Gate-controlled neuromodulatory optical synaptic transistor for adaptive learning and energy-accuracy balance

Area of Science:

• Materials Science
• Neuroscience
• Computer Science

Background:

• Neuromorphic vision systems require efficient optical signal processing and adaptive, energy-aware learning.
• Optical synaptic transistors are key for in-sensor computing, mimicking synaptic functions like plasticity.
• Existing devices have fixed synaptic gain, hindering adaptive learning for diverse tasks.

Purpose of the Study:

• To develop a gate-tunable optical synaptic transistor inspired by biological neuromodulation.
• To enable voltage-dependent modulation of learning sensitivity in neuromorphic vision systems.
• To integrate synaptic behaviors with gate-controlled gain modulation for adaptive computing.

Main Methods:

• Fabrication of a gate-tunable optical synaptic transistor using Indium Gallium Zinc Oxide (IGZO).
• Demonstration of conventional synaptic behaviors (EPSC, PPF, plasticity).
• Implementation of gate bias for pre-conditioning synaptic response and mimicking neuromodulation.
• Training a Convolutional Neural Network (CNN) on the CIFAR-10 dataset to evaluate device performance.

Main Results:

• The IGZO phototransistor successfully mimicked synaptic functions and exhibited gate-tunable gain modulation.
• Higher gate biases improved CNN classification accuracy but increased energy consumption.
• Lower gate biases reduced energy use, offering an adaptive accuracy-energy tradeoff.

Conclusions:

• The developed device integrates essential synaptic behaviors with gate-controlled gain modulation, emulating neuromodulation.
• This offers a practical and energy-efficient approach for adaptive neuromorphic vision systems.
• The device presents a pathway towards more sophisticated and adaptable in-sensor computing for visual processing.