《2016 Multi-instance multi-label learning with application to scene classification》.pdf

Multi-Instance Multi-Label Learning with Application to Scene Classification Zhi-Hua Zhou Min-Ling Zhang National Laboratory for Novel Software Technology Nanjing University, Nanjing 210093, China {zhouzh,zhangml}@ Abstract In this paper, we formalize multi-instance multi-label learning, where each train- ing example is associated with not only multiple instances but also multiple class labels. Such a problem can occur in many real-world tasks, e.g. an image usually contains multiple patches each of which can be described by a feature vector, and the image can belong to multiple categories since its semantics can be recognized in different ways. We analyze the relationship between multi-instance multi-label learning and the learning frameworks of traditional supervised learning, multi- instance learning and multi-label learning. Then, we propose the MIMLBOOST and MIML SVM algorithms which achieve good performance in an application to scene classification. 1 Introduction In traditional supervised learning, an object is represented by an instance (or feature vector) and associated with a class label. Formally, let X denote the instance space (or feature space) and Y the set of class labels. Then the task is to learn a function f : X → Y from a given data set {(x , y ), (x , y ), · · · , (x , y )}, where x ∈ X is an instance and y ∈ Y the known label of x . 1 1 2 2 m m i i i Although the above formalization is prevailing and successful, there are many real-world problems which do not fit this