Solving the multiple instance problem with axis-parallel rectangles
Artificial Intelligence
A framework for multiple-instance learning
NIPS '97 Proceedings of the 1997 conference on Advances in neural information processing systems 10
Content-Based Image Retrieval Using Multiple-Instance Learning
ICML '02 Proceedings of the Nineteenth International Conference on Machine Learning
Solving the Multiple-Instance Problem: A Lazy Learning Approach
ICML '00 Proceedings of the Seventeenth International Conference on Machine Learning
Image Categorization by Learning and Reasoning with Regions
The Journal of Machine Learning Research
MILES: Multiple-Instance Learning via Embedded Instance Selection
IEEE Transactions on Pattern Analysis and Machine Intelligence
Localized Content-Based Image Retrieval
IEEE Transactions on Pattern Analysis and Machine Intelligence
MILD: Multiple-Instance Learning via Disambiguation
IEEE Transactions on Knowledge and Data Engineering
MILIS: Multiple Instance Learning with Instance Selection
IEEE Transactions on Pattern Analysis and Machine Intelligence
Robust Object Tracking with Online Multiple Instance Learning
IEEE Transactions on Pattern Analysis and Machine Intelligence
Multiple-instance learning with instance selection via dominant sets
SIMBAD'11 Proceedings of the First international conference on Similarity-based pattern recognition
| Hi-index | 0.00 |
Multiple-instance learning (MIL) is a variant of traditional supervised learning, where training examples are bags of instances. In this learning framework, only the labels of bags are known while the labels of instances in bags are unknown. This ambiguity in labels of instances leads to significant challenges in MIL. In this paper, we propose an efficient instance selection method to solve this problem, called Salient Instance Selection for Multiple-Instance Learning (MILSIS). MILSIS has two roles: first, selecting discriminative instances and eliminating redundant or irrelevant instances from each bag; second, selecting an instance prototype from each positive bag to construct an embedding space in order to convert the MIL problem to the standard single instance learning problem. Accordingly, based on the first role, we present two novel MIL methods, called MILSIS-kNN-C and MILSIS-kNN-B; based on the second role, we present another new MIL method, called MILSIS-SVM. Experimental results on some synthetic and benchmark data-sets demonstrate the effectiveness of our methods as compared to others.