Multi-View Representation Learning

Abstract

Representation of multiple views on machine learning and data mining are two fields where learning is a very fast-growing direction. The data for multi-view representation learning is gathered. In a single object, there are several views of data in detail. During the testing process, only one view of the data will be used. In this chapter, multi-view representation alignment and fusion are the methods of multi-view representation learning. As a result, multi-view representation learning utilizes a correlation-based alignment perspective. As compared to other representation learning algorithms, canonical correlation analysis (CCA) is one of the most successful. From the standpoint of representation fusion, the progression of multi-view representation learning, which involves generative methods such as multimodal topic learning, is discussed. Multi-view convolutional neural networks and multimodal recurrent neural networks, as well as multi-view sparse coding and multi-view latent space Markov networks, are neural network-based approaches. The theoretical foundations and current developments in the field of multi-view representation learning were addressed in this chapter, as well as a number of important applications of multi-view representation learning.