|
Research Institute of Intelligent Computer Systems Ternopil National Economic University |
|
2006, Vol. 5, Issue 3 |
Contents and abstracts
PREFACE Guest Editors: Vladimir Glovko 1), Robert Hiromoto 2), Akira Imada 3), Kurosh Madani 4) 1) Brest State Technical University,
Moskovskaja str. 267, 224017 Brest, Belarus,gva@bstu.by ARTIFICIAL INTELLIGENCE AND NEURAL NETWORKS THE LEGACY OF ALAN TURING AND JOHN VON NEUMANN Heinz Muhlenbein Fraunhofer Institut Autonomous intelligent Systems Schloss Birlinghoven 53757 Sankt Augustin, Germany heinz.muehlenbein@online.de, http://www.ais.fraunhofer.de/~muehlen The work of Alan Turing and John von Neumann on machine intelligence and artificial automata is reviewed. Turing's proposal to create a child machine with the ability to learn is discussed. Von Neumann had doubts that with teacher based learning it will be possible to create artificial intelligence. He concentrated his research on the issue of complication, probabilistic logic, and self-reproducing automata. The problem of creating artificial intelligence is far from being solved. In the last sections of the paper I review the state of the art in probabilistic logic, complexity research, and transfer learning. These topics have been identified as essential components of artificial intelligence by Turing and von Neumann. PREDICTION OF FATIGUE CRACK GROWTH PROCESS VIA ARTIFICIAL NEURAL NETWORK TECHNIQUE Konstantin N. Nechval 1), Nicholas A. Nechval 2), Irina Bausova 3), Daina Skiltere 3), Vladimir F. Strelchonok 4) 1) Applied Mathematics Dept, Transport
Institute, Lomonosov Street 1, LV-1019, Riga, Latvia, e-mail:
konstan@tsi.lv Failure analysis and prevention are important to all of the engineering disciplines, especially for the aerospace industry. Aircraft accidents are remembered by the public because of the unusually high loss of life and broad extent of damage. In this paper, the artificial neural network (ANN) technique for the data processing of on-line fatigue crack growth monitoring is proposed after analyzing the general technique for fatigue crack growth data. A model for predicting the fatigue crack growth by ANN is presented, which does not need all kinds of materials and environment parameters, and only needs to measure the relation between a (length of crack) and N (cyclic times of loading) in-service. The feasibility of this model was verified by some examples. It makes up the inadequacy of data processing for current technique and on-line monitoring. Hence it has definite realistic meaning for engineering application. INDUSTRIAL APPLICATIONS OF COMPUTATIONAL INTELLIGENCE Colin M. Frayn CERCIA, School of Computer Science, University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK. cmf@cercia.ac.uk, http://www.cs.bham.ac.uk/~cmf/ In this paper, I cover in more detail two specific applications where Computational Intelligence systems have been used in industry. In particular, I consider the problems of path optimization through an inhomogeneous road network, and data analysis for financial applications. This paper will deal with several important questions about the applicability of those techniques in real-world scenarios, and will show how some of these issues have been directly addressed in order to create value for our business partners. COMPUTER AIDED DIAGNOSIS USING SOFT-COMPUTING TECHNIQUES AND IMAGE’S ISSUED REPRESENTATION: APPLICATION TO BIOMEDICAL AND INDUSTRIAL FIELDS Kurosh Madani 1), Matthieu Voiry 1) 2), Veronique Amarger 1), Nadia Kanaoui 1), Amine Chohra 1), Francois Houbre 2) 1) Images, Signals and Intelligent Systems
Laboratory (LISSI / EA 3956), Senart Institute of Technology, University
PARIS XII, Av. Pierre Point, F-77127 Lieusaint, France {madani, amarger,
chohra, kanaoui}@univ-paris12.fr, http://www.univ-paris12.fr/ It is interesting to notice that from “problem’s formulation” point of view “Industrial Computer Aided Diagnosis” and “Biomedical Computer Aided Diagnosis” could be formulated as a same diagnosis riddle: “How point out a correct diagnosis from a set of symptoms?”. The only difference between the two above-mentioned groups of problems is the nature of the monitored (diagnosed) system: in the first group the monitored system is an artificial machinery (plant, industrial process, etc…), while in the second, the monitored system is a living body (animal or human).One of the most appealing classes of approaches allowing handling the Computer Aided Diagnosis Systems’ design in the frame of the aforementioned dual point of view is Soft-Computing based techniques, especially those dealing with neural networks and fuzzy logic. In this article, we present two soft-computing based approaches dealing with CADS design. One aims designing a biomedical oriented CADS and the other sets sights on conceiving a CADS to overcome a real-world industrial quality control dilemma. The goal of the first system is to diagnose the human’s auditory pathway’s health. The target of the second is to detect and diagnose the high tech optical devices’ defects. HOLISTIC SELF-REPROGRAMMING OF NEURAL NETWORKS: BETWEEN SELF-ORGANIZATION AND SELF-OBSERVATION Jean-Jacques Mariage CSAR research group, AI Laboratory, Paris 8 University 2 rue de la Liberte, St Denis, Cdx 93526, France jam@ai.univ-paris8.fr Neural networks (NNs) are inspired – at least metaphorically –from biological solutions nature selected by evolution. On one hand, learning algorithms' efficacy has been widely demonstrated experimentally, even if the mathematical proof of their convergence is not always very easy to establish (SOM). On the other hand, biological mechanisms like brain wiring or embryology remain partly understood and how life or the bases of consciousness are related to the underlying biological substrate remains a total mystery. The same goes for memory. We don’t really know how information is stored in and recovered from biological neural structures. We therein paradoxically use complex systems, the hard core of which we still don't always fully understand, both regarding the models we build, as well as their former roots in the leaving world. In this theoretical paper, we resort to a few biological encoding schemata that bring insights into neural structures' growth, plasticity and reorganization, and we suggest reconsidering the metaphor in an adaptive developmental view. LEARNING FROM THE ENVIRONMENT WITH A UNIVERSAL REINFORCEMENT FUNCTION Diego Ariel Bendersky, Juan Miguel Santos Departamento de Computacion, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires. Pabellon I, Ciudad Universitaria 1428 Ciudad de Buenos Aires, Argentina. {dbenders, jmsantos}@dc.uba.ar Traditionally, in Reinforcement Learning, the specification of the task is contained in the reinforcement function (RF), and each new task requires the definition of a new RF. But in the nature, explicit reward signals are limited, and the characteristics of the environment affects not only “how” animals perform particular tasks, but also “what” skills an animal will develop during its life. In this work, we propose a novel use of Reinforcement Learning that consists in the learning of different abilities or skills, based on the characteristics of the environment, using a fixed and universal reinforcement function. We also show a method to build a RF for a skill using information from the optimal policy learned in a particular environment and we prove that this method is correct, i.e., the RF constructed in this way produces the same optimal policy. A MODULAR NEUROCONTROLLER FOR A SENSOR-DRIVEN REACTIVE BEHAVIOR OF BIOLOGICALLY INSPIRED WALKING MACHINES Poramate Manoonpong 1, 2), Frank Pasemann 1), Hubert Roth 3) 1) Fraunhofer Institut fur Autonome
Intelligente Systeme (AIS), Sankt Augustin, Germany
frank.pasemann@ais.fraunhofer.de, http://www.ais.fraunhofer.de/INDY In this article, a modular neurocontroller is presented. It has the capability to generate a reactive behavior of walking machines. The neurocontroller is formed on the basis of a modular structure. It consists of the three different functionality modules: neural preprocessing, a neural oscillator network and velocity regulating networks. Neural preprocessing is for sensory signal processing. The neural oscillator network, based on a central pattern generator, generates the rhythmic movement for basic locomotion of the walking machines while the velocity regulating networks change the walking directions of the machines with respect to the sensory inputs. As a result, this neurocontroller enables the machines to explore in- and out-door environments by avoiding obstacles and escaping from corners or deadlock situations. It was firstly developed and tested on a physical simulation environment, and then was successfully transferred to the six-legged walking machine AMOS-WD06. INFORMATION-BASED ALGORITHMIC DESIGN OF A NEURAL NETWORK CLASSIFIER Robert E. Hiromoto 1), Milos Manic 2) 1) University of Idaho, Moscow, Idaho
83844-1010, USA, hiromoto@cs.uidaho.edu An information-based design principle is presented that provides a framework for the design of both parallel and sequential algorithms. In this presentation, the notion of information (data) organization and canonical separation are examined and used in the design of an iterative line method for pattern grouping. In addition this technique is compared to the Winner Take All (WTA) method and shown to have many advantages. HANDWRITTEN SCRIPT AND WORD RECOGNITION – A VIEW-BASED APPROACH Marek Tabedzki, Khalid Saeed Faculty of Computer Science, Bialystok Technical University Wiejska 45A, 15-351 Bialystok, Poland {abedzki, aida}@ii.pb.bialystok.pl, http://aragorn.pb.bialystok.pl/~zspinfo/ This paper presents a hybrid system for character and word recognition. It is based on a modification to the view-based approach presented in authors’ previous works. The algorithm is appropriate for dealing with whole, unsegmented words or isolated characters. The characteristic vectors taken from views of the tested image are processed with the method of minimal eigenvalues of Toeplitz matrices. The obtained series of minimal eigenvalues are used for classification with Artificial Neural Networks. The results of the experiments on different sets of words and letters are presented. THE DEFECT AND PROJECT RULES INSPECTION ON PCB LAYOUT IMAGE Alexander Doudkin, Alexander Inyutin United Institute of Informatics Problems, National Academy of Science of Belarus, 6 Surganov st., Minsk, 220012, doudkin@newman.bas-net.by, avin@lsi.bas-net.by, www.lsi.bas-net.by A technique of PCB layout optical inspection based on image comparison and mathematical morphology methods is proposed. The unique feature of the technique is that the inspection is performed at different stages of image processing. The presence of all layout elements is checked up, then positions of found elements and their conformity to project rules are verified, the breakouts and shorts are found. The inspection of mousebits, spur and pinholes on conductors is also carried out. ESTIMATION OF THE CROSS CORRELATION BASED OPTICAL FLOW FOR VIDEO SURVEILLANCE Rauf Sadykhov 1), Denis Lamovsky 2) 1) United Institute of Informatics Problems
National Academy of Science of Belarus, The Laboratory of System
Identification, 6, Surganov st., Minsk, 220012, Belarus, URL:
http://lsi.bas-net.by This paper describes a new algorithm to calculate cross-correlation function. We combined box filtering technique for calculation of cross correlation coefficients with parallel processing using MMX/SSE technology of modern general purpose processors. We have used this algorithm for real time optical flow estimation between frames of video sequence. Our algorithm was tested on real world video sequences obtained from the cameras of video surveillance system. NEURAL NETWORK APPROACHES FOR INTRUSION DETECTION AND RECOGNITION Vladimir Golovko, Leanid Vaitsekhovich Brest State Technical University, Moskovskaja str. 267, 224017 Brest, Belarus gva@bstu.by, vspika@rambler.ru Most current Intrusion Detection Systems (IDS) examine all data features to detect intrusion. Also existing intrusion detection approaches have some limitations, namely impossibility to process large number of audit data for real-time operation, low detection and recognition accuracy. To overcome these limitations, we apply modular neural network models to detect and recognize attacks in computer networks. It is based on combination of principal component analysis (PCA) neural networks and multilayer perceptrons (MLP). PCA networks are employed for important data extraction and to reduce high dimensional data vectors. We present two PCA neural networks for feature extraction: linear PCA (LPCA) and nonlinear PCA (NPCA). MLP is employed to detect and recognize attacks using feature-extracted data instead of original data. The proposed approaches are tested using KDD-99 dataset. The experimental results demonstrate that the designed models are promising in terms of accuracy and computational time for real world intrusion detection. HOW MANY PARACHUTISTS WILL BE NECESSARY TO FIND A NEEDLE IN A PASTORAL - WHO IS A LUCKY ONE? Akira Imada Brest State Technical University, Moskowskaja 267 Brest 224017 Belarus, akira@bstu.by, http://neuro.bstu.by/ai/akira.html This article is a consideration on computer network intrusion detection using artificial neural networks, or whatever else using machine learning techniques. We assume an intrusion to a network is like a needle in a haystack not like a family of iris flower, and we consider how an attack can be detected by an intelligent way, if any. SIMULATION MODELING OF INTERPLANETARY SHOCKS ARRIVAL TIME PREDICTION ON HISTORICAL DATA SET Volodymyr Turchenko, Viktor Demchuk, Anatoly Sachenko Research Institute of Intelligent Computer Systems, Ternopil National Economic University, Peremoga Square 3, 46004, Ternopil, Ukraine, {vtu, vde, as}@tanet.edu.te.ua An approach to prediction of the arrival time of interplanetary shocks using neural networks based on the data gathered from single EPAM (Electron, Proton and Alpha Monitor) channel of NASA’s ACE (Advanced Composition Explorer) spacecraft is proposed in this paper. A short description of ACE spacecraft and the data, published online on the appropriate web-site, are considered. A data choice to fulfill a prediction of interplanetary shocks is proven and structure of neural network is described. The results of simulation modeling in MATLAB are considered in the end of the paper. |