Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Introduction to learning classifier systems
Urbanowicz R., Browne W., Springer Publishing Company, Incorporated, New York, NY, 2017. 123 pp. Type: Book (978-3-662550-06-9)
Date Reviewed: Jul 11 2018

In the preface, Introduction to learning classifier systems presents itself as an introductory textbook for undergraduate or graduate students, as well as for practitioners. The scope of the work is encompassing, and its brevity splendid.

Learning classifier systems (LCS) are a family of related techniques sharing the goal of finding and improving a solution to some encoded problem as more and more information about that problem becomes available. The distinguishing features of LCS are in how the problem domain is encoded and in how the solution domain is searched. The common representation of problem and solution is abstracted into a rule set. Rules are built with composable symbols grouped in alphabets. The application of rules to the problem domain results in a classifier, the outcome of which establishes some partition of the solution domain. Rules may be overgeneric, producing the equivalent of false positive matches, or overspecific, producing false negative matches. Learning entails selecting the best rules and generating new ones as necessary. Learning is fully successful when the outcome of a classifier applying a certain rule set to a problem yields a globally optimal solution. Remarkably, that rule set remains amenable to analytical interpretation, whereby an expert technologist can infer from examination a mechanistic understanding of how an optimal solution operates. The concepts underpinning all of the preceding are enumerated and composed in around 100 pages of terse explanations.

The strong suit of this work--brevity--is also its Achilles’ heel. Perhaps an alternate apt title for this work could have been Lecture notes on LCS. The practical details useful to crystallize the concepts and taxonomies into applicative confidence are left wanting. Of course, all works represent a balanced compromise, and Introduction to learning classifier systems aims at being conceptually comprehensive and brief, rather than hands-on. Only chapter 1, an overview of the subject matter, provides end-of-chapter exercises, which are tied to computer code therein referenced and independently available online. On the other hand, the theoretical approach to solving a practical nontrivial example problem is the thread unifying the conceptual presentation throughout the book. While unraveling this sample problem--the operation of a multiplexer--the chapters unfold as follows (after the already mentioned overview in chapter 1).

Chapter 2 is a taxonomy of concepts in LCS composed of learning, classifiers, systems, and applicability to problem domains, with observations on advantages and disadvantages. Chapter 3 presents the anatomy of how a functional cycle operates. It introduces the nature of encoding via alphabets and rules, rule fitness, matching and deletion, and discovery of new rules. Chapter 4 is a review of existing proven approaches to composing functional cycles. Chapter 5 consists of guidelines for parameter choices and for running computer codes, and also includes a compendium of selected references for books, journals, and reviews, some available in print, some online.

Introduction to learning classifier systems fits well as a general guide to the subject matter, even if it may fall short of filling the role of a textbook, except perhaps with substantial help from an instructor or great diligence from a self-learner approaching this topic.

More reviews about this item: Amazon

Reviewer:  A. Squassabia Review #: CR146143 (1809-0486)
Bookmark and Share
  Reviewer Selected
Featured Reviewer
 
 
Learning (I.2.6 )
 
 
Classifier Design And Evaluation (I.5.2 ... )
 
 
Reference (A.2 )
 
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