490.txt

This complete matlab for neural network

发信人: fervvac (高远), 信区: DataMining
标  题: Re: 关于聚类
发信站: 南京大学小百合站 (Mon Apr 22 09:29:09 2002), 站内信件

Can you elaborate on this?

At the first sight, this seems not nature. Most cluster definitions are based
on distance or density in the original space, not on the frequent domain. For
example, two clusters (one dimensional) might be of the same shape, but with 
different offsex in the x-axis, so they only differ in the phases. Moreoever, 
if you have a fairly accurate approximation, you will include high-frequency
component, which in many case is regarded as noise.

Anyway, there may be a solution there.

【 在 ttux (tux) 的大作中提到: 】
: 
: 偶最近有个问题:是否可以通过用数学逼近的方法(如小波等)来描述每一个cluster,
: 从而为进一步的预测提供了可能,不知道这样的做法复杂度在多少阿?请教.


--
※ 来源:．南京大学小百合站 bbs.nju.edu.cn．[FROM: 饮水思源BBS]