The performance of machine learning algorithms is heavily dependent on the quality of the data representation. Learning discriminative data representations becomes a critical step in the machine learning task pipeline. Data representations can be obtained from multiple views. For example, in a complex network model, the representation of each node can be viewed as a combination of the correlations between itself and its neighbors and the node attributes. A comprehensive representation can be obtained because data from different views usually contain complementary information.This dissertation tackles the problem of multi-view representation learning. How to generate the data representations by incorporating multiple aspects of data, such as text, network topological information, and data derived from various sources. Multi-view representations are capable of uncovering the implicit and complex explanatory factors of variation behind the data.