当前位置: X-MOL 学术Mater. Today Phys. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Flexible transparent InZnO synapse transistor based on Li1.3Al0.3Ti0.7(PO4)3/polyvinyl pyrrolidone nanocomposites electrolyte film for neuromorphic computing
Materials Today Physics ( IF 11.5 ) Pub Date : 2020-12-01 , DOI: 10.1016/j.mtphys.2020.100264
J. Li , Y.-H. Yang , W.-H. Fu , Q. Chen , D.-L. Jiang , W.-Q. Zhu , J.-H. Zhang

Abstract The hardware implementation of neuromorphic computing has attracted growing interest as a promising candidate for confronting the bottleneck of traditional von Neumann computers. Moreover, flexible artificial synapses with learning capabilities are easier to achieve massive parallelism and structural flexibility of the human brain. Most previous reports have focused on the use of electric or light stimulation methods to simulate synaptic behavior through a single mode. To improve the memory and learning characteristics of synaptic transistors, devices with ion-electron coupled electric-double-layer are frequently used. Here we report an organic composite nanoparticle electrolyte TFT based on an aqueous solution is proposed. Some important synaptic behaviors are successfully simulated in our PVP: LATP TFT artificial synapses, including paired pulse promotion (PPF), paired pulse depression (PPD) and signal filtering characteristics. The PPF and PPD index can be modulated by the spike width and spike interval of the presynaptic pulse and the pulse intensity. The minimum energy consumption of the pulse spike of the PVP: LATP-based synaptic transistor is calculated to be 2.28 pJ. We use the synaptic device to simulate “OR” and “YES” logic. Furthermore, we also conducted the classical conditional Pavlov's dog experiments to mimic the associative memory process of human brain using light and electricity combined stimulation. These results migvht provide an alternative route for neural computing.

中文翻译:

基于Li1.3Al0.3Ti0.7(PO4)3/聚乙烯吡咯烷酮纳米复合电解质膜的柔性透明InZnO突触晶体管用于神经形态计算

摘要 神经拟态计算的硬件实现作为解决传统冯诺依曼计算机瓶颈的一个有前途的候选者,引起了越来越多的兴趣。而且,具有学习能力的灵活人工突触更容易实现人脑的大规模并行和结构灵活性。大多数以前的报告都集中在使用电或光刺激方法通过单一模式来模拟突触行为。为了改善突触晶体管的记忆和学习特性,经常使用具有离子电子耦合双电层的器件。在这里,我们报告了一种基于水溶液的有机复合纳米粒子电解质 TFT。在我们的 PVP 中成功模拟了一些重要的突触行为:LATP TFT 人工突触,包括成对脉冲促进(PPF)、成对脉冲抑制(PPD)和信号滤波特性。PPF 和 PPD 指数可以通过突触前脉冲的尖峰宽度和尖峰间隔以及脉冲强度进行调制。PVP:基于 LATP 的突触晶体管的脉冲尖峰的最小能量消耗计算为 2.28 pJ。我们使用突触设备来模拟“OR”和“YES”逻辑。此外,我们还进行了经典的条件巴甫洛夫狗实验,以利用光和电联合刺激来模拟人脑的联想记忆过程。这些结果可能为神经计算提供了另一种途径。PPF 和 PPD 指数可以通过突触前脉冲的尖峰宽度和尖峰间隔以及脉冲强度进行调制。PVP:基于 LATP 的突触晶体管的脉冲尖峰的最小能量消耗计算为 2.28 pJ。我们使用突触设备来模拟“OR”和“YES”逻辑。此外,我们还进行了经典的条件巴甫洛夫狗实验,以利用光和电联合刺激来模拟人脑的联想记忆过程。这些结果可能为神经计算提供了另一种途径。PPF 和 PPD 指数可以通过突触前脉冲的尖峰宽度和尖峰间隔以及脉冲强度进行调制。PVP:基于 LATP 的突触晶体管的脉冲尖峰的最小能量消耗计算为 2.28 pJ。我们使用突触设备来模拟“OR”和“YES”逻辑。此外,我们还进行了经典的条件巴甫洛夫狗实验,以利用光和电联合刺激来模拟人脑的联想记忆过程。这些结果可能为神经计算提供了另一种途径。我们还进行了经典的条件巴甫洛夫狗实验,利用光和电联合刺激来模拟人脑的联想记忆过程。这些结果可能为神经计算提供了另一种途径。我们还进行了经典的条件巴甫洛夫狗实验,利用光和电联合刺激来模拟人脑的联想记忆过程。这些结果可能为神经计算提供了另一种途径。
更新日期:2020-12-01
down
wechat
bug