How to find an optimal of Gaussian mixture component proportion number?

5 次查看（过去 30 天）
Pradya Panyainkaew
Pradya Panyainkaew 2018-4-13
Hi everyone
I have a weekday data that I attached file in this question. I try to find an optimal number of Gaussian mixture component proportions by using the code below.
AIC = zeros(1,10);
GMModels = cell(1,10);
for k = 1:10
GMModels{k} = fitgmdist(weekday_data,k,'Options',statset('Display','final','MaxIter',1500,'TolFun',1e-5));
AIC(k)= GMModels{k}.AIC;
end
[minAIC,numComponents] = min(AIC);
numComponents
BestModel = GMModels{numComponents}
Unfortunately, the Matlab software shows the error like this :
"Ill-conditioned covariance created at iteration 2."
and others descpriptions :
"Error in gmdistribution.fit (line 98) gmcluster(X,k,start,reps,CovType,SharedCov,RegV,options,probtol);
Error in fitgmdist (line 130) gm = gmdistribution.fit(X,k,varargin{:});
Error in test (line 58) GMModels{k} = fitgmdist(weekday_data,k,'Options',statset('Display','final','MaxIter',1500,'TolFun',1e-5));" I don't know how to fix these error.
Are there anyone can solve this problem ?
Thank you very much.

请先登录，再回答此问题。

回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Gaussian Mixture Models 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by