Related Experiment Video
Updated: May 1, 2026
08:48
Optical Recording of Suprathreshold Neural Activity with Single-cell and Single-spike Resolution
Published on: September 5, 2012
11.3K
Enhanced In-Sensor Computing with Spike Number-Dependent Plasticity Characteristics in an InGaSnO Optical
Min Ho Park1,2, Yeojin Kim3, Min Jung Choi1,2
1Department of Advanced Materials Engineering for Information and Electronics, Kyung Hee University, Yongin 17104, Republic of Korea.
ACS Nano
|March 27, 2025
Summary
This study introduces indium gallium tin oxide (IGTO) to improve optical neuromorphic devices. The new material enhances spike number-dependent plasticity (SNDP), boosting machine vision efficiency and accuracy.
Area of Science:
- Materials Science
- Neuromorphic Engineering
- Computer Vision
Background:
- Optical neuromorphic devices are key for efficient machine vision.
- Spike timing-dependent plasticity (STDP) in current devices complicates in-sensor computing and reduces accuracy.
Purpose of the Study:
- To develop an optical neuromorphic device with enhanced spike number-dependent plasticity (SNDP) and eliminate STDP.
- To investigate the potential of indium gallium tin oxide (IGTO) for this application.
Main Methods:
- Fabrication of an optical neuromorphic device using IGTO semiconductor.
- Characterization of synaptic plasticity using light pulse stimuli.
- Verification of material properties using photoemission spectroscopy (PES).
- Evaluation of in-sensor computing performance in a multilayer perceptron (MLP) model.
Main Results:
- The IGTO-based device demonstrated robust SNDP, independent of light pulse timing.
- Device conduction state reliably correlated with the number of input light pulses.
- PES analysis confirmed strong Sn-O bonding contributing to SNDP.
- In-sensor computing with the SNDP-enhanced device reduced MLP training time by 87.7% with high classification accuracy.
Conclusions:
- IGTO-based optical neuromorphic devices exhibit significant SNDP, overcoming STDP limitations.
- These devices offer a promising pathway for accelerating machine learning in highly efficient machine vision systems.
