The Neuromorphic Networks are electronic circuits that mime the physical structure of the nervous system through elements that act as synapses and neurons . The core innovation is the realization of a physical system, instead of a virtual simulation, that reproduces brain-like processes, able to react to external stimulations.
In the present invention, two transistors and one memnistor (2T1R) are the fundamental parts of the artificial synapses of the neuromorphic network. The transistors, miming the activities of the ionic channels in neuros, can generate short electrical pulses, called spikes ( comparable to neurons action potential) while the memnistor is the resistive element that describe the degree of electrical conductivity between two neurons. The time lapse between spikes determine the Synaptic Plasticity, the basis phenomenon of all cognitive functions, including learning and memory.
The neuronal structure 2T1R represents the first experimental realization of a neuronal circuit with biological learning spike-timing dependent plasticity (STDP). Moreover, the association of two transistors and one memnistor allows, contrary to other circuit models, a higher possibility of miniaturization with low complexity and low energy consumption.
The invention finds application in visual and auditory recognition of patterns for sensors in smartphones, smartwatches, vehicles, robots and drones and/or energy-autonomous for monitoring remote locations (deep sea, space orbits, ect.). The main application fields are Internet of Things, automotive, environmental monitoring and security in public places.