Pattern-recognition by an artificial network derived from biologic neuronal systems

Abstract

A novel artificial neural network, derived from neurobiological observations, is described and examples of its performance are presented. This DYnamically STable Associative Learning (DYSTAL) network associatively learns both correlations and anticorrelations, and can be configured to classify or restore patterns with only a change in the number of output units. DYSTAL exhibits some particularly desirable properties: computational effort scales linearly with the number of connections, i.e., it is O(N) in complexity; performance of the network is stable with respect to network parameters over wide ranges of their values and over the size of the input field; storage of a very large number of patterns is possible; patterns need not be orthogonal; network connections are not restricted to multi-layer feed-forward or any other specific structure; and, for a known set of deterministic input patterns, the network weights can be computed, a priori, in closed form. The network has been associatively trained to perform the XOR function as well as other classification tasks. The network has also been trained to restore patterns obscured by binary or analog noise. Neither global nor local feedback connections are required during learning; hence the network is particularly suitable for hardware (VLSI) implementation.

Synchrony detection in neural assemblies

“…Both the nature of postsynaptic processing and the stimulation of long-term potentiation (LTP) effects are influenced by temporal synchronies of arriving impulses (Hall 1992). Temporal synchronies also appear in neural processing models as well as in recorded multiunit data (Alkon et al 1990;Lindsey et al 1989). Synchronies are, in addition, a possible means of superimposing multiple signals, independent of the firing rate codes (Hall 1992).…”

The magnetic and electric fields induced by superparamagnetic magnetite in honeybees

“…Then for each image, a certain configuration of neurons will receive visual and magnetic inputs simultaneously. On repetition, synaptic patches on these neurons may memorize the coincidence of such inputs (Alkon et al 1990) from vision and from the magnetic sensors. Correlation could be associatively learned, such that with an overcast sky the magnetic input alone may be enough for determining the position of the sun or, vice versa, the visual input alone may be enough for orientation.…”

“…We propose a model for associative learning and the development of specific activation patterns in a neuron population. This model is similar to a combination of the model for learning correlations between simultaneous inputs in adjacent synaptic patches (Alkon et al 1990) and the EEG recordings showing the activity of neuron ensembles in the vertebrate olfactory system (Freeman 1991). The model thus derived would explain some of the mysteries in the electrophysiological recordings from single neurons as well as in experiments involving training of bees .…”

Faust:

“…A diagram of the FAUST system is shown in figure 1. This [13]. The correlation based method is the one used in [4] (2).…”