Biochemical neuron: hardware implementation of functional devices by mimicking the natural intelligence such as metabolic control systems

Abstract

Mimicking the switching property of cyclic enzyme systems in metabolic pathways, we have proposed a different type of molecular switching device of which mechanism can be represented by threshold-logic function capable of storing short-term memory or ‘Hebbian rule’ or ‘post-synaptic neuron’ with synaptic history. The great advantage of this device is that it can process the time-variant input signals. We have named this system a ‘biochemical neuron’ and have developed the board-leveled analog circuit. In the present study, building the integrated artificial neural network system being composed of biochemical neurons, we have examined whether this network can recognize the pattern similarity in time-variant external analog signals or not. Our neural network showed highly recognition of total time-variant patterns of external analog signals even if signals involve an uniformed random noise.