谷歌浏览器插件
订阅小程序
在清言上使用

1-Phototransistor-1-Threshold Switching Optoelectronic Neuron for In-Sensor Compression via Spiking Neuron Network

2023 International Electron Devices Meeting (IEDM)(2023)

引用 0|浏览12
暂无评分
摘要
In this work, we demonstrate a 1-Phototransistor-1-Threshold Switching (1PT1TS) optoelectronic neuron with an In 2 O 3 PT and NbO x TS device for in-sensor compression paradigm, which is achieved by exploiting continuous-time data representation and frequency multiplexing—without external electrical circuits nor ADCs. The In 2 O3 PT and NbO x TS devices exhibit high photoresponsivity (10 4 mA/W) and endurance (10 8 cycles), respectively. Our approach enables a single channel to perceive sensory information from multiple inputs in parallel with read errors of only 2.2%. Compared to the conventional sensory processing framework with separated sensing and compression modules, our 1PT1TS optoelectronic neuron arrays can directly sense, compress, encode, and interpret sensory data (sampling rate <20%) with ultra-low energy consumption (0.1 nJ), high recognition accuracy (>80%), and nearly zero accuracy loss (~0%) based on the spiking neuron network.
更多
查看译文
关键词
Spiking Neural Networks,Spiking Neuronal Networks,Sensory Information,Sensory Processing,Recognition Accuracy,High Recognition Accuracy,Parallel Input,High Photoresponsivity,Visual Information,Processing Unit,Error Distribution,Edge Computing,Reconstruction Results,Sparse Sampling,Speed Distribution,Signal Compression,Spike Signals,Phototransistor,Extra Unit
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要