Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Unsupervised learning algorithms
Celebi M., Aydin K., Springer International Publishing, New York, NY, 2016. 558 pp. Type: Book (978-3-319242-09-5)
Date Reviewed: Aug 22 2016

This book offers a comprehensive overview of unsupervised learning; editors Celebi and Aydin provide an introduction. Many of the chapters, which are detailed below, describe clustering algorithms. Other chapters consider anomaly detection or are dedicated to specific applications or areas of application.

Although not formally divided into parts, the sequencing of the chapters effectively groups them into three areas. The first of these deals with anomaly detection--here, distinguished from outlier detection. This is followed by several chapters on clustering itself, with an emphasis on techniques rather than applications. The final chapters are application oriented.

Of the two chapters on anomaly detection, the first, by Deepak, deals with cases where data has some spatial attributes. The chapter gives a taxonomy of techniques that covers both object anomalies, basically outliers, and region anomalies. Several of these techniques are described in some detail. The second chapter, by Clémençon et al., concerns itself with the evaluation of the degree of anomaly, which is of crucial importance in applications such as fraud surveillance. The aim is to train data to produce an optimal mass-volume curve that serves as a way to evaluate the algorithm. Numerical experiments with synthetic data are presented.

The next 11 chapters are devoted to clustering techniques. Each chapter presents an overview of clustering techniques suited to a particular class of problem. Thus, Scrucca describes a genetic algorithm for model-based clustering where it is assumed that the data can be represented by a finite mixture model. Tomašev and Radovanović review various “quality measures and evaluate them in the ... context of high-dimensional data clustering.” This is followed by a chapter by Festa, which emphasizes combinatorial optimization. Langone et al. review the use of kernels in clustering. This is followed by a chapter by Keivani and Peña on the use of Bayesian networks for uni- and multi-dimensional clustering. Niros and Tsekouras describe a neural network approach to pattern classification. Dinler and Tural’s survey chapter considers the problem of constrained clustering, where prior knowledge about data is incorporated into clustering methods so as to enforce restrictions such as the requirement that two items belong (or do not belong) to the same cluster. Torra et al. give an overview of clustering applied to data privacy. Wang and Lai discuss nonlinear clustering where the clusters cannot be separated by linear functions; they describe an algorithm, GMPCL, and the chapter contains quite striking illustrations of the effectiveness of the algorithm at segmenting images. İnkaya et al. survey the use of swarm-based methods in clustering; the chapter includes as an appendix a useful table of particle swarm algorithms for clustering. Huang et al. describe how one can incorporate both intra-cluster compactness and inter-cluster separation into k-means techniques.

The remaining chapters are each oriented toward specific applications of clustering. Tsolakis and Tsekouras describe a fuzzy soft learning approach to image compression. A critical idea is to move the code words so as to better capture areas where more discrimination is required. Wong et al. consider the use of unsupervised learning in genome informatics. The challenge here is to identify sites where transcription factors may bind to deoxyribonucleic acid (DNA). The case where these bindings are imperfect can be an explanation for cancers. The chapter’s goal is to identify potential binding sites and groups of binding sites, which can then be explored using wet lab experiments. Martin et al. describe an application for the automatic grading of written (essay) responses to test questions using latent semantic analysis. The insight here is that the existence of huge datasets allows one to deduce commonalities between words based on co-occurrence in multiple documents, thus allowing the application to take into account that different words may refer to the same item or concept. Ahmed and Karypis describe algorithms that can extract patterns in evolving networks. In real-world applications, finding such patterns can provide insight into the evolution of real networks, for example, inter-country trade. In the final chapter, “Trentin and Bongini ... survey probabilistic interpretations of artificial neural networks (ANNs),” particularly those ANNs that can be used to estimate probability density functions. The application chapters have a tendency to be somewhat dense. Readers who are familiar with the application area will have less difficulty following the ideas than those who are unfamiliar with the area.

The book provides a valuable survey of an area of both research and application, particularly as massive datasets have become available. These large datasets have made unsupervised techniques necessary, and have made it possible to implement application ideas that previously could not be acted upon. The application to automated text response grading is a striking example of this in its use of large datasets to group words by meaning. The book can be recommended to anyone interested in getting an overview of this fast-moving research and application area. Because each chapter has a comprehensive bibliography, the book can serve as an entry point for those wishing to work in or with unsupervised learning.

Reviewer:  J. P. E. Hodgson Review #: CR144701 (1611-0792)
Bookmark and Share
  Featured Reviewer  
 
Learning (I.2.6 )
 
 
Applications And Expert Systems (I.2.1 )
 
Would you recommend this review?
yes
no
Other reviews under "Learning": Date
Learning in parallel networks: simulating learning in a probabilistic system
Hinton G. (ed) BYTE 10(4): 265-273, 1985. Type: Article
Nov 1 1985
Macro-operators: a weak method for learning
Korf R. Artificial Intelligence 26(1): 35-77, 1985. Type: Article
Feb 1 1986
Inferring (mal) rules from pupils’ protocols
Sleeman D.  Progress in artificial intelligence (, Orsay, France,391985. Type: Proceedings
Dec 1 1985
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy