Adaptive p-Posterior Mixture-Model Kernels for Multiple Instance Learning

In multiple instance learning (MIL), how the instances determine the bag-labels is an essential issue, both algorithmically and intrinsically. In this paper, we show that the mechanism of how the inst