No label defined (Q4338957)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 1016747
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 1016747

      Statements

      4 June 1997
      0 references
      London (GB)
      0 references
      Proceedings
      0 references
      Conference
      0 references
      ADT '95
      0 references
      Applied decision technologies
      0 references
      genetic algorithms
      0 references
      probabilistic reasoning
      0 references
      computational learning
      0 references
      Bayesian belief networks
      0 references
      ``Computational Learning'' and ``Probabilistic Reasoning'' are two well established domains in the field of Artificial Intelligence and its theoretical foundations. In the Zentralblatt the corresponding reviews may be found mainly under 68T05 and 68T30, respectively. Computational Learning is the more theoretical and more abstract one of both domains. Although the reviewer himself has not been working in either one, he has been following the more or less independent development of both subdisciplines with an interested eye from the halfdistance. When he found this book in his mailing heap together with the request to review it, a cheerful expectation started to grow in him: Should this conference/book be the starting point of a mutual fertilization of both domains? But already a quick pass through the first pages considerably dampened the joy of anticipation. Although the first sentence of the preface still reads ``This book is devoted to two INTERRELATED techniques for solving \(\dots\)'' (accentuation by the reviewer), the table of contents shows the picture of a ``biased'' meeting: Only one of eighteen contributions stems from Computational Learning, namely Section 1 by V. Vapnik. The ``remainder'' of the book is written by several researchers from Probabilistic Reasoning and a few illustrious guests from outside, like for instance J. Rissanen. Unfortunately just these two sections by Vapnik and Rissanen have almost no concrete cross relations to all the other pieces of the book. Probably it would have been more exact not to mention the name ``Computational Learning'' in the book title at a par with ``Probabilistic Reasoning''. The book originated from the Conference ADT '95, Applied Decision Technologies, which took place in London in April 1995. This conference had four major streams, and one of them was devoted to Computational Learning and Probabilistic Reasoning. Besides the conference proceedings, it was decided to develop some of the papers of this special stream and publish them in a separate volume, together with several extra sections written by other scientists in the field(s). The result is the book under review. The book has four parts with altogether 18 sections, contributed by 18 different authors and/or teams of authors. We give the content. Part I--Generalisation Principles and Learning: Structure of Statistical Learning Theory (V. Vapnik, 29 pages), Stochastic Complexity -- an Introduction (J. Rissanen, 9 p.), MML Inference of Predictive Trees, Graphs and Nets (C. S. Wallace, 24 p.), Learning and Reasoning as Information Compression by Multiple Alignment, Unification and Search (J. G. Wolff, 19 p.), Probabilistic Association and Denotation in Machine Learning of Natural Language (P. Suppes and L. Liang, 14 p.). Part II--Causation and Model Selection: Causation, Action, and Counterfactuals (J. Pearl, 21 p.), Another Semantics for Pearl's Action Calculus (V. G. Vovk, 20p.), Efficient Estimation and Model Selection in Large Graphical Models (D. Wedelin, 15 p.), T-Normal Distribution on the Bayesian Belief Networks (Yu. N. Blagoveschensky, 6 p.). Part III--Bayesian Belief Networks and Hybrid Systems: Bayesian Belief Networks with an Application in Specific Case Analysis (C. G. G. Aitken et al, 16 p.), Bayesian Belief Networks and Patient Treatment (L. D. Meshalkin and E. K. Tsybulkin, 13 p.), A Higher Order Bayesian Neural Network for Classification and Diagnosis (A. Host and A. Lansner, 11 p.), Genetic Algorithms Applied to Bayesian Networks (P. Larranaga et al, 24 p.). Part IV--Decision-Making, Optimization and Classification: Rationality, Conditional Independence and Statistical Models of Competition (J. Q. Smith and C. T. J. Allard, 22 p.), Axioms for Dynamic Programming (P. P. Shenoy, 17 p.), Mixture-Model Cluster Analysis Using the Projection Pursuit Method (S. Aivazian, 18 p.), A Parallel -- Nearest Neighbour Classifier for Estimation of Non-Linear Decision Regions (A. Kovalenko, 8 p.), Extreme Values of Functionals Characterizing Stability of Statistical Decisions (A. V. Nagaev, 13 p.). With respect to the degree of difficulty, the book is not very homogeneous. Different sections have rather different demands on the reader, concerning both the amount of previous knowledge and the intellectual level. Especially some of the more theoretical contributions might be understood only by very insiders. But this reviewer had special problems also with the section on alignments. The reviewer was not a participant in the ADT '95 conference, and so he can not report, up to what extent participants with different backgrounds broadened each others horizons by their diverse talks and in discussions (he knows from his own fields of research how difficult it is to look and to think across rather firm boundaries of an established domain). However, on his backburner he keeps the hope that one day people from Computational Learning and from Probabilistic Reasoning might really start to understand and fertilize each other.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references