontaneous sequences are built up predominantly from materials supplied by the experience, and may be called “free associations”. These associations depend on the characteristics of the biological noise. Confronted with the question of problem solving procedures, the flow of associations can be visualized as a sort of random experimentation on the level of imaginative activity.
CitationDownload as .RIS
MCB UP Ltd
Copyright © 1974, MCB UP Limited