A multi-instance multi-label learning algorithm based on instance correlations

摘要

Existing multi-instance multi-label learning algorithms generally assume that instances in a bag are independent of each other, which is difficult to be guaranteed in practical applications. A novel multi-instance multi-label learning algorithm is proposed by modeling instance correlations in each bag. First, instance correlations are introduced in multi-instance multi-label learning by constructing graphs. Then, different kernel matrices are derived from kernel functions based on graphs at different scales, which are employed to train Multiple Kernel Support Vector Machine (MKSVM) classifiers. Experimental results on different datasets show that the proposed method significantly improves the accuracy of the multi-label classification compared with the state-of-the-art methods.