Supervised adapted factor analysis algorithm for multimodal data classification

In recent years, multimodal data processing has enjoyed an increasing attention. Multimodal document means different modal multimedia data representing the same semantics. In this paper, the problem of multimodal document classification is studied. As the text features have more obvious advantages than the image features over the classification problem, on the basis of factor analysis, the paper proposes a supervised algorithm which projects image space into text space and then learns a linear classifier for classification in the text space. Encouraging experiment results on three benchmark datasets demonstrate the superiority and effectiveness of proposed methods over most existing algorithms.