Multiple instance learning

http://www.cs.cmu.edu/~juny/MILL/review.htm

1. Introduction

Multiple Instance Learning (MIL) is proposed as a variation of supervised learning for problems with incomplete knowledge about labels of training examples. In supervised learning, every training instance is assigned with a discrete or real-valued label. In comparison, in MIL the labels are only assigned to bags of instances. In the binary case, a bag is labeled positive if at least one instance in that bag is positive, and the bag is labeled negative if all the instances in it are negative. There are no labels on the individual instances. The goal of MIL is to classify unseen bags or instances based on the labeled bags as the training data.

            The study on MIL was first motivated by the problem of predicting the drug molecule activity level. After that, many MIL methods have been proposed, such as learning axis-parallel concepts [Dietterich et al., 1997], diverse density [Maron and Lozano-Perez, 1998], extended Citation kNN [Wang and Zucker, 2000], etc. They have been applied to a wide spectrum of applications ranging from image concept learning and text categorization to stock market prediction. We review the several popular MIL methods and their applications in this document.