A new approach for multiple instance learning based on a homogeneity bag operator

Multiple Instance Learning (MIL) proposes a new paradigm when instance labeling, in the learning step, is not possible or infeasible, by assigning a single label (positive or negative) to a set of instances called bag. In this paper, an operator based on homogeneity of positive bags for MIL is introduced. Our method consists in removing instances from the positives bags according to their similarity with the ones from the negative bags. The experimental results show that our operator always increases the accuracy of the Citation kNN algorithm achieving the best results in 2 out of 4 datasets when compared with other classic methods in the literature.