Key Instance Detection in Multi-Instance Learning

Abstract

The goal of traditional multi-instance learning (MIL) is to predict the labels of the bags,
\nwhereas in many real applications, it is desirable to get the instance labels, especially the
\nlabels of key instances that trigger the bag labels, in addition to getting bag labels. Such a
\nproblem has been largely unexplored before. In this paper, we formulate the Key Instance
\nDetection (KID) problem, and propose a voting framework (VF) solution to KID. The key
\nof VF is to exploit the relationship among instances, represented by a citer kNN graph. This
\ngraph is different from commonly used nearest neighbor graphs, but is suitable for KID.
\nExperiments validate the effectiveness of VF for KID. Additionally, VF also outperforms state-of-the-art MIL approaches on the performance of bag label prediction.