Learnable Graph Guided Deep Multi-View Representation Learning via Information Bottleneck

In real world applications, multi-view data has attracted intensive attention due to the complex and complementary relationship across views. Multi-view representation learning (MvRL) focuses on obtaining consistent feature representation from multi-view data, and becomes a popular topic in multi-view research field. However, the relationship between different samples, i.e., the graph information, is usually ignored or excavated insufficiently in most existing MvRL methods, which only regard graph structure as regularization items instead of graph embedding for multi-view data. Besides, the limited learning capacity of the adopted shallow models is another challenge for MvRL. To tackle them, in this paper, we propose a novel unsupervised deep multi-view representation learning model guided by learnable graph structure, termed as LGG-DMRL. It first captures a multi-view consistent graph from original data based on self-representation learning, and explores the view-specific feature representation of each view by the designed graph guided attention network using the learnt graph. After that, the information bottleneck principle is employed to identify the shared representation across views integrated with the view-specific feature representations, promoting the multi-view complementarity and completeness. Experimental results on five real-world datasets demonstrate the superiority and effectiveness of our proposed LGG-DMRL compared with the recent state-of-the-art multi-view approaches.

I. Introduction

Data in real world applications naturally contain complex interactions and are described with multiple modalities or types of features, which can be considered as multiple views. Most multi-view methods aim at exploring the common hidden structure of cross-view data to better integrate multiple features. Over the past decades, a great number of methods have been proposed and achieved promising results. Some researchers focus on consistent graph learning [1], [2], [3], [4], [5], [6], [7], [8], [9] or consistency and specificity from the data distribution [10], [11], [12], [13] for multi-view clustering or classification. And some other methods projects different views into one common space and learns the latent representation [14], [15], [16], i.e., the multi-view representation learning (MvRL) methods. MvRL methods have the higher generalization capability since the learnt representation can be used for downstream tasks, including clustering and classification.