The Receptron: A device for the implementation of information processing systems based on complex nanostructured systems

Abstract

Unconventional Computing (UComp) identifies several data processing paradigms focused on exploiting emergent complexity and collective phenomena from various classes of physical substrates. Among UComp platforms, neuromorphic artificial systems aim at the reproduction of the human brain functions in terms of classification and pattern recognition capabilities, overcoming the limitations of traditional digital computers and closing the gap with the energetic efficiency of biological systems. Here we present a model, the receptron, and its physical implementation via a neuromorphic system which opens the way for the exploitation of complex networks of reconfigurable elements. Recently we have reported that nanostructured Au films, fabricated from gold clusters produced in the gas phase, have non-linear and non-local electric conduction properties caused by the extremely high density of grain boundaries and the resulting complex arrangement of nanojunctions. Exploiting these non-linear and non-local properties we produced and tested a device, based on a generalization of the perceptron, named receptron, that can receive inputs from different electrode configurations and generate a complete set of Boolean functions of n variables for classification tasks. The receptron allows also the classification of non-linearly separable functions without previous training of the device. Spatial correlations and the re-organization of the nanojunctions of the cluster-assembled film upon the application of suitable electrical stimuli are the enabling features for the efficient exploration of an extremely large number of weights configurations and hence the capability of the receptron to perform complex tasks.