How much time for training the classifier

Working with OpenViBE signal processing scenarios and doing scenario/BCI design
Post Reply
rencarcos
Posts: 4
Joined: Tue Oct 25, 2011 2:24 pm

How much time for training the classifier

Post by rencarcos »

Hi everyone!
I'm using the motor-imagery scenarios with emotiv epoc. In the classifier, I changed to select F3 and F4 channels, no reference channel since emotiv doesn't need. But I think i'm having problems with the classifier trainer, how long it takes to finish?
Using fastforward it's running for 95h and didn't ended yet

In the lua script i found box:send_stimulation(1, OVTK_StimulationId_ExperimentStop, t, 0) but in the scenario the trigger of the classifier trainer is OVTK_GDF_End_Of_Session. Should I changed it to OVTK_StimulationId_ExperimentStop?

Thank you very much
RenCarrasco

ddvlamin
Posts: 160
Joined: Thu Aug 13, 2009 8:39 am
Location: Ghent University
Contact:

Re: How much time for training the classifier

Post by ddvlamin »

Hi,

Which scenario are you using and did you change something?

Could you also post the log file of the designer (you can find this in the log directory)?

Training the classifier should go quite fast (at least the linear discriminant analysis) and it will output lines in the log or the console such as the following:

Code: Select all

[  INF  ] At time 2280124317696 (0x212e2000000) <Box algorithm::Classifier trainer> Received train stimulation
[  INF  ] At time 2280124317696 (0x212e2000000) <Box algorithm::Classifier trainer> Classifier performance on whole set is 97.469388% (trained without k-fold test - performance estimation may not be accurate)
[  INF  ] Importing scenario...
My scenario also keeps running so I have to press stop after I see these lines in the output. However, it could be that in the new scenarios it is stopped automatically after it has been finished?

Indeed, you could also quickly try other stimulation ends in the meantime, I use the gdf reader and consequently OVTK_GDF_End_Of_Session as the training trigger. You could also try OVTK_StimulationId_EndOfFile?

Best regards,
Dieter Devlaminck

rencarcos
Posts: 4
Joined: Tue Oct 25, 2011 2:24 pm

Re: How much time for training the classifier

Post by rencarcos »

now i tried to use OVTK_StimulationId_ExperimentStop and it took just some minutes to finish
looks like it worked

but I was using the bci/motor-imagery/motor-imagery-bci-2-classifier-training.xml
just removed the reference channel box and removed the surface laplacian box, and changed the channel selector to F3;F4

thanks!
RenCarrasco

Post Reply