Nonlinear Dimensionality Reduction Techniques: A Data Structure Preservation Approach

This book provides tools for analyzing multidimensional and metric data, focusing on visual exploration through dimensionality reduction. It reviews existing techniques and proposes new solutions, such as ASKI, ClassNeRV, ClassJSE, and MING, for representing expert-designed fault indicators, I-V curves, and acoustic signals.

Format: Paperback / softback
Length: 247 pages
Publication date: 04 December 2022
Publisher: Springer Nature Switzerland AG

This comprehensive book offers a valuable resource for researchers and practitioners seeking advanced tools and techniques for analyzing multidimensional and metric data. By providing a state-of-the-art overview of existing solutions and developing innovative new ones, it empowers users to effectively explore and understand these complex data sets.

The book's primary focus lies in facilitating visual exploration of multidimensional and metric data by human analysts. It relies on the generation of 2D or 3D scatter plot displays through dimensionality reduction techniques, enabling domain experts to analyze the relationships between observed monitoring variables and the underlying internal state of the system.

Dimensionality reduction plays a crucial role in this analysis, as it allows for the representation of visually a multidimensional dataset, making it easier for experts to identify patterns, trends, and correlations. The book reviews existing techniques for visual data exploration and dimensionality reduction, including t-SNE and Isomap, and proposes novel solutions to address challenges in this field.

In particular, the book introduces the new unsupervised technique called ASKI, as well as the supervised methods ClassNeRV and ClassJSE. Additionally, a novel approach for local map quality evaluation called MING is presented. These methods are then applied to diverse real-world applications, such as the representation of expert-designed fault indicators for smart buildings, I-V curves for photovoltaic systems, and acoustic signals for Li-ion batteries.

By presenting a comprehensive and practical approach to multidimensional data analysis, this book serves as a valuable resource for researchers, engineers, and practitioners in various fields, including data science, machine learning, and engineering. It equips users with the tools and knowledge necessary to unlock the full potential of multidimensional and metric data and make informed decisions based on their insights.

Weight: 450g
Dimension: 235 x 155 (mm)
ISBN-13: 9783030810283
Edition number: 1st ed. 2022