bci2000.org BBS

Hi there,
I am trying to setup a 1D cursor task experiment in BCI2000 V2, but the cursor moves very slowly, how I can speed the cursor movement. I know in BCI2000 V1 there was the statistic tab where some parameters can be set to control the cursor movement.

Another question is how I can interpret the weight parameter in the MLR matrix of the Linear Classifier, what affect does the weight value has on the work of the classifier. and how I can set the adaptation parameter for the Normalizer in the filter tab. Any hints appreciated.

I appreciate your help,

Omar AL-Zoubi

Omar,

we provide a list of parameter changes from version 1 to version 2 at
http://www.bci2000.org/wiki/index.php/U ... _Version_2

The parameter that governs cursor speed is now called FeedbackDuration, and allows to specify the desired mean duration of feedback in seconds.
For more details, see http://www.bci2000.org/wiki/index.php/U ... ckDuration

Normalizer behavior is now controlled by the Adaptation parameter on the Filtering tab. Please see http://www.bci2000.org/wiki/index.php/U ... Normalizer for more details.

The weight value in the LinearClassifier matrix is just a multiplicative factor that enters into computation of the feedback signal. Its sign determines the orientation of the cursor response (i.e. whether it goes up or down when the brain signal amplitude increases). For a single entry in the LinearClassifier matrix, its absolute value is irrelevant when the Normalizer filter is used to adapt signal gain; when there are multiple entries in the LinearClassifier matrix, their relative absolute values determine the relative weight of each feature in the feedback signal.

For manual configuration, you will typically set the weight value to +1 or -1 for the features of interest. When using a training algorithm to optimize classification, the algorithm will produce an individual weight value for each input feature.

Regards,
Juergen

Thanks Juergen,
this realy clarify many things to go with my experemnt for the moment.

omar,