Survey on Deep Multi-modal Data Analytics: Collaboration, Rivalry, and Fusion

With the development of web technology, multi-modal or multi-view data has surged as a major stream for big data, where each modal/view encodes individual property objects. Often, different modalities are complementary to other. This fact motivated lot research attention on fusing feature spaces comprehensively characterize Most existing state-of-the-arts focused how fuse energy information from deliver superior performance over their counterparts with single modal. Recently, deep neural networks have been exhibited powerful architecture well capture nonlinear distribution high-dimensional multimedia so naturally does data. Substantial empirical studies carried out demonstrate its advantages that benefited methods, which can essentially deepen fusion spaces. In this article, we provide substantial overview in field analytics shallow Throughout survey, further indicate critical components go collaboration, adversarial competition, and Finally, share our viewpoints regarding some future directions field.