Multiview machine learning

During the past two decades, multiview learning as an emerging direction in machine learning became a prevailing research topic in artificial intelligence (AI). Its success and popularity were largely motivated by the fact that real-world applications generate various data as different views while people try to manipulate and integrate those data for performance improvements. In the data era, this situation will continue. We think the multiview learning research will be active for a long time, and further development and in-depth studies are needed to make it more effective and practical. In 2013, a review paper of mine, entitled “A Survey of Multi-view Machine Learning” (Neural Computing and Applications, 2013), was published. It generates a good dissemination and promotion of multiview learning and has been well cited. Since then, much more research has been developed. This book aims to provide an in-depth and comprehensive introduction to multiview learning and hope to be helpful for AI researchers and practitioners. I have been working in the machine learning area for more than 15 years. Most of my work introduced in this book was completed after I graduated from Tsinghua University and joined East China Normal University in 2007. And we also include many important and representative works from other researchers to make the book content complete and comprehensive. Due to space and time limits, we may not be able to include all relevant works. I owe many thanks to the past and current members of my Pattern Recognition and Machine Learning Research Group, East China Normal University, for their hard work to make research done in time. The relationship between me and them is not just professors and students, but also comrades-in-arms. Content: Intro Preface Contents 1 Introduction 1.1 Background 1.2 Definition of Multiview Machine Learning and Related Concepts 1.3 Typical Application Fields in Artificial Intelligence 1.4 Why Can Multiview Learning Be Useful 1.5 Book Structure References 2 Multiview Semi-supervised Learning 2.1 Introduction 2.2 Co-training Style Methods 2.2.1 Co-training 2.2.2 Co-EM 2.2.3 Robust Co-training 2.3 Co-regularization Style Methods 2.3.1 Co-regularization 2.3.2 Bayesian Co-training 2.3.3 Multiview Laplacian SVM 2.3.4 Multiview Laplacian Twin SVM 2.4 Other Methods References 3 Multiview Subspace Learning3.1 Introduction 3.2 Canonical Correlation Analysis and Related Methods 3.2.1 Canonical Correlation Analysis 3.2.2 Kernel Canonical Correlation Analysis 3.2.3 Probabilistic Canonical Correlation Analysis 3.2.4 Bayesian Canonical Correlation Analysis 3.3 Multiview Subspace Learning with Supervision 3.3.1 Multiview Linear Discriminant Analysis 3.3.2 Multiview Uncorrelated Linear Discriminant Analysis 3.3.3 Hierarchical Multiview Fisher Discriminant Analysis 3.4 Other Methods References 4 Multiview Supervised Learning 4.1 Introduction 4.2 Multiview Large Margin Classifiers4.2.1 SVM-2K 4.2.2 Multiview Maximum Entropy Discriminant 4.2.3 Soft Margin-Consistency-Based Multiview Maximum Entropy Discrimination 4.3 Multiple Kernel Learning 4.3.1 Kernel Combination 4.3.2 Linear Combination of Kernels and Support Kernel Machine 4.3.3 SimpleMKL 4.4 Multiview Probabilistic Models 4.4.1 Multiview Regularized Gaussian Processes 4.4.2 Sparse Multiview Gaussian Processes 4.5 Other Methods References 5 Multiview Clustering 5.1 Introduction 5.2 Multiview Spectral Clustering 5.2.1 Co-trained Spectral Clustering 5.2.2 Co-regularized Spectral Clustering5.3 Multiview Subspace Clustering 5.3.1 Multiview Clustering via Canonical Correlation Analysis 5.3.2 Multiview Subspace Clustering 5.3.3 Joint Nonnegative Matrix Factorization 5.4 Distributed Multiview Clustering 5.5 Multiview Clustering Ensemble 5.6 Other Methods References 6 Multiview Active Learning 6.1 Introduction 6.2 Co-testing 6.3 Bayesian Co-training 6.4 Multiple-View Multiple-Learner 6.5 Active Learning with Extremely Spare Labeled Examples 6.6 Combining Active Learning with Semi-supervising Learning 6.7 Other Methods This book provides a unique, in-depth discussion of multiview learning, one of the fastest developing branches in machine learning. Multiview Learning has been proved to have good theoretical underpinnings and great practical success. This book describes the models and algorithms of multiview learning in real data analysis. Incorporating multiple views to improve the generalization performance, multiview learning is also known as data fusion or data integration from multiple feature sets. This self-contained book is applicable for multi-modal learning research, and requires minimal prior knowledge of the basic concepts in the field. It is also a valuable reference resource for researchers working in the field of machine learning and also those in various application domains. Read more... Abstract: This book provides a unique, in-depth discussion of multiview learning, one of the fastest developing branches in machine learning. Incorporating multiple views to improve the generalization performance, multiview learning is also known as data fusion or data integration from multiple feature sets. Read more...