classification with unequal weights Options

[Tsur]

Hello,
The problem i am trying to solve is as follows:

two data sets one labels 'wins' and another labeled 'loses' , each point is a vector a1,a2,...,an
however the cost of misclassifying a 'lose' is constant 1, while the cost of misclassifying a win depends on parameter a1.

It seems to me that a modified version of knn can solve this problem, however i am interested to do it using a SVM with a custom performance function.
is this possible in synapse?

I have been following the development of Synapse for some time and was waiting for a problem to emerge which suits your solution