How do I find cluster center of each cluster gained from Spectral Clustering?

Question

Riley 2021-3-29

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/786286-how-do-i-find-cluster-center-of-each-cluster-gained-from-spectral-clustering

评论： Megh 2022-8-2

Hi, I am using Spectral Clustering for data analysis recently and I found that Spectral Clustering function built in MATLAB only returns the cluster indicator (that is, "idx"). However, I actually need the cluster center of each cluster (similar to the centroid locations gained by k-means). How can I get it? Any idea or code would be very helpful.

Regards, Ning

2 个评论
显示无隐藏无

Image Analyst 2021-3-29

Attach your data and what you've tried so far. Attach screenshots.

Riley 2021-3-29

在 MATLAB Online 中打开

xi.mat

clear
clc
load('xi.mat');
K=100;
[idx]=spectralcluster(xi, K);

Thank you for reminding. This is code and data xi. 'idx' contains cluster indices of each sampling in xi.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

KSSV 2021-3-29

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/786286-how-do-i-find-cluster-center-of-each-cluster-gained-from-spectral-clustering#answer_661036

在 MATLAB Online 中打开

As you have indices in hand, you can get the centers....

clear
clc
load('xi.mat');
K=100;
[idx]=spectralcluster(xi, K);
C = zeros(k,2) ; 
for i = 1:k
    c = xi(idx==i) ;  % this will give cluster 
    C(i,:) = mean(c) ;    % calculate the mean 
end

2 个评论
显示无隐藏无

Riley 2021-3-29

Thanks for your code and idea. It's not what I exactly want but inspiring me!

Megh 2022-8-2

I dont think it makes sense to take the mean since the mean is specifially optimized to minimize the euclidean distance from all the points in the given cluster to the cluster center, which turns out to be the mean after differentiating the MSE loss function and setting it to zero, as in k-means.

请先登录，再进行评论。