bci2000.org BBS

Where can I find the detailed tutorial about the adaptation algorithm in STATISTIC?
I really want to fully understand how the Mean and Gain are adapted online in order to better configure my system for mu rhythm application.
Thanks for help!

Shixiaoyue,

Unfortunately, we currently do not have a good documentation for this capacity. This is a prime target for the ongoing system improvements.

At the same time, it is a good start if you use the D2Box batch file and parameter file, and then load the appropriate 1D or 2D parameter file fragments on top of them. Parameters will be set appropriately if you do that.

In addition, The BaselineCfg parameter defines which periods are used to adapt, e.g.

TargetCode 1 Feedback 1
TargetCode 2 Feedback 1

Would set the intercept to the mean of the signal values when the target is on top and the cursor is moving (first line) and when the target is on the bottom and the cursor is moving (2nd line).

The gain is updated so as to produce an output range that is proportional to X/YPixelsPerSecond.

Please do not hesitate to ask additional questions on this subject. Also, there have been similar questions on the board that may answer some of your questions.

Gerv[/code]

Hi,Gerv:
Thank you very much for your help! And now I can fully understand how the Mean and Gain are adapted online when Y(X)TrendControl=1. I have also noticed that when Y(X)TrendControl=1, TargetWinLth, BaselineHits, TrendControlRate, OutcomeDirection and Y(X)MeanProportion will not be used. Then, I have another question, when should I set Y(X)TrendControl to 2 or 3? How does StatFilter modify the algorithm when Y(X)TrendControl = 2 or 3? In the D2Box parameter file, Y(X)TrendControl is also set to 1, does that mean it is enough for ordinary use of D2Box?
Best wishes!

Shixiaoyue,

At present, I would not suggest to set X/YTrendControl to 2 or 3. This requires complex and precise adjustments to make this work correctly. As I mentioned, we are currently working hard to improve both the adaptation framework and also its documentation.

The Gerv