Fuzzy and evidence reasoning (Q1382416)

The book is devoted to the theory of nonclassical inference methods based on fuzzy logic and the theory of evidence initiated by Dempster and Shafer. It consists of 9 chapters. Chapters 1 and 2 are introductory and they provide an overview of some notions of fuzzy set and possibility theory, the concepts of a linguistic variable, fuzzy truth space, and also knowledge representation using special linguistically specified rules pertaining to modification, composition, quantification, and qualification. Chapters 3, 4, and 5 are devoted to fuzzy reasoning, i.e. inference based on the generalized rule of modus ponens and inference on the basis of more general schemes of rules such as projection, entailment, compositional rule of inference and other ones. The remaining chapters are devoted to the Dempster-Shafer theory of evidence, evidence reasoning based on two-valued as well as multivalued logic and, finally, interval-valued evidence theory. The book is evidently written by non-mathematicians and contains various mistakes and imprecisenesses. For example, on p. 5 an example of a complement of a fuzzy set is provided but without specification of the universe! Among various kinds of many-valued implications we can also find ``Mamdani implication'' (the operation of minimum). In general the book seems to be a compilation from a lot of sources, however quite often without understanding, and, therefore, the explanation is often misleading. For example, on pp. 78/79 the authors present generalized modus ponens using a max-min composition of a fuzzy set with a fuzzy relation derived on the basis of some implication. The example which follows uses Łukasiewicz implication. However, this is incorrect and gives counterintuitive results. This can be demonstrated immediately using the presented data: the authors consider the rule ``IF \(X\) is medium THEN \(Y\) is small'' where ``medium'' is interpreted by the fuzzy set \(\{0.6/2, 1/3, 0.5/4\}\) and ``small'' by \(\{1/2, 0.6/3, 0.2/4\}\). Using max-min composition, we derive from ``\(X\) is medium'' the fuzzy set \(\{0.5/1, 1/2, 0.6/3, 0.6/4\}\), which is far from being ``small''. Chapters 3-5 are a mixture of various operations without presenting a clear and justified system. Some discussed properties are already outdone, e.g., forcing ``IF \(A\) THEN \(B\)'' to imply ``very B'' from ``very A''; take as an example ``IF it is warm then I am fine''. What about the case when it is ``very warm'' -- does it follow that ``I am very fine''? Namely, the authors ignore the latest deep mathematical results in fuzzy logic, which is apparent when looking at the citations, since most cited works are rather old and the works of such distinguished mathematicians like Di Nola, Hájek, Höhle, Klement, Mesiar, Mundici, Pavelka, and others are completely missing. I could continue to list deficiencies of the book. For example, the concept of t-norm is only very vaguely mentioned and the authors incorrectly state that min and product operations with fuzzy sets reduce to classical operations when restricted to \(\{0, 1\}\) in contradistinction to other ones, i.e. t-norms! To my surprise, the Appendix contains the most important t-norms (and t-conorms), but incomplete and without mentioning that t-norms are in concern! Many objections can be raised also towards the linguistic considerations and philosophical discussions concerning fuzziness, possibility and probability. All in all, the book belongs to those works which raised the (unfair and unjustified) hostility of mathematicians towards fuzzy logic.