基于仿生复眼的自生长人工神经网络

摘要

为探索具有特定功能的仿生式可生长型人工神经网络,提出一套仿生式、可自生长自组织神经网络的生成算法.基于自生成网络的思想设计一套神经网络模拟器,采用GPU进行加速仿真,使得每个神经元能进行并行运算,针对果蝇的复眼神经网络的视觉处理过程进行解析,抽象提取出一个五层的网络结构,该仿生昆虫视觉神经网络具有对物体距离、运动方向判断的功能,在模拟器中完成对仿生果蝇视觉的功能验证.下一步,将在该网路基础上,实现自生成算法,完成对网络结构的优化.%To explore the bionic artificial neural network with specific functions,an algorithm to build self-growing and self-organizing bionic neuron network was presented.A neural network simulator was designed based on the idea of self-generated network Graphics processing unit was used to accelerate the simulation,so that each neuron performed parallel operation.The visual processing of the compound eye network of drosophila melanogaster was analyzed,and a five layer network structure was extracted,which had the function of judging object's distance and moving direction.The next step,a self-generation algorithm to complete the optimization of the network structure will be achieved on the basic neural network.