Linear kernel only one that does not return poor classification

asked 2018-04-06 15:04:05 -0500

LeBorzi gravatar image

updated 2018-04-07 02:44:16 -0500

After trying different SVM kernels in OpenCV, I realised that whenever I switched from a linear kernel to another, the classifier would stop functioning properly and classify all data as the same label. I currently have six different labels and am trying to classify features from three dimensional arrays. Is there anything I'm missing? Thank you in advance!

edit retag flag offensive close merge delete

Comments

@LeBorzi please remove all others posts

LBerger gravatar imageLBerger ( 2018-04-06 15:07:41 -0500 )edit
2

Sorry, I did not realise it posted this many times...

LeBorzi gravatar imageLeBorzi ( 2018-04-06 15:10:51 -0500 )edit

If data are easy to classify then switched method will give same results. When method is switched you train model then you can save model using FileStorage and compare models

LBerger gravatar imageLBerger ( 2018-04-07 03:19:35 -0500 )edit

How come that it seems like switching kernels seems to break the classifier though? Surely, switching from linear to another kernel doesn't explain why the classifier starts to make the same label prediction every time. Keep in mind that I am redoing training from scratch with the new kernel, so I am not training the SVM with a linear kernel, then switching to another.

LeBorzi gravatar imageLeBorzi ( 2018-04-07 14:14:55 -0500 )edit