Optoelectronic neuron based on transistor combined with volatile threshold switching memristors for neuromorphic...

Area of Science:

• Neuromorphic Engineering
• Biomimetic Vision Systems
• Artificial Intelligence Hardware

Background:

• The human visual system, particularly the retina, is crucial for preprocessing visual information, inspiring the development of neuromorphic vision hardware.
• Existing neuromorphic approaches aim to replicate the neurobiological functions of the retina for enhanced visual processing capabilities.

Purpose of the Study:

• To develop a novel optoelectronic neuron that mimics the functional behavior of retinal neurons.
• To integrate electrical and optical modulation for biologically plausible neuronal signal transmission.
• To evaluate the performance of a spiking neural network composed of these artificial neurons for image recognition tasks.

Main Methods:

• Fabrication of an optoelectronic neuron by combining a gate-modulated PDVT-10 channel with a volatile threshold switching memristor.
• Implementation of a resistance-matching mechanism to achieve optoelectronic performance.
• Simulation of a perceptron spiking neural network using the developed leaky integrate-and-fire neurons for image classification.

Main Results:

• The developed optoelectronic spiking neuron successfully altered its spiking behavior, mimicking retinal processing.
• The artificial neuron demonstrated biologically plausible signal transmission through combined electrical and optical modulation.
• Neuronal firing was inhibited in darkness and activated upon light exposure, replicating retinal responses.
• The perceptron spiking neural network showed capability in classifying image recognition algorithms.

Conclusions:

• The novel optoelectronic neuron effectively replicates key retinal functions, offering a promising pathway for biomimetic vision.
• This research paves the way for developing expandable and hierarchical spiking electronics for neuromorphic hardware applications.
• The findings contribute to advancing the field of neuromorphic vision with potential applications in image recognition and beyond.