A Two-Phase Method for Solving the Multiple Instance Learning Problem
Abstract
The multiple instance learning problem is getting more attention recently in the field of machine learning. In multiple instance learning, the training set comprises labeled bags that are composed of unlabeled instances, and the task is to predict the labels of unseen bags through analyzing the training bags with known labels. In this paper, we proposed a two-phase method for solving the multiple instance learning problem. Two phases of our method are independent of each other and therefore the proposed method has a high flexibility for future optimization and extension. The proposed method can discover the specific concept pattern using only a few training samples. This method presents the learned concept as a hypercube in an n dimensional feature space. We compare our proposed method with other existing algorithms using the Musk data, which is a popularly used real-world benchmark for multiple instance learners. The result of experiments shows the validity of the proposed method. Moreover, we provide the experimental results on image classification problem. Accuracy is evaluated and the effectiveness of the proposed method has been shown through comparative studies.
Keywords
Machine Learning, Multiple Instance Learning, Training Examples, Feature Space
