The GNG is compared to other self-organising networks such as Kohonen and Neural Gas (NG) maps and results are given for the training set of hand outlines, showing that the proposed method preserves accurate models. 展开 关键词: image segmentation classification growing neural gas registration ...
然后这16个值就可以用4位的二进制码来编码表示(2^4=16)。因此,这是个2-dimensional, 4-bit VQ,它的速率同样是2bits/dimension。上面这些红星点就是量化矢量,表示图中的任意一个点都可以量化为这16个矢量中的其中一个。 设计问题(写出优化目标) 优化准则 LBG算法(迭代) 论文阅读(A Growing Neural Gas Netwo...
library(GrowingNeuralGas)#Load datadata(wine,package="rattle")scaled_wine<-scale(wine[-1])#Train in an offline mannergng<-GNG(scaled_wine,labels=wine$Type,max.nodes=20)#Find closest node to vector [1,...,1]vct=rep(1,13) predict(gng,vct)#Convert to igraph and print mean degree of...