How so? We give it different variables and let it change them incrementally. Calmness is a 0.0 - 1.0 floating point output from the band.
This was custom-made by the way.
Depends on what exactly you did, but the SVM doesn't maximize calmness - it finds a maximally separating hyperplane that separates calm data from not-calm data. If you had real-time biofeedback then the person wearing the handband might have used the feedback to help them relax, but in that case they are the one maximizing calmness, not the SVM.
Ok. My only experience with SVMs was reading the wikipedia article a year ago. Our implementation evolved pretty organically and I only thought later that it was like a SVM. thanks for the info.