Matrix Factorization for Multimedia Clustering - (Computing and Networks) by Hangjun Che & Xin Wang & Xing He & Man-Fai Leung & Baicheng Pan

About the Book

This book explores clustering principles through advanced data analysis techniques, such as matrix and tensor factorization in multimedia information processing. The authors present methods to address these challenges, examine popular regularization techniques, and explore the relationship between regularization and clustering performance.

Book Synopsis

Due to the rapid development of the internet of things, mobile communication and social media, more multimedia content (images, texts, videos and audios) is being created from multimedia platforms. To discover the intrinsic structure in unlabeled multimedia big data, it is crucial to partition objects into different groups and downstream tasks for application such as misinformation, epidemiology, user recommendation, and so on. Clustering is a fundamental problem in multimedia information processing.

This co-authored book explores clustering principles through advanced data analysis techniques such as matrix and tensor factorization which are highly relevant for multimedia information processing. Multimedia data may exhibit various forms of noise represented from multiple perspectives, making traditional clustering approaches less effective. The authors consider complex conditions such as noise sensitivity and discuss methods to address these challenges in the context of multimedia data. They examine popular regularization techniques, providing theoretical analyses that demonstrate the relationship between regularization and clustering performance.

The book will serve as a solid advanced reference for researchers, scientists, engineers and advanced students who wish to implement practical tasks through clustering formulations. Additionally, the authors provide a detailed description of convergence theory to enable readers to conduct the corresponding algorithm analyses. They investigate novel regularization techniques, such as self-paced learning, optimal graph learning, and diversity regularization, to uncover the geometric structure of data. These techniques are beneficial for enhancing clustering performance. Furthermore, they demonstrate the efficiency of these regularization techniques through theoretical analyses, practical experiments and applications in real-world datasets.