============================= VOLUME 18 NUMBER 2 JUNE 1994 ============================= 1. PROFILES: B. SOUCEK ---------------------- 2. NEUROLOGICAL DIAGNOSES BASED ON EVOKED BRAIN WINDOWS AND ON HOLOGRAPHIC LEARNING B. Soucek, STAR SERVICE S.p.A., Via Amandola 162/1, 70126 Bari, Italy Phone: 3980.5484555, Fax: 3980.5484556 pp. 109-114 Keywords: Brain-windows, evoked potentials, holographic learning, diagnosis, neurology Abstract: The evoked potentials have been generated in response to auditory stimuli to a person, and light stimuli to insects, resulting in two datasets, HUMAN and INSECT. In both datasets the responses are composed of several peaks with variable latencies. The brain-window logic is used to explain the evoked responses. Brain-windows are generated through mutual coupling of biological oscillators, and modulated by the memory that stores the past history and the present behavior. Latencies of the peaks provide necessary information to discriminate between normal subject and pathological states resulting from injury, tumor or multiple sclerosis. The holographic neural network classifies the subjects, based on the peak latencies. Combining brain-window theory with the holographic learning opens new possibilities for neurological diagnoses, as well as for a new kind of fuzzy neural networks. ---------------------- 3. APPROXIMATING KNOWLEDGE IN A MULTI-AGENT SYSTEM M. Kubat, S. Parsons, Ludwig-Boltzmann Institute of Medical Informatics and Neuroinformatics, Department of Medical Informatics, Institute of Biomedical Engineering, Graz University of Technology, Brockmanngasse 41, A-8010 Graz, Austria, email mirek@dpmi.tu-graz.ac.at Advanced Computation Laboratory, Imperial Cancer Research Fund, P.O. Box 123, Lincoln's Inn Fields, London WC2A 3PX, United Kingdom, email sp@acl.lif.icnet.uk Department of Electronic Engineering, Queen Mary and Westfield College, Mile End Road, London E1 4NS, United Kingdom. pp. 115-129 Keywords: Artificial Intelligence, abstraction, granularity of knowledge, dl-cut, rough concepts Abstract: This paper is concerned with establishing a common language that can be used to communicate between the different members of a multi-agent system. We suggest that this may be done by successively approximating the concepts that each agent in the system deals with, and the paper gives algorithms which make this possible. Along the way we introduce the notion of a description language cut, or dl-cut, which is an abstraction to which a rich class of languages may be mapped. The idea of a dl-cut is then used to introduce rough concepts - rough descriptions of the concepts used by the agents. Finally we discuss the way in which rough concepts can be logically combined and used in deductive reasoning, also debating the scope of the validity of inferences using the concepts. ---------------------- 4. THE THEORY OF DYNAMIC CONCEPTUAL MAPPINGS AND ITS SIGNIFICANCE FOR EDUCATION, COGNITIVE SCIENCE, AND ARTIFICIAL INTELLIGENCE V. A. Fomichov, O. S. Fomichova, Moscow State Institute of Electronics and Mathematics (Technical University), Moscow State University, Bolshoj Vuzovsky pereulok, 3/12, 109028 Moscow, Russia, root@onti.miem.msk.su Moscow Children and Teenagers Palace for Creative Work, Russia pp. 131-148 Keywords: artificial intelligence, cognitive science, theory of teaching, dynamic conceptual mapping, teaching children foreign languages, emotionally-imaginative teaching English, intelligent tutoring system, knowledge archives Abstract: Important advancements in the theory of teaching and practical teaching young children foreign languages are described and discussed in a broad context of finding the most effective ways of conveying information in various areas of human activity. It is reported about the development of a new theory of teaching called the theory of dynamic conceptual mappings (the DCM-theory). The basic principles of the DCM-theory are set forth, and its composition is shortly described. The DCM-theory may be characterized as a theory of bridging gaps between conceptual systems of a teacher and a learner. It became the ground for creating new, highly effective methods of teaching young children and teenagers to read and communicate in English. These methods are called the methods of emotionally-imaginative teaching (the EIT-methods) and are based on ideas of artificial intelligence and cognitive science. ---------------------- 5. INFORMATIONAL BEING-IN A. P. Zeleznikar, Volariceva ulica 8, 61111 Ljubljana, Slovenia, anton.p.zeleznikar@ijs.si pp. 149-173 Keywords: abduction, Being-in, Being-in-the-world, circularity, decomposition, deduction, externalism, informational includedness (involvement, embedding), induction, inference, informational modi (ponens, tollens, rectus, obliquus), internalism, metaphysicalism, parallelism, phenomenalism, reasoning, serialism Abstract: In this paper the phenomenon of informational Being-in, that is, includedness is studied in a formally recursive (informational) way, dealing with basic definitions of includedness (informational in-volvement, em-bedding) and their consequences. It seems that the informational includedness is a phenomenon of informational entities, which involves them in a perplexedly recursive way and offers the richness of the informationally spontaneous parallelism, serialism, and circularity. In this respect, together with its informational openness and recursiveness, informational Being-in can come semantically as close as possible to its philosophical notion (concept). Some includable structured phenomena of inference or reasoning (deduction, induction, abduction, modus ponens, tollens, rectus, and obliquus) are shown in a formal manner. The disposed formal apparatus enables an unbounded and even deepened philosophical investigation of the phenomenon of Being-in and its consequences. So, a formalistic investigation of informational Being-in can enrich its philosophical understanding. ---------------------- 6. GRAPHS AND THE THIRD NORMAL FORM FOR RELATIONAL DATABASE J. Nemec, J. Grad, University of Maribor, College of Agriculture, 62000 Maribor, Vrbanska 30, Slovenia, Phone: +386 62 226 611, Fax: +386 62 23 363, E-mail: joze.nemec@uni-mb.si University of Ljubljana, Faculty of Economics, 61109 Ljubljana, Slovenia, Phone: +386 61 1683 333, Fax +386 61 301 110 pp. 175-182 Keywords: relational database, relations, normal forms, normalization process, DB graphs, matrices Abstract: Determination of higher order normal forms for a relational database (RDB) is frequently a time-consuming process. We can solve this problem by applying graph theory. In the paper the necessary characteristics of graphs that represent a given RDB are analyzed. The connectivity matrices for these graphs and the properties they must satisfy are also introduced and discussed. The established RDB graphs and the corresponding relationship matrices form an important basis of the algorithm for designing RDB with no redundant relations. ---------------------- 7. ON BAYESIAN NEURAL NETWORKS I. Kononenko, University of Ljubljana, Faculty of electrical engineering and computer science, Trzaska 25, SI-61001 Ljubljana, Slovenia, Phone: +386 61 1231121, Fax: +386 61 264990, e-mail: igor.kononenko@ninurta.fer.uni-lj.si pp. 183-195 Keywords: Bayesian neural network, Hopfield's neural network, naive Bayesian classifier, continuous neural network, probability, entropy, machine learning, artificial intelligence, overview In the paper the contribution of the work on Bayesian neural networks is discussed with respect to previous, current, and potential future research in machine learning. The discrete and the continuous Bayesian neural network model is compared with Hopfield's models. It is shown that the Bayesian neural network's equations are analogous to equations used to describe Hopfield's model. Two different models of the Bayesian neural network are compared with Hopfield's model, one based on Shannon's entropy (probability) and the other based on Good's plausibility (odds). A generalization of the naive-Bayesian classifier is described that enable the basic algorithm to detect the dependencies between neurons. ---------------------- 8. LECTURE NOTES IN MACHINE LEARNING Xindong Wu, Department of Computer Science, James Cook University, Townsville, QLD 4811, Australia, Email: xindong@cs.jcu.edu.au pp. 197-218 Keywords: Artificial intelligence, machine learning, symbolic approaches, educational paper Abstract: Machine learning is a major area in artificial intelligence (AI), and has seen sustained research and a growing presence in lecture syllabuses over recent years. It has been commonly recognized as a feasible solution to the so called knowledge bottleneck problem in transforming knowledge from human experts to knowledge-based systems. Also, as learning is the essence of human intelligence, only when we have computer systems that can learn can we have real AI. Researchers have devised quite a few sound and efficient learning algorithms (such as ID3 and HCV); a number of universities have opened machine learning courses in their AI-related undergraduate and/or Master of Science programs. This document contains a compressed set of lecture notes designed for the Machine Learning module in our Advanced Artificial Intelligence course at James Cook University. They are biased towards a basic exposition of the practical symbolic approaches only. ---------------------- 9. COMPILER DETECTION OF FUNCTION CALL SIDE EFFECTS D. A. Spuler, A. S. M. Sajeev, Department of Computer Science, James Cook University, Townsville, QLD 4811, Australia, Department of Software Development, Monash University, Caulfield East, VIC 3145, Australia pp. 219-227 Keywords: Side effect, function calls, static analysis Abstract: The determination of whether an operation in a procedural language such as Pascal causes side effects has important applications in static error detection, program optimization, program verification and other software tool issues. For simple operations, such as assignment, there is obviously a side effect. However, to detect whether a function call causes a side effect, the body of the function must be analyzed. A function call can produce a side effect in a number of ways, such as: modifying a global variable, modifying an internal static variable, modifying one of its pass-by-reference arguments, producing output, consuming input etc. This paper explains the algorithms and data structures used to determine if a function call causes a side effect. ---------------------- 10. CONTROL ABSTRACTIONS IN MODULA-2: A CASE STUDY USING ADVANCED BACKTRACKING Dipartimento di Elettronica, Informatica e Sistemistica, Universita' della Calabria, I-87036 Rende(CS) - Italy, E-mail: nigro@ccusc1.unical.it pp. 229-243 Keywords: Control abstractions, Modula-2, reusable modules, backtracking, simulation Abstract: This paper shows that Modula-2, extended with a general control abstraction called a thread, supports the construction of programmer-defined control modules. As an example, a realistic control regime providing advanced backtracking is presented.