论文部分内容阅读
忆阻器是一种具有天然信息存储能力和纳米级尺寸的类突触元件,由忆阻器构成的桥式突触电路结构简单、控制精确,有望应用于人工神经网络、超大规模集成电路、图像处理、模式识别等领域,然而传统的忆阻桥突触电路在突触模拟的过程中,由于单极性脉冲造成的通量积累会对权值模拟产生一定的误差.本文提出了一种新型的忆阻突触桥电路,该电路有效克服了传统桥电路忆阻值漂移的问题,由该电路构成的新型忆阻神经网络应用于图像处理中有着更为明显的优势,文中通过仿真实验验证了该方案的可行性,说明基于双极性脉冲发生器忆阻桥神经网络具有更加高效的仿生特性,而且集成度更高、模板更易更换,有望更有效解决实时的复杂的智能问题.
Memristor is a kind of synaptic device with natural information storage capacity and nano-scale size. The bridge synaptic circuit composed of memristor has the advantages of simple structure and precise control and is expected to be applied to artificial neural networks, VLSI circuits, Image processing and pattern recognition. However, in the process of synaptic simulation, the conventional memristor synaptic circuit has some errors due to the accumulation of flux caused by unipolar pulses, The new memristor synaptic bridge circuit, the circuit effectively overcome the traditional bridge circuit memristor drift value of the problem, the circuit consisting of a new memristor neural network used in image processing has a more obvious advantages, the paper through the simulation experiment The feasibility of this scheme is verified. It shows that neural network based on bipolar pulser bridge with memristor bridge has more efficient bionics, more integration and easier template replacement, which is expected to solve the real-time complex and intelligent problem more effectively.