- Email: [email protected]
Elie Sanchez, 1944–2014
Artificial Intelligence in Medicine 62 (2014) 73–77
Contents lists available at ScienceDirect
Artificial Intelligence in Medicine journal homepage: www.elsevier.com/locate/aiim
In memoriam
Elie Sanchez, 1944–2014
On March 6, 2014 the French mathematician, pioneer of fuzzy set theory and one of the founders of the fuzzy community, Professor Elie Sanchez, died of leukemia. He was 70 years old (Fig. 1). After high school Elie Sanchez passed the preparatory classes “Mathématiques supérieures” and “Mathématiques spéciales” at the Lycée Thiers in Marseille, and then attended the Facultés des Sciences de Saint-Charles and the Facultés des Sciences de Luminy in the same city. After studying mathematics at the Laboratoire de Biomathématiques, Statistique et Informatique Médicale of the Faculté de Médecine in Marseille in 1972, he acquired a Ph.D. in Mathematics from the Faculty of Sciences of that university with his thesis entitled Matrices et Fonctions en Logique Symbolique [1]. In this work he had considered Boolean matrix equations to be related to ternary logic, but shortly afterwards he switched to fuzzy logic. However, the term “fuzzy logic” did not exist before the Berkeley linguist Lakoff suggested it in 1972 [2,3]. In 1973 Elie Sanchez became familiar with the field of fuzzy sets, as he said in one of his most recent papers: “The first time I heard of fuzzy sets was during a seminar of Arnold Kaufmann (later on I also read an interview of L.A. Zadeh, called “Les ensembles Flous–un concept précis”, in l’Informatique, 1970) while he was presenting Zadeh’s theory and his own research. He was carrying with him the manuscript of his first book on the subject. The book was entitled “Introduction à la Théorie des Sous Ensembles Flous. Tome I, Eléments Théoriques de Base” (Masson, Paris, 1973). He called them “sous ensembles flous”, i.e., fuzzy subsets, for the class of points, or objects, on which membership functions were defined was not fuzzy. This manuscript was written by hand in India ink on tracing papers. So it was one of my early influences.” [4]1 Elie Sanchez then extended his work on Boolean matrix equations to fuzzy relation equations in his second Ph.D. thesis (in human biology from the Faculty of Medicine at the University of Méditeranée (University of Aix-Marseille II, 1974) Equations de Relations Floues [6]. Here he proposed an application of the principle for medical diagnosis (Fig. 2). Zadeh had devised his max-min rule as a composition rule for fuzzy relations. He expected interesting results from the application of fuzzy relations “in transportation problems and in belief systems” [7]. Assilian and Mamdani used it in 1972 to calculate inference rule relationships once they had implemented
1
For Kaufmann’s book see [5].
http://dx.doi.org/10.1016/j.artmed.2014.09.001 0933-3657/
Fig. 1. Elie Sanchez at the IIZUKA-88 International Video-Session, KIT-NASA at the Kyushu Institute of Technology in Iizuka, Japan. Photograph taken from E. Sanchez’ private collection.
Fig. 2. Elie Sanchez and Arnold Kaufmann in France 1980.
the fuzzy IF–THEN rules for their fuzzy control algorithm as fuzzy relations [8,9]. Elie Sanchez now pursued a different direction as he later wrote in [10,p. 47]: “We plan to investigate medical aspects of fuzzy relations at some future time.” He assumed that
74
In memoriam / Artificial Intelligence in Medicine 62 (2014) 73–77
Fig. 3. Elie Sanchez and Lotfi A. Zadeh at the IEEE Conference on Decision and Control, Fairmont Hotel, New Orleans, Louisiana, 1977. Photograph taken from E. Sanchez’ private collection.
a doctor translates his knowledge and his experience into degrees of association between symptoms and diagnoses. Already in 1976, in a Master’s thesis entitled Medical Knowledge Network: A Database for Computer-aided Diagnosis, Alonso PerezOjeda at the Department of Industrial Engineering of the University of Toronto had suggested that medical knowledge could be represented as a network in which symptoms and diseases were linked to one another by relations [11]. In order to model the “relation strength” (labeled as, e.g., “usually”, “occasionally”, “almost always”) mathematically, he identified them with probability modifiers interpreted using frequency theory. Probably without any knowledge of Perez-Ojeda’s work, Elie introduced the relationships between the set of symptoms and the set of diseases as fuzzy relations; these fuzzy relations, he found, represented the corpus of medical knowledge: “In a given pathology, we denote by S a set of symptoms, D a set of diagnoses, and P a set of patients. What we call “medical knowledge” is a fuzzy relation, generally denoted by R, from S to D expressing associations between symptoms, or syndromes, and diagnoses, or groups of diagnoses (Fig. 3).”[12]. Elie interpreted Zadeh’s max–min rule directly to develop diagnoses. Given symptom and diagnosis sets S and D and an existing fuzzy relation R ⊂ S × D between them, the max–min composition may serve as an “inference rule”, which makes it possible to deduce imprecise descriptions of a patient’s illnesses (fuzzy sets of D) from imprecise symptom descriptions (fuzzy sets of S). With this inference rule, medical diagnoses Dj about a patient’s disease can be derived by fuzzy logic from symptoms with the help of the medical knowledge represented by the fuzzy relation R. By taking into account a set P of all patients considered and another fuzzy relation Q between P and the symptom set S, it was now possible, with the aid of the max–min composition rule, to obtain a fuzzy relation T = Q × R. The fuzzy relation R can be expressed as a matrix, the entries of which can be made after interviewing doctors about their diagnostic experience. This expert medical knowledge must additionally be translated into degrees of association between symptoms and diagnoses. Later he reported: “This work was still related to Zadeh’s original fuzzy sets with [0,1]-valued membership functions. Then I extended it to L-fuzzy sets,2 where L was a complete Brouwerian lattice. The aim was to propose the largest class of lattice-valued fuzzy sets to which the resolution methodology could apply.” He published his results in an article entitled “Resolution of Composite Fuzzy Relation Equations” [10].
2
For “L-sets” see Goguen [13].
Three years later he had turned the problem around: Instead of calculating the fuzzy relation T = Q × R from the available fuzzy relations Q and R, it was much more interesting and closer to clinical practice, at least in terms of diagnostics, to determine the fuzzy relation R from the known fuzzy relations T and Q. This would mean, of course, acquiring “medical knowledge” on the basis of knowledge about symptoms and diseases (clinical discharge diagnosis) from a patient set; in other words, to discover the relationships between symptoms and diseases. For the volume about Advances in Fuzzy Set Theory and its Applications [14], he shaped this plan in two closely related papers in which he demonstrated how the max–min composition rules Zadeh had introduced could be used as a rule of inference, in particular in medical diagnostics. “Compositions of Fuzzy Relations” [15] is the first text, an abstract mathematical foundation for the second: “Medical Diagnosis and Composite Fuzzy Relations” [12] In the second article Elie Sanchez referred to the fact that medical diagnoses often had to be made without any precise analysis being possible. One or more illnesses then had to be inferred from a patient’s symptoms, which most often cannot be described in any exact way. In so doing, neither the set of diseases taken into consideration nor the conclusion about the disease(s) drawn from the symptoms can be precise. In these two (interrelated) texts in 1979, Elie then highlights how it is possible to determine the largest relation matrix R for which: T = Q × R. Here, he also examined the remaining possibility, namely that of calculating the fuzzy relation Q from a given T and R. His question was: In what range can a patient’s symptoms vary without the patient dropping out of the diagnosis cluster for the respective disease? This meant determining the largest relation matrix Q for which T = Q × R. Motivated by Elie’s work on fuzzy relations that can be used in the medical field, in Vienna, Austria, Klaus-Peter Adlassnig fuzzified the CADIAG-I system–the computer-assisted DIAGnosis system that was being developed since 1968 at the department for medical computer sciences at the University of Vienna’s Medical School: “Fuzzy Set Theory with its capability of defining inexact medical entities as fuzzy sets, with its linguistic approach providing an excellent approximation to medical texts as well as its power of approximate reasoning, seems to be perfectly appropriate for designing and developing Computer-Assisted diagnostic, prognostic and treatment recommendation systems.” [16] In CADIAG-II, the relationships “occurrence” and “strength of confirmation” that were considered between symptoms and diagnoses were represented as fuzzy relations with the aid of his devised mathematical ammunition. In addition, the “occurrence” of a symptom Si in the case of a diagnosis Dj was defined as the frequency with which it occurs, while the “strength of confirmation” of the symptom was established as the importance it has for the diagnosis. The mainframe computer on which CADIAG-II had been running was shut down in 2004; this also marked the end of the CADIAG-II system. Now, in Vienna General Hospital, the knowledge representation and inference process in the medical consultation system MedFrame/CADIAG-IV [17] and also the MONI system (Monitoring of nosocomial infections)—a database and monitoring system for surveillance of nosocomial infections [18,19]—are based on fuzzy set theory and fuzzy logic, and thus on Elie Sanchez’ pioneering theoretical work. In the 1980s, following Kaufmann’s lectures, Elie gave lectures in China and Japan to introduce fuzzy sets. Fig. 4 reproduces a transparency that he used in those times. “The baldness of the human was meant to recall somebody” he wrote, when he published this illustration in [4]. At the end of the 1970s and in the 1980s, Elie Sanchez co-edited several volumes on fuzzy sets, fuzzy logic, approximate reasoning,
In memoriam / Artificial Intelligence in Medicine 62 (2014) 73–77
75
Fig. 5. From left to right: Lotfi A. Zadeh, Michio Sugeno, Elie Sanchez and Takeshi Yamakawa. Photograph taken from E. Sanchez’ private collection.
Fig. 4. Elie Sanchez’ drawing in the 1980s. Picture taken from E. Sanchez’ private collection.
genetic algorithms in decision processes, knowledge engineering, intelligent systems, and similar topics [20–25]. In the 1990s Zadeh suggested considering “Soft Computing” as a “coalition of methodologies which are drawn together by a quest for accommodation with the pervasive imprecision of the real world. At this juncture, the principal members of the coalition are fuzzy logic, neuro-computing, evolutionary computing, probabilistic computing, chaotic computing and machine learning.” ([26], p. 1–2) Elie supported this idea and he published intensively not only on fundamentals in fuzzy set theory [27–33] and with applications to medicine and biology [34,35], but also on soft computing and its disciplines [36–38]. Furthermore, he established new research areas, such as “DNA biosoft computing”. Elie was a visiting research associate at the University of California, Berkeley, a council member of the Kyushu Institute of Technology, Iizuka (Japan), and an invited professor at the University of Buenos Aires, Argentina. He was Director of the Neurinfo Research Department at the Institut Mediterraneen de Technologie, and President of the Neural & Fuzzy Systems Institute in Marseille, that worked successfully for many practical applications including fuzzy-logic-based decision support systems (“FuzzIA”), hand-written pattern recognition (“Eureca”), fuzzy-logic-controlled mobile robots (“Vitaur”) and automotive applications (particularly with Peugeot), vocational guidance systems (“FuzzADe”), and fuzzy information retrieval systems with optimization of weighted queries by genetic algorithms. Motivated by the vision of “The Semantic Web. A new form of Web content that is meaningful to computers will unleash a revolution of new possibilities,” published by Berners-Lee, Hendler, and Lassila in the Scientific American article in 2001 [39], and by Zadeh’s ideas on “Web Intelligence, World Knowledge and Fuzzy Logic” [40,41], Elie started new research interests on fuzzy logic in the Internet, the world wide web [39], and web intelligence [42]. He organized special sessions and workshops on “Fuzzy Logic in the Semantic Web” and published first papers on this new approach [43–48], which were “presented as an extension of the current web in which information is given well-defined meaning, better enabling computers and people to work in cooperation”. Elie
believed in “the positive role fuzzy logic, and more generally soft computing, can play in the development of the Semantic Web, filling a gap and facing a new challenge” [46]. In 2006, he edited the book Fuzzy Logic and the Semantic Web as the first volume in the new book series “Capturing Intelligence” [49]. In his preface to this book, series editor van Harmelen wrote: “The insight that any realistic approach will have to take into account the lessons in a wide community, and slowly but surely, the Semantic Web community is waking up to this fact.” [49, p. vii]. Elie Sanchez was a former President of the International Fuzzy Systems Association (IFSA) and was elected IFSA Fellow. In 1995 he received the International Grigore Moisil Gold Medal and Award for “capital contributions to computer and information sciences, namely in the fields of Fuzzy Systems and Artificial Intelligence”. He was co-editor of ten books and served on the editorial board of 20 international journals, including the present journal entitled Artificial Intelligence in Medicine. On March 10, Professor Takeshi Yamakawa (see Fig. 5) informed the BISC mailing group about the death of Elie Sanchez. Many members of the fuzzy community expressed their condolences. Henri Prade added in his personal note: “It is thanks to Elie that Didier and I [Henri] heard about possibility theory for the first time in 1977.” Lotfi Zadeh wrote in his message that he “saw him in
Fig. 6. Elie Sanchez (middle) with Enric Trillas (left) and José Angel Olivas (right) in Madrid, when Lotfi A. Zadeh was granted to the BBVA Foundation Frontiers of Knowledge Award 2012. Photo: R. Seising.
76
In memoriam / Artificial Intelligence in Medicine 62 (2014) 73–77
Madrid a couple of [months]3 ago. He looked good. His illness was in remission. I feel deeply saddened by his leaving our community.” The last meeting took place when the fifth edition of the BBVA Foundation Frontiers of Knowledge Award 2012 in the category of Information and Communication Technologies (ICT) was given to Lotfi Zadeh “for the invention and development of fuzzy logic”. It was in June 2013 and it was the day I saw Elie for the last time too. After the official part of the celebration I took the photograph that is published here as a last picture (Fig. 6).
Acknowledgments I am very grateful to his widow Mrs. Marie Sanchez, their son Guillaume Sanchez, and their daughter Emilie Sanchez for helping me to write this obituary.
References [1] Sanchez E. Matrices et Fonctions en Logique Symbolique, Thèse de 3e cycle: Mathématiques appliquées: Université Aix-Marseille 2. Faculté des Sciences; 1972. [2] Lakoff George. Hedges: a study in meaning criteria and the logic of fuzzy concepts. In: Peranteau Paul M, Levi, Judith N, Phares Gloria C, editors. Papers from the 8th regional meeting. Chicago, IL: Chicago Linguistic Society; 1972. p. 183–228. [3] Lakoff G. Hedges: a study in meaning criteria and the logic of fuzzy concepts. J Philos Logic 1973;2:458–508. [4] Sanchez E. The robot and the butterfly. In: Seising R, Trillas E, Termini S, Moraga C, editors. On fuzziness. a homage to Lotfi A. Zadeh, vol. II. Berlin (Studies in Fuzziness and Soft Computing, 299): Springer; 2013. p. 625–30. [5] Kaufmann A. Introduction à la théorie des sous-ensembles flous à l’usage des ingénieurs, Vol. 1 Éléments théoriques de base. Paris: Masson; 1977. [6] Sanchez E. Equations de Relations Floues, Thèse Biologie Humaine. Marseille: Faculté de Médecine de Marseille; 1974. [7] Zadeh LA. The concepts of system, aggregate, and state in system theory. In: Zadeh LA, Polak E, editors. System theory. New York, NY: McGraw-Hill; 1969. p. 3–42. [8] Assilian S. Artificial intelligence in the control of real dynamic systems. London: Queen Mary College, University of London; 1974 (August) (Ph.D. Thesis, No. DX193553). [9] Mamdani EH, Assilian S. An experiment in linguistic synthesis with a fuzzy logic controller. Int J Man Mach Stud 1975;7(1):1–13. [10] Sanchez E. Resolution of composite fuzzy relation equations. Inf Control 1976;30(1):38–48. [11] Perez-Ojeda A. Medical knowledge network. A database for computer aided diagnosis. Canada: Department of Industrial Engineering, University of Toronto; 1976 (Master Thesis). [12] Sanchez E. Medical diagnosis and composite fuzzy relations. In: Gupta MM, Ragade RK, Yager RR, editors. Advances in fuzzy set theory and applications. Amsterdam: North-Holland; 1979. p. 437–44. [13] Goguen JA. L-fuzzy sets. J Math Anal Appl 1967;18:145–74. [14] Gupta MM, Rommohan K Ragade, Ronald R Yager, editors. Advances in fuzzy set theory and applications. Amsterdam: North-Holland; 1979. [15] Sanchez E. Compositions of fuzzy relations. In: Gupta MM, Ragade RK, Yager RR, editors. Advances in fuzzy set theory and applications. Amsterdam: NorthHolland; 1979. p. 421–33. [16] Adlassnig K-P. A survey on medical diagnosis and fuzzy subsets. In: Gupta M, Sanchez E, editors. Approximate reasoning in decision analysis. Amsterdam: North-Holland; 1982. p. 203–17, 205. [17] Boegl K, Leitich H, Kolousek G, Rothenfluh ThE, Adlassnig K-P. Knowledge acquisition in MedFrame/Cadiag: a generalized fuzzy approach. In: Cimino JJ, editor. AMIA annual fall symposium. Philadelphia: Hanley & Belfus Inc.; 1996 (Supplement, 833). [18] Adlassnig K-P, Blacky A, Koller W. Fuzzy-based nosocomial infection control. In: Nikravesh M, Kacprzyk J, Zadeh LA, editors. Forging new frontiers: fuzzy pioneers II. Berlin: Springer; 2008. p. 343–50 (Studies in Fuzziness and Soft Computing, 218). [19] Adlassnig K-P, Blacky A, Mandl H, Rappelsberger A, Koller W. Fuzziness in Healthcare-Associated Infection Monitoring and Surveillance. In: Winston A, Melek W, Hall P, McKenzie S, Gibbs M, Adamson G, editors. Proceedings of the 2014 IEEE conference on Norbert Wiener in the 21st century—driving technology’s future. Piscataway: Institute of Electrical and Electronics Engineers (IEEE); 2014 (86.pdf.).
3
Lotfi Zadeh wrote “years” but it was actually months.
[20] Gupta MM, Sanchez E, editors. Fuzzy information and decision processes. Amsterdam: North-Holland; 1982. [21] Gupta MM, Sanchez E, editors. Approximate reasoning in decision analysis. Amsterdam: North-Holland; 1982. [22] Sanchez E, editor. Fuzzy information, knowledge representation and decision analysis: proceedings of the International Federation of Automatic Control (IFAC) symposium. Oxford: Pergamon Press; 1983. [23] Sanchez E, Zadeh LA, editors. Approximate reasoning in intelligent systems, decision & control. Elmsford: Pergamon Press; 1987. [24] Di Nola A, Pedrycz W, Sanchez E, Sessa S. Fuzzy relation equations and their applications to knowledge engineering. Dordrecht: Kluwer Academic Publishers; 1989. [25] Sanchez E, Shibata T, Zadeh LA. Genetic algorithms and fuzzy logic systems: soft computing perspectives. Singapore: World Scientific Publishing Company; 1997. [26] Zadeh LA. Foreword. Appl Soft Comput 2001;1(1):1–2. [27] Sanchez E. Importance in knowledge systems. Inf Syst 1989;14(6): 455–64. [28] Sanchez E. On possibility qualification in natural languages. Inf Sci 1978;15(1):45–76. [29] Sanchez E. Non standard fuzzy arithmetic. In: Wang PZ, Loe KF, editors. Between mind and computer: fuzzy science and engineering. Singapore: World Scientific Publishing Company; 1994. p. 271–82. [30] Avrachenkov KE, Sanchez E. Fuzzy markov chains and decision-making. Fuzzy Optim Decis Making 2002;1(2):143–59. [31] Sanchez E. Eigen fuzzy sets and fuzzy relations. J Math Anal Appl 1981;81(2):399–421. [32] Di Nola A, Pedrycz W, Sessa S, Sanchez E. Fuzzy relation equations theory as a basis of fuzzy modelling: an overview. Fuzzy Sets Syst 1991;40(3): 415–29. [33] Sanchez E. Truth-qualification and fuzzy relations in natural languages, application to medical diagnosis. Fuzzy Sets Syst 1996;84(2):155–67. [34] Bartolin R, Bonniol V, Sanchez E. Inflammatory protein variations: medical knowledge representation and approximate reasoning. In: Bouchon B, Saitta L, Yager RR, editors. Uncertainty and Intelligent Systems—Proceedings of the 2nd international conference on information processing and management of uncertainty in knowledge-based systems (IPMU’88). Berlin: Springer; 1988. p. 306–13. [35] Sanchez E, Pierre Ph. Intelligent decision making systems: from medical diagnosis to vocational guidance. In: Szczepaniak PS, Lisboa PJG, Kacprzyk J, editors. Fuzzy systems in medicine. Heidelberg: Physica-Verlag; 2000. p. 204–23. [36] Sanchez E. Soft computing perspectives. In: Proceedings of the twenty-forth international symposium on multiple-valued logic. Boston: IEEE; 1994. p. 276–81. [37] Sanchez E. DNA biosoft computing. In: Yamakawa T, Matsumoto G, editors. Methodologies for the conception, design, and application of intelligent systems. Proceedings of the fourth international conference on soft computing. Singapore: World Scientific Publishing Company; 1996. p. 30–7. [38] Sanchez E. Fuzzy logic knowledge systems and artificial neural networks in medicine and biology. In: Yager RR, Zadeh LA, editors. An introduction to fuzzy logic applications in intelligent systems. Boston, MA: Kluwer Academic Publishers; 1992. p. 235–51. [39] Berners-Lee T, Hendler J, Lassila O. The Semantic Web. A new form of web content that is meaningful to computers will unleash a revolution of new possibilities. Sci Am 2001;284((5) May):34–43. [40] Zadeh LA. Web intelligence, world knowledge and fuzzy logic—the concept of web IQ (WIQ). In: Negoita MGh, Howlett RJ, Jain LC, editors. Knowledgebased intelligent information and engineering systems, Proceedings of the 8th International Conference, KES 2004, September 20–25, 2004. Wellington, New Zealand, Part I: Springer, Berlin; 2004. p. 1–5. [41] Zadeh LA. A note on web intelligence, world knowledge and fuzzy logic. Data & Knowledge Engineering—Special jubilee issue: DKE 50 2004;50(3): 291–304. [42] Sanchez E. Fuzzy logic e-motion. In: Nikravesh M, Azvine B, editors. FLINT 2001—New directions in enhancing the power of internet. Proceedings of the 2001 BISC international workshop on fuzzy logic and the internet, Berkeley, California UC Berkeley Electronics Research Laboratory Memorandum No. UCB/ERL M01/28. 2001. p. 10–4. [43] IPMU. Special session “Fuzzy logic in the Semantic Web: a new challenge”. In: Bouchon-Meunier B, Coletti G, Yager RR, editors. Proceedings of the 10th international conference on information processing and management of uncertainty in knowledge-based systems (IPMU’04). Perugia, Italy: Casa Editrice Universita’ la Sapienza, Roma; 2004. p. 1017–38. [44] Fuzzy Logic and the Semantic Web. Workshop Marseille, France 2005. Extended abstracts available at: http://www.lif.univ-mrs.fr/FLSW. [45] FLSW-II. Fuzzy logic and the Semantic Web, second workshop abstracts. In: Bouchon-Meunier B, Yager RR, editors. Proceedings of the 11th international conference on information processing and management of uncertainty in knowledge-based systems (IPMU’2006). Paris, France: EDK; 2006. [46] Sanchez E, Yamanoi T. D’Amico fuzzy ontologies for the Semantic Web. In: Larsen HL, Pasi G, Ortiz-Arroyo T, Andreasen T, Christiansen H, editors. Flexible query answering systems, proceedings of the 7th International conference FQAS 2006. Berlin: Springer; 2006. p. 691–9.
In memoriam / Artificial Intelligence in Medicine 62 (2014) 73–77 [47] Calegari S, Sanchez E. A fuzzy ontology-approach to improve semantic information retrieval. In: Bobillo F, da Costa PCG, D’Amato C, Fanizzi N, Fung F, Lukasiewicz T, et al., editors. Proceedings of the Third ISWC Workshop on uncertainty reasoning for the Semantic Web (URSW’07), CEUR-Workshop Proceedings, Vol. 327. Nov. 12, 2007, Busan, Korea; 2007. Available at: http://ceur-ws.org/Vol-327/. [48] Calegari S, Sanchez E. Object-fuzzy concept network: an enrichment of ontologies in semantic information retrieval. J Assoc Info Sci Technol 2008;59(13):2171–80. [49] Sanchez E, editor. Fuzzy logic and the Semantic Web. Amsterdam: Elsevier; 2006 (Capturing Intelligence 1).
77
Rudolf Seising ∗ European Centre for Soft Computing, Edificio de Investigación, Calle Gonzalo Gutiérrez Quirós S/N, 33600 Mieres, Asturias, Spain ∗ Tel.: +34 985456545; fax: +34 985456699. E-mail address: [email protected]
2 September 2014
Contents lists available at ScienceDirect
Artificial Intelligence in Medicine journal homepage: www.elsevier.com/locate/aiim
In memoriam
Elie Sanchez, 1944–2014
On March 6, 2014 the French mathematician, pioneer of fuzzy set theory and one of the founders of the fuzzy community, Professor Elie Sanchez, died of leukemia. He was 70 years old (Fig. 1). After high school Elie Sanchez passed the preparatory classes “Mathématiques supérieures” and “Mathématiques spéciales” at the Lycée Thiers in Marseille, and then attended the Facultés des Sciences de Saint-Charles and the Facultés des Sciences de Luminy in the same city. After studying mathematics at the Laboratoire de Biomathématiques, Statistique et Informatique Médicale of the Faculté de Médecine in Marseille in 1972, he acquired a Ph.D. in Mathematics from the Faculty of Sciences of that university with his thesis entitled Matrices et Fonctions en Logique Symbolique [1]. In this work he had considered Boolean matrix equations to be related to ternary logic, but shortly afterwards he switched to fuzzy logic. However, the term “fuzzy logic” did not exist before the Berkeley linguist Lakoff suggested it in 1972 [2,3]. In 1973 Elie Sanchez became familiar with the field of fuzzy sets, as he said in one of his most recent papers: “The first time I heard of fuzzy sets was during a seminar of Arnold Kaufmann (later on I also read an interview of L.A. Zadeh, called “Les ensembles Flous–un concept précis”, in l’Informatique, 1970) while he was presenting Zadeh’s theory and his own research. He was carrying with him the manuscript of his first book on the subject. The book was entitled “Introduction à la Théorie des Sous Ensembles Flous. Tome I, Eléments Théoriques de Base” (Masson, Paris, 1973). He called them “sous ensembles flous”, i.e., fuzzy subsets, for the class of points, or objects, on which membership functions were defined was not fuzzy. This manuscript was written by hand in India ink on tracing papers. So it was one of my early influences.” [4]1 Elie Sanchez then extended his work on Boolean matrix equations to fuzzy relation equations in his second Ph.D. thesis (in human biology from the Faculty of Medicine at the University of Méditeranée (University of Aix-Marseille II, 1974) Equations de Relations Floues [6]. Here he proposed an application of the principle for medical diagnosis (Fig. 2). Zadeh had devised his max-min rule as a composition rule for fuzzy relations. He expected interesting results from the application of fuzzy relations “in transportation problems and in belief systems” [7]. Assilian and Mamdani used it in 1972 to calculate inference rule relationships once they had implemented
1
For Kaufmann’s book see [5].
http://dx.doi.org/10.1016/j.artmed.2014.09.001 0933-3657/
Fig. 1. Elie Sanchez at the IIZUKA-88 International Video-Session, KIT-NASA at the Kyushu Institute of Technology in Iizuka, Japan. Photograph taken from E. Sanchez’ private collection.
Fig. 2. Elie Sanchez and Arnold Kaufmann in France 1980.
the fuzzy IF–THEN rules for their fuzzy control algorithm as fuzzy relations [8,9]. Elie Sanchez now pursued a different direction as he later wrote in [10,p. 47]: “We plan to investigate medical aspects of fuzzy relations at some future time.” He assumed that
74
In memoriam / Artificial Intelligence in Medicine 62 (2014) 73–77
Fig. 3. Elie Sanchez and Lotfi A. Zadeh at the IEEE Conference on Decision and Control, Fairmont Hotel, New Orleans, Louisiana, 1977. Photograph taken from E. Sanchez’ private collection.
a doctor translates his knowledge and his experience into degrees of association between symptoms and diagnoses. Already in 1976, in a Master’s thesis entitled Medical Knowledge Network: A Database for Computer-aided Diagnosis, Alonso PerezOjeda at the Department of Industrial Engineering of the University of Toronto had suggested that medical knowledge could be represented as a network in which symptoms and diseases were linked to one another by relations [11]. In order to model the “relation strength” (labeled as, e.g., “usually”, “occasionally”, “almost always”) mathematically, he identified them with probability modifiers interpreted using frequency theory. Probably without any knowledge of Perez-Ojeda’s work, Elie introduced the relationships between the set of symptoms and the set of diseases as fuzzy relations; these fuzzy relations, he found, represented the corpus of medical knowledge: “In a given pathology, we denote by S a set of symptoms, D a set of diagnoses, and P a set of patients. What we call “medical knowledge” is a fuzzy relation, generally denoted by R, from S to D expressing associations between symptoms, or syndromes, and diagnoses, or groups of diagnoses (Fig. 3).”[12]. Elie interpreted Zadeh’s max–min rule directly to develop diagnoses. Given symptom and diagnosis sets S and D and an existing fuzzy relation R ⊂ S × D between them, the max–min composition may serve as an “inference rule”, which makes it possible to deduce imprecise descriptions of a patient’s illnesses (fuzzy sets of D) from imprecise symptom descriptions (fuzzy sets of S). With this inference rule, medical diagnoses Dj about a patient’s disease can be derived by fuzzy logic from symptoms with the help of the medical knowledge represented by the fuzzy relation R. By taking into account a set P of all patients considered and another fuzzy relation Q between P and the symptom set S, it was now possible, with the aid of the max–min composition rule, to obtain a fuzzy relation T = Q × R. The fuzzy relation R can be expressed as a matrix, the entries of which can be made after interviewing doctors about their diagnostic experience. This expert medical knowledge must additionally be translated into degrees of association between symptoms and diagnoses. Later he reported: “This work was still related to Zadeh’s original fuzzy sets with [0,1]-valued membership functions. Then I extended it to L-fuzzy sets,2 where L was a complete Brouwerian lattice. The aim was to propose the largest class of lattice-valued fuzzy sets to which the resolution methodology could apply.” He published his results in an article entitled “Resolution of Composite Fuzzy Relation Equations” [10].
2
For “L-sets” see Goguen [13].
Three years later he had turned the problem around: Instead of calculating the fuzzy relation T = Q × R from the available fuzzy relations Q and R, it was much more interesting and closer to clinical practice, at least in terms of diagnostics, to determine the fuzzy relation R from the known fuzzy relations T and Q. This would mean, of course, acquiring “medical knowledge” on the basis of knowledge about symptoms and diseases (clinical discharge diagnosis) from a patient set; in other words, to discover the relationships between symptoms and diseases. For the volume about Advances in Fuzzy Set Theory and its Applications [14], he shaped this plan in two closely related papers in which he demonstrated how the max–min composition rules Zadeh had introduced could be used as a rule of inference, in particular in medical diagnostics. “Compositions of Fuzzy Relations” [15] is the first text, an abstract mathematical foundation for the second: “Medical Diagnosis and Composite Fuzzy Relations” [12] In the second article Elie Sanchez referred to the fact that medical diagnoses often had to be made without any precise analysis being possible. One or more illnesses then had to be inferred from a patient’s symptoms, which most often cannot be described in any exact way. In so doing, neither the set of diseases taken into consideration nor the conclusion about the disease(s) drawn from the symptoms can be precise. In these two (interrelated) texts in 1979, Elie then highlights how it is possible to determine the largest relation matrix R for which: T = Q × R. Here, he also examined the remaining possibility, namely that of calculating the fuzzy relation Q from a given T and R. His question was: In what range can a patient’s symptoms vary without the patient dropping out of the diagnosis cluster for the respective disease? This meant determining the largest relation matrix Q for which T = Q × R. Motivated by Elie’s work on fuzzy relations that can be used in the medical field, in Vienna, Austria, Klaus-Peter Adlassnig fuzzified the CADIAG-I system–the computer-assisted DIAGnosis system that was being developed since 1968 at the department for medical computer sciences at the University of Vienna’s Medical School: “Fuzzy Set Theory with its capability of defining inexact medical entities as fuzzy sets, with its linguistic approach providing an excellent approximation to medical texts as well as its power of approximate reasoning, seems to be perfectly appropriate for designing and developing Computer-Assisted diagnostic, prognostic and treatment recommendation systems.” [16] In CADIAG-II, the relationships “occurrence” and “strength of confirmation” that were considered between symptoms and diagnoses were represented as fuzzy relations with the aid of his devised mathematical ammunition. In addition, the “occurrence” of a symptom Si in the case of a diagnosis Dj was defined as the frequency with which it occurs, while the “strength of confirmation” of the symptom was established as the importance it has for the diagnosis. The mainframe computer on which CADIAG-II had been running was shut down in 2004; this also marked the end of the CADIAG-II system. Now, in Vienna General Hospital, the knowledge representation and inference process in the medical consultation system MedFrame/CADIAG-IV [17] and also the MONI system (Monitoring of nosocomial infections)—a database and monitoring system for surveillance of nosocomial infections [18,19]—are based on fuzzy set theory and fuzzy logic, and thus on Elie Sanchez’ pioneering theoretical work. In the 1980s, following Kaufmann’s lectures, Elie gave lectures in China and Japan to introduce fuzzy sets. Fig. 4 reproduces a transparency that he used in those times. “The baldness of the human was meant to recall somebody” he wrote, when he published this illustration in [4]. At the end of the 1970s and in the 1980s, Elie Sanchez co-edited several volumes on fuzzy sets, fuzzy logic, approximate reasoning,
In memoriam / Artificial Intelligence in Medicine 62 (2014) 73–77
75
Fig. 5. From left to right: Lotfi A. Zadeh, Michio Sugeno, Elie Sanchez and Takeshi Yamakawa. Photograph taken from E. Sanchez’ private collection.
Fig. 4. Elie Sanchez’ drawing in the 1980s. Picture taken from E. Sanchez’ private collection.
genetic algorithms in decision processes, knowledge engineering, intelligent systems, and similar topics [20–25]. In the 1990s Zadeh suggested considering “Soft Computing” as a “coalition of methodologies which are drawn together by a quest for accommodation with the pervasive imprecision of the real world. At this juncture, the principal members of the coalition are fuzzy logic, neuro-computing, evolutionary computing, probabilistic computing, chaotic computing and machine learning.” ([26], p. 1–2) Elie supported this idea and he published intensively not only on fundamentals in fuzzy set theory [27–33] and with applications to medicine and biology [34,35], but also on soft computing and its disciplines [36–38]. Furthermore, he established new research areas, such as “DNA biosoft computing”. Elie was a visiting research associate at the University of California, Berkeley, a council member of the Kyushu Institute of Technology, Iizuka (Japan), and an invited professor at the University of Buenos Aires, Argentina. He was Director of the Neurinfo Research Department at the Institut Mediterraneen de Technologie, and President of the Neural & Fuzzy Systems Institute in Marseille, that worked successfully for many practical applications including fuzzy-logic-based decision support systems (“FuzzIA”), hand-written pattern recognition (“Eureca”), fuzzy-logic-controlled mobile robots (“Vitaur”) and automotive applications (particularly with Peugeot), vocational guidance systems (“FuzzADe”), and fuzzy information retrieval systems with optimization of weighted queries by genetic algorithms. Motivated by the vision of “The Semantic Web. A new form of Web content that is meaningful to computers will unleash a revolution of new possibilities,” published by Berners-Lee, Hendler, and Lassila in the Scientific American article in 2001 [39], and by Zadeh’s ideas on “Web Intelligence, World Knowledge and Fuzzy Logic” [40,41], Elie started new research interests on fuzzy logic in the Internet, the world wide web [39], and web intelligence [42]. He organized special sessions and workshops on “Fuzzy Logic in the Semantic Web” and published first papers on this new approach [43–48], which were “presented as an extension of the current web in which information is given well-defined meaning, better enabling computers and people to work in cooperation”. Elie
believed in “the positive role fuzzy logic, and more generally soft computing, can play in the development of the Semantic Web, filling a gap and facing a new challenge” [46]. In 2006, he edited the book Fuzzy Logic and the Semantic Web as the first volume in the new book series “Capturing Intelligence” [49]. In his preface to this book, series editor van Harmelen wrote: “The insight that any realistic approach will have to take into account the lessons in a wide community, and slowly but surely, the Semantic Web community is waking up to this fact.” [49, p. vii]. Elie Sanchez was a former President of the International Fuzzy Systems Association (IFSA) and was elected IFSA Fellow. In 1995 he received the International Grigore Moisil Gold Medal and Award for “capital contributions to computer and information sciences, namely in the fields of Fuzzy Systems and Artificial Intelligence”. He was co-editor of ten books and served on the editorial board of 20 international journals, including the present journal entitled Artificial Intelligence in Medicine. On March 10, Professor Takeshi Yamakawa (see Fig. 5) informed the BISC mailing group about the death of Elie Sanchez. Many members of the fuzzy community expressed their condolences. Henri Prade added in his personal note: “It is thanks to Elie that Didier and I [Henri] heard about possibility theory for the first time in 1977.” Lotfi Zadeh wrote in his message that he “saw him in
Fig. 6. Elie Sanchez (middle) with Enric Trillas (left) and José Angel Olivas (right) in Madrid, when Lotfi A. Zadeh was granted to the BBVA Foundation Frontiers of Knowledge Award 2012. Photo: R. Seising.
76
In memoriam / Artificial Intelligence in Medicine 62 (2014) 73–77
Madrid a couple of [months]3 ago. He looked good. His illness was in remission. I feel deeply saddened by his leaving our community.” The last meeting took place when the fifth edition of the BBVA Foundation Frontiers of Knowledge Award 2012 in the category of Information and Communication Technologies (ICT) was given to Lotfi Zadeh “for the invention and development of fuzzy logic”. It was in June 2013 and it was the day I saw Elie for the last time too. After the official part of the celebration I took the photograph that is published here as a last picture (Fig. 6).
Acknowledgments I am very grateful to his widow Mrs. Marie Sanchez, their son Guillaume Sanchez, and their daughter Emilie Sanchez for helping me to write this obituary.
References [1] Sanchez E. Matrices et Fonctions en Logique Symbolique, Thèse de 3e cycle: Mathématiques appliquées: Université Aix-Marseille 2. Faculté des Sciences; 1972. [2] Lakoff George. Hedges: a study in meaning criteria and the logic of fuzzy concepts. In: Peranteau Paul M, Levi, Judith N, Phares Gloria C, editors. Papers from the 8th regional meeting. Chicago, IL: Chicago Linguistic Society; 1972. p. 183–228. [3] Lakoff G. Hedges: a study in meaning criteria and the logic of fuzzy concepts. J Philos Logic 1973;2:458–508. [4] Sanchez E. The robot and the butterfly. In: Seising R, Trillas E, Termini S, Moraga C, editors. On fuzziness. a homage to Lotfi A. Zadeh, vol. II. Berlin (Studies in Fuzziness and Soft Computing, 299): Springer; 2013. p. 625–30. [5] Kaufmann A. Introduction à la théorie des sous-ensembles flous à l’usage des ingénieurs, Vol. 1 Éléments théoriques de base. Paris: Masson; 1977. [6] Sanchez E. Equations de Relations Floues, Thèse Biologie Humaine. Marseille: Faculté de Médecine de Marseille; 1974. [7] Zadeh LA. The concepts of system, aggregate, and state in system theory. In: Zadeh LA, Polak E, editors. System theory. New York, NY: McGraw-Hill; 1969. p. 3–42. [8] Assilian S. Artificial intelligence in the control of real dynamic systems. London: Queen Mary College, University of London; 1974 (August) (Ph.D. Thesis, No. DX193553). [9] Mamdani EH, Assilian S. An experiment in linguistic synthesis with a fuzzy logic controller. Int J Man Mach Stud 1975;7(1):1–13. [10] Sanchez E. Resolution of composite fuzzy relation equations. Inf Control 1976;30(1):38–48. [11] Perez-Ojeda A. Medical knowledge network. A database for computer aided diagnosis. Canada: Department of Industrial Engineering, University of Toronto; 1976 (Master Thesis). [12] Sanchez E. Medical diagnosis and composite fuzzy relations. In: Gupta MM, Ragade RK, Yager RR, editors. Advances in fuzzy set theory and applications. Amsterdam: North-Holland; 1979. p. 437–44. [13] Goguen JA. L-fuzzy sets. J Math Anal Appl 1967;18:145–74. [14] Gupta MM, Rommohan K Ragade, Ronald R Yager, editors. Advances in fuzzy set theory and applications. Amsterdam: North-Holland; 1979. [15] Sanchez E. Compositions of fuzzy relations. In: Gupta MM, Ragade RK, Yager RR, editors. Advances in fuzzy set theory and applications. Amsterdam: NorthHolland; 1979. p. 421–33. [16] Adlassnig K-P. A survey on medical diagnosis and fuzzy subsets. In: Gupta M, Sanchez E, editors. Approximate reasoning in decision analysis. Amsterdam: North-Holland; 1982. p. 203–17, 205. [17] Boegl K, Leitich H, Kolousek G, Rothenfluh ThE, Adlassnig K-P. Knowledge acquisition in MedFrame/Cadiag: a generalized fuzzy approach. In: Cimino JJ, editor. AMIA annual fall symposium. Philadelphia: Hanley & Belfus Inc.; 1996 (Supplement, 833). [18] Adlassnig K-P, Blacky A, Koller W. Fuzzy-based nosocomial infection control. In: Nikravesh M, Kacprzyk J, Zadeh LA, editors. Forging new frontiers: fuzzy pioneers II. Berlin: Springer; 2008. p. 343–50 (Studies in Fuzziness and Soft Computing, 218). [19] Adlassnig K-P, Blacky A, Mandl H, Rappelsberger A, Koller W. Fuzziness in Healthcare-Associated Infection Monitoring and Surveillance. In: Winston A, Melek W, Hall P, McKenzie S, Gibbs M, Adamson G, editors. Proceedings of the 2014 IEEE conference on Norbert Wiener in the 21st century—driving technology’s future. Piscataway: Institute of Electrical and Electronics Engineers (IEEE); 2014 (86.pdf.).
3
Lotfi Zadeh wrote “years” but it was actually months.
[20] Gupta MM, Sanchez E, editors. Fuzzy information and decision processes. Amsterdam: North-Holland; 1982. [21] Gupta MM, Sanchez E, editors. Approximate reasoning in decision analysis. Amsterdam: North-Holland; 1982. [22] Sanchez E, editor. Fuzzy information, knowledge representation and decision analysis: proceedings of the International Federation of Automatic Control (IFAC) symposium. Oxford: Pergamon Press; 1983. [23] Sanchez E, Zadeh LA, editors. Approximate reasoning in intelligent systems, decision & control. Elmsford: Pergamon Press; 1987. [24] Di Nola A, Pedrycz W, Sanchez E, Sessa S. Fuzzy relation equations and their applications to knowledge engineering. Dordrecht: Kluwer Academic Publishers; 1989. [25] Sanchez E, Shibata T, Zadeh LA. Genetic algorithms and fuzzy logic systems: soft computing perspectives. Singapore: World Scientific Publishing Company; 1997. [26] Zadeh LA. Foreword. Appl Soft Comput 2001;1(1):1–2. [27] Sanchez E. Importance in knowledge systems. Inf Syst 1989;14(6): 455–64. [28] Sanchez E. On possibility qualification in natural languages. Inf Sci 1978;15(1):45–76. [29] Sanchez E. Non standard fuzzy arithmetic. In: Wang PZ, Loe KF, editors. Between mind and computer: fuzzy science and engineering. Singapore: World Scientific Publishing Company; 1994. p. 271–82. [30] Avrachenkov KE, Sanchez E. Fuzzy markov chains and decision-making. Fuzzy Optim Decis Making 2002;1(2):143–59. [31] Sanchez E. Eigen fuzzy sets and fuzzy relations. J Math Anal Appl 1981;81(2):399–421. [32] Di Nola A, Pedrycz W, Sessa S, Sanchez E. Fuzzy relation equations theory as a basis of fuzzy modelling: an overview. Fuzzy Sets Syst 1991;40(3): 415–29. [33] Sanchez E. Truth-qualification and fuzzy relations in natural languages, application to medical diagnosis. Fuzzy Sets Syst 1996;84(2):155–67. [34] Bartolin R, Bonniol V, Sanchez E. Inflammatory protein variations: medical knowledge representation and approximate reasoning. In: Bouchon B, Saitta L, Yager RR, editors. Uncertainty and Intelligent Systems—Proceedings of the 2nd international conference on information processing and management of uncertainty in knowledge-based systems (IPMU’88). Berlin: Springer; 1988. p. 306–13. [35] Sanchez E, Pierre Ph. Intelligent decision making systems: from medical diagnosis to vocational guidance. In: Szczepaniak PS, Lisboa PJG, Kacprzyk J, editors. Fuzzy systems in medicine. Heidelberg: Physica-Verlag; 2000. p. 204–23. [36] Sanchez E. Soft computing perspectives. In: Proceedings of the twenty-forth international symposium on multiple-valued logic. Boston: IEEE; 1994. p. 276–81. [37] Sanchez E. DNA biosoft computing. In: Yamakawa T, Matsumoto G, editors. Methodologies for the conception, design, and application of intelligent systems. Proceedings of the fourth international conference on soft computing. Singapore: World Scientific Publishing Company; 1996. p. 30–7. [38] Sanchez E. Fuzzy logic knowledge systems and artificial neural networks in medicine and biology. In: Yager RR, Zadeh LA, editors. An introduction to fuzzy logic applications in intelligent systems. Boston, MA: Kluwer Academic Publishers; 1992. p. 235–51. [39] Berners-Lee T, Hendler J, Lassila O. The Semantic Web. A new form of web content that is meaningful to computers will unleash a revolution of new possibilities. Sci Am 2001;284((5) May):34–43. [40] Zadeh LA. Web intelligence, world knowledge and fuzzy logic—the concept of web IQ (WIQ). In: Negoita MGh, Howlett RJ, Jain LC, editors. Knowledgebased intelligent information and engineering systems, Proceedings of the 8th International Conference, KES 2004, September 20–25, 2004. Wellington, New Zealand, Part I: Springer, Berlin; 2004. p. 1–5. [41] Zadeh LA. A note on web intelligence, world knowledge and fuzzy logic. Data & Knowledge Engineering—Special jubilee issue: DKE 50 2004;50(3): 291–304. [42] Sanchez E. Fuzzy logic e-motion. In: Nikravesh M, Azvine B, editors. FLINT 2001—New directions in enhancing the power of internet. Proceedings of the 2001 BISC international workshop on fuzzy logic and the internet, Berkeley, California UC Berkeley Electronics Research Laboratory Memorandum No. UCB/ERL M01/28. 2001. p. 10–4. [43] IPMU. Special session “Fuzzy logic in the Semantic Web: a new challenge”. In: Bouchon-Meunier B, Coletti G, Yager RR, editors. Proceedings of the 10th international conference on information processing and management of uncertainty in knowledge-based systems (IPMU’04). Perugia, Italy: Casa Editrice Universita’ la Sapienza, Roma; 2004. p. 1017–38. [44] Fuzzy Logic and the Semantic Web. Workshop Marseille, France 2005. Extended abstracts available at: http://www.lif.univ-mrs.fr/FLSW. [45] FLSW-II. Fuzzy logic and the Semantic Web, second workshop abstracts. In: Bouchon-Meunier B, Yager RR, editors. Proceedings of the 11th international conference on information processing and management of uncertainty in knowledge-based systems (IPMU’2006). Paris, France: EDK; 2006. [46] Sanchez E, Yamanoi T. D’Amico fuzzy ontologies for the Semantic Web. In: Larsen HL, Pasi G, Ortiz-Arroyo T, Andreasen T, Christiansen H, editors. Flexible query answering systems, proceedings of the 7th International conference FQAS 2006. Berlin: Springer; 2006. p. 691–9.
In memoriam / Artificial Intelligence in Medicine 62 (2014) 73–77 [47] Calegari S, Sanchez E. A fuzzy ontology-approach to improve semantic information retrieval. In: Bobillo F, da Costa PCG, D’Amato C, Fanizzi N, Fung F, Lukasiewicz T, et al., editors. Proceedings of the Third ISWC Workshop on uncertainty reasoning for the Semantic Web (URSW’07), CEUR-Workshop Proceedings, Vol. 327. Nov. 12, 2007, Busan, Korea; 2007. Available at: http://ceur-ws.org/Vol-327/. [48] Calegari S, Sanchez E. Object-fuzzy concept network: an enrichment of ontologies in semantic information retrieval. J Assoc Info Sci Technol 2008;59(13):2171–80. [49] Sanchez E, editor. Fuzzy logic and the Semantic Web. Amsterdam: Elsevier; 2006 (Capturing Intelligence 1).
77
Rudolf Seising ∗ European Centre for Soft Computing, Edificio de Investigación, Calle Gonzalo Gutiérrez Quirós S/N, 33600 Mieres, Asturias, Spain ∗ Tel.: +34 985456545; fax: +34 985456699. E-mail address: [email protected]
2 September 2014