Home fitgmdist gives to big sigma - matlab
Reply: 0

fitgmdist gives to big sigma - matlab

Sarah Lockenmarie
1#
Sarah Lockenmarie Published in 2017-12-07 18:23:24Z

I have a lot of dot areas in my data file. And I want to fit a GMM but everytime I try to calculate my mu and sigma, sigma is far to big. The strange thing however is that mu fits pretty good. By multiplying my data with a scalfactor of 10000... the fit seems to discribe the data better, at least the red line is visible. But why?? Can I somehow minimize my sigma without using an scale factor?

enter code here

step = 1;
bin_centers = 25000:2000:47500;
data = Einzelbilder.TL1(1).area;
data = data(data > 30000);
counts = hist(data,bin_centers);
bar(bin_centers,counts);
hold on;

GMM = fitgmdist(data,1);%,'Regularizationvalue',0.1

y1 = pdf('Normal',bin_centers,GMM.mu(1),GMM.Sigma(1))*GMM.ComponentProportion(1)*numel(data)*step;
plot(bin_centers,y1,'-r');
hold on



data =

   1.0e+04 *

    4.2300
    3.0758
    3.4027
    3.9044
    4.1913
    3.8030
    3.2079
    4.5503
    4.3902
    3.4120
    4.3621
    3.0451
    3.3960
    3.4814
    4.2420
    4.0312
    3.0184
    3.8150
    3.4227
    3.0264
    3.3707
    3.7176
    3.1999
    3.9685
    4.2554
    4.2300
    3.0691
    3.7163
    3.7203
    4.5810
    4.2794
    4.5570
    3.1625
    3.4454
    3.7603
    3.9151
    4.0125
    3.2893
    3.0691
    3.6549
    3.4054
    3.1065
    3.4761
    4.5449
    3.0438
    3.1452
    4.3568
    3.1265
    3.0451
    3.8564
    4.3741
    3.0424
    3.0264
    3.5748
    3.3066
    3.3000
    3.5241
You need to login account before you can post.

About| Privacy statement| Terms of Service| Advertising| Contact us| Help| Sitemap|
Processed in 0.333425 second(s) , Gzip On .

© 2016 Powered by mzan.com design MATCHINFO