摘 要：摘 要：该文介绍一种可以不断升级神经环路的规模的设计原则。作者将单个视网膜神经节细胞的树突区域定义为类似于数码相机CCD上的一个“像素”。决定这个像素的大小（即树突区域所占的大小）存在两种相互竞争的因素：为得到最高的分辨率，像素应越小越好；另一方面，像素越大越利于平均化输入信号，以得到信噪比精确度最高的光强度值递呈给大脑。作者列举了三种可能的策略来阐释如何设计一个视网膜像素大小，并且为进化中实际选择的策略提供了证据。
关键词：神经环路；视网膜神经节细胞；像素
中图分类号：Q436; R339.146; 文献标识码：A
Abstract: Abstract: I will describe one of the design principles that govern circuits that are scalable in this way. The dendritic arbor of a RGC, then, defines a pixel much like a "smart"pixel in a digital camera. The pixel size (territory size of the dendritic arbor) is subject to competing demands: the pixel should be as small as possible to give the best resolution, and as large as possible to average over the most inputs and derive the most accurate estimate of the light intensity it reports to the brain. I formulate three possible strategies for dealing with the problem of how large to make a retinal pixel, and present evidence for the one that evolution has selected.
Key words: neural circuits; RGC; pixel size
