A survey on canonical correlation analysis based multi-view learning

Abstract

Multi-view learning (MVL) is a strategy for fusing data from different sources or subsets.Canonical correlation analysis (CCA) is very important in MVL, whose main idea is to maximize the correlation of different views.The traditional CCA can only calculate the linear correlation between two views.Moreover, it is unsupervised, and the label information is wasted in supervised learning tasks.Many nonlinear, supervised, or generalized extensions have been proposed to accommodate these limitations.Firstly, a comprehensive overview of representative CCA approaches was provided.Then their classical applications in pattern recognition, cross-modal retrieval and classification, and multi-view embedding were described.Finally, the challenges and future research directions of CCA-based MVL approaches were pointed out.

Keywords

• canonical correlation analysis;multi-view learning;multi-modal retrieval;multi-view embedding