Learning From Weakly Supervised Data by The Expectation Loss SVM (e-SVM) algorithm

In many situations we have some measurement of confidence on positiveness is a continuous value whose range is a bounded interval.