Matrix Methods in Data Mining and Pattern Recognition

Lars Elden

Publisher:

SIAM

Publication Date:

2019

Number of Pages:

229

Format:

Paperback

Edition:

2

Price:

69.00

ISBN:

978-1-611975-85-7

Category:

Textbook

MAA Review
Table of Contents

[Reviewed by

Bill Satzer

, on

12/14/2019

]

This is the second edition of a book first published in 2007; our review of the first edition is here. It maintains the structure, approach and tone of the first edition, and adds two new chapters. The author re-emphasizes his conservative approach to the topics he includes. He notes in particular that he has not included any discussion of neural networks in this volume, and suggests that he has misgivings with the whole subject of deep learning because its theoretical underpinnings are not strong.

One new chapter on graphs and matrices has been added to Part I of the book, the more theoretical piece. This provides an introduction to graph theory and its connections to matrix methods. An important part of this chapter introduces graph Laplacians and spectral partitioning. The idea here is to identify a way to partition a graph into multiple pieces by identifying the parts that are least connected to one another.

The other new chapter is a parallel one in Part II that further develops the idea of spectral partitioning and applies it in two extended applications: a social network (political blogs) and text classification.

Apart from these additions, the author has updated his figures and introduced some color. Citations have been added and updated, primarily in the area of applications.

This continues to be a book of interest. Some topics - particularly the more theoretical ones - are treated so succinctly that less experienced readers may need to look for greater detail elsewhere. It is clear that the author’s primary interest is with the applications. These are particularly well done, both as examples in Part I and in more extended forms in Part II. The book is worth consideration as a supplement to linear algebra courses or as a text for a more advanced class.

Bill Satzer (bsatzer@gmail.com), now retired from 3M Company, spent most of his career as a mathematician working in industry on a variety of applications ranging from speech recognition and network modeling to optical films and material science. He did his PhD work in dynamical systems and celestial mechanics.

See the publisher's web page.

Tags:

Data Mining