Having:
- set of soft fuzzy classifiers (classification onto overlapping sets) C_i(x) -> [0,1]
- corresponding set of weak estimators R_i(z) of the form R_i(z) = EX(y|z).
The estimators R_i are just some kind of regression, kalman or particle filters. The classifiers C_i are fixed and static.
How to make a strong estimator out of a weighted combination of the form L(x, z) = SUM_i: C_i(x)R_i(z)Q_i ?
In other words how to choose the weights Q_i?
Is there some kind of online approach to this problem?
[link][comment]