Hello,
I want to compare the accuracy that I get from motor imagery scenarios with different algorithms.
When I use SVM as my classification algorithm it crushes (terminates and closes Openvibe). I read here in the forum that if I use version 2.2.0 of the "Classifier trainer" box it will work.
My question is how to change the version of the specific box only?
Also, when I use MLP as my classification algorithm, I get errors. (please see attached image here: https://prnt.sc/LCRILxaOdPdq)
What am I doing wrong?
Thanks
SVM - crushes, MLP - gets default values
Re: SVM - crushes, MLP - gets default values
Hi,
It is not possible to change the version of a specific box only. If you want to use the SVM classification, you need to download OpenViBE 2.2.0, and use the boxes that come with it.
The feature will be fixed in the next version of OpenViBE which should be released by end of April.
In the meantime, you can have access to it by compiling the development branch of OpenViBE on your machine.
Here is how to get the code from gitlab: http://openvibe.inria.fr/repo-instructions/
Make sure to replace the clone command by :
Once you have the code, here are the build instructions: http://openvibe.inria.fr/build-instructions/
As for the MLP, can you confirm that the three parameter "Number of neurons", "Learning stop condition" and "Learning coefficient" have positive values ?
Hope that all helps.
Cheers,
Thomas
It is not possible to change the version of a specific box only. If you want to use the SVM classification, you need to download OpenViBE 2.2.0, and use the boxes that come with it.
The feature will be fixed in the next version of OpenViBE which should be released by end of April.
In the meantime, you can have access to it by compiling the development branch of OpenViBE on your machine.
Here is how to get the code from gitlab: http://openvibe.inria.fr/repo-instructions/
Make sure to replace the clone command by :
Code: Select all
git clone -b development https://gitlab.inria.fr/openvibe/meta.gitAs for the MLP, can you confirm that the three parameter "Number of neurons", "Learning stop condition" and "Learning coefficient" have positive values ?
Hope that all helps.
Cheers,
Thomas
Re: SVM - crushes, MLP - gets default values
Thanks for your response Thomas.
I am using version 3.1.0 of Openvibe.
I am planning to download version 2.2.0 in order to get the results for SVM. Do you know if this problem -not showing these parameters of MLP- is present in version 2.2.0 as well?
I would appreciate your help,
Sofia
In the configurations of the Classifier Trainer box, there are not these parameters. Please find a screenshot here https://prnt.sc/rtNPVIYY3e05 .As for the MLP, can you confirm that the three parameter "Number of neurons", "Learning stop condition" and "Learning coefficient" have positive values ?
I am using version 3.1.0 of Openvibe.
I am planning to download version 2.2.0 in order to get the results for SVM. Do you know if this problem -not showing these parameters of MLP- is present in version 2.2.0 as well?
I would appreciate your help,
Sofia
Re: SVM - crushes, MLP - gets default values
Hi,
The hack to use OV 2.2.0 is usefull to set up your box. After box is setting up you can go back to 3.1.0. You see the option change when you change classifier (in 2.2.0) the update fail in 3.1.0 it's why it's crash. Burt if settings are update (by previous version of OV) it's good.
Thibaut
The hack to use OV 2.2.0 is usefull to set up your box. After box is setting up you can go back to 3.1.0. You see the option change when you change classifier (in 2.2.0) the update fail in 3.1.0 it's why it's crash. Burt if settings are update (by previous version of OV) it's good.
Thibaut