Computing Journal

Міжнародний науковий журнал "Комп'ютинг"

Науково-дослідний інститут Інтелектуальних комп'ютерних систем

Тернопільський Національний Економічний Університет

2004, Том 3, Випуск 1

Зміст і резюме

V. Golovko. Editorial.
Kurosh Madani. Industrial Applications of Artificial Neural Networks.
A. Brückmann, F. Klefenz, A. Wünsche. A Neural Net For 2d-Slope And Sinusoidal Shape Detection.
Mahinda Pathegama, Özdemir Göl. An Artificial Neural Process To Create Continuous Object Boundaries In Medical Image Analysis.
Akira Imada. How a Peak Can be Searched for in an Almost Everywher Flatland of Altitude Zero? — Tiny Flat Island in Huge Lake.
Helmut A. Mayer. Ontogenetic Teaching Of Mobile Autonomous Robots With Dynamic Neurocontrollers.
Ruslan R. Zholtikov, Mikhail M.Tatur. Some Models Of Raster Correlators Of Binary Images.
Khalid Saeed, Marek Tabędzki. A New Hybrid System for Recognition of Handwritten-Scrip.
Jean-Jacques Mariage. Learning To Teach To Neural Networks How To Learn Well With Soh, A Self-Observing Heuristic.
Lamine Thiaw, Mariusz Rybnik, Rachid Malti, Abdennasser Chebira, Kurosh Madani. A Comparative Study Between A Multi-Models Based Approach And An Artificial Neural Network Based Technique For Nonlinear Systems Identification.
Petrovsky A.A., Likhachov D.S., W.Wan. An Anthropomorphic Speech Processing Based On The Cochlear Model And Its Application For Coding Task.
Qiangfu Zhao. Learning And Understanding Based On Neural Network Trees.
Vladimir Golovko, Yury Savitsky. Computing Of Lyapunov Exponents Techniques Using Neural Networks.
Leonid Makhnist, Nikolaj Maniakov, Vladimir Rubanov. Some Methods Of Adaptive Multilayer Neural Networks Training.
A. Reznik, R.Kh. Sadykhov. The Steganographic System For Hidden Transfer Of The Color Images.

Примітка: прізвища та координати авторів подано мовою оригіналу, а назви та резюме – українською мовою.

Editorial
Introduction to the special issue on
“Neural Networks and Artificial Intelligence 2003”
Guest Editors Vladimir Golovko

This special issue is collection of recent contributions in theory and applications presented at the International Conference on Neural Networks and Artificial Intelligence (ICNNAI)’2003, held in Minsk (Belarus). The key aims of the ICNNAI 2003 were to present and discuss with the researchers from various countries scientific results and their application in the broad field of neural computation and artificial intelligence, as well to review our past and to define new perspectives. The venue this year was the beautiful city of Minsk – capital of Belarus. The conference was made in the Belarus State University of Informatics and Radioelectronics – one of the leading Universities in Belarus. During this meeting the researchers from different countries had opportunity of discussing the theoretical foundations and the practical using of neural technologies and artificial intelligence. The various social events enhanced the technical discussions that greatly contributed to the success of the conference and to making for further dialogue between researches. The program structure of ICNNAI’2003 was organized under the following topic areas: artificial intelligence, neural networks architecture and learning algorithms, pattern recognition and image processing, signal processing, data analysis and classification, neural networks in technical systems and applications. The following series of papers was selected for presentation in this journal.

Kurosh Madani: “Industrial applications of artificial neural networks” describes real application capability of main ANN models and based techniques in real world industrial tasks. Inspired from biological nervous systems and brain structure, these models take advantage from their learning and generalization capabilities, overcoming difficulties and limitations related to conventional techniques. Several examples, namely intelligent adaptive control, IBM neuro-processor for image processing, yield prediction in VLSI industry have been presented and discussed.

Andreas Brueckman, Frank Klefenz, Andreas Wuensche: “A neural net for 2D-slope and sinusoidal shape detection”, proposes a neural network approach, which is able to train a set of different slopes or a set of sinusoids of different frequencies and to detect test patterns after the training stage. Unsupervised learning with a Boltzmann temperature is assumed. The weight settings are either analytically derived by the Hough transform equations or are self-learned by neural network.

Mahinda Pathegama, Oezdemir Goel: “An Artificial neural process to create continuous object boundaries in medical image analysis” presents a novel edge-linking technique based on an artificial neural process, which uses directional sensitivity derivatives from an edged image. The proposed edge-linking technique, implemented as an image-processing algorithm for direction-sensitive selectiveness, provides an effective solution to the problem of porous boundaries encountered in biological cell image analysis.

Akira Imada: “How a peak can be searched for in an almost everywher flatland of altitude zero? — Tiny flatisland in huge lake”, explores a weight configuration space searching for solutions to make a neural network with spiking neurons do some tasks. The weight configuration we already knew for the task of associative memory is found to be like a tiny-flat-island-in-a-huge-lake. In short, this is a problem of finding so called a needle in a haystack, and author calls for proposals how we find a solution to it.

Helmut Mayer: “Ontogenetic teaching of mobile autonomous robots with dynamics neurocontrollers” presents experiments employing a standard sensor–motor neurocontroller with self–adapting weights. The focus of investigations is put on the mechanisms of the interaction of teaching input and structural changes. A well–known concept for this interaction is Hebbian learning, which is regulated by artificial neuromodulators (ANMs) in the presented approach. The results show that ontogenetic learning of mobile autonomous robots with neurocontrollers regulated by external feedback mediated by ANMs is sufficient to teach robots simple tasks.

Ruslan Zoltikov, Michail Tatur: “Some models of raster correlators of binary images”, addresses the problem of statistical recognition of binary images. The main hypothesis is that the pixels located on object boundary practically do not carry the information on an image. The results of experiments are discussed.

Khalid Saeed, Marek Tabedzki: “A new hybrid system for recognition of Handwritten script”, describes a new approach for capital Latin-letter classification and recognition. This approach is based on multilauer perceptron and algorithm of minimal eigenvalues of Toeplitz matrices. The obtained results are discussed.

Jean-Jacques Mariage: “Learning to teach to neural networks. How to learn well with SOH, a self-observing heuristic”, presents an adaptive learning approach based on neo-Darwinian evolution of neural units. First of all author examines the main properties of SOM algorithm and its evolutionary growing variants. In the second part a self –observing heuristic as a minimal system capable of adaptive learning are proposed. Finally the minimal set of properties in order to obtain the emergence of Darwinian evolution among elementary constituents is extracted.

Lamine Thiaw, Mariusz Rybnik, Rachid Malti, Abdennasser Chebira, Kurosh Madani: “A comparative study between a multi-models based approach and an artificial neural networks based technique for nonlinear system identification”, presents a comparative study between a conventional multi-model architecture and an ANN based multi-model structure, in the frame of nonlinear system identification. The validation has been performed on an ARMA based generated model. The experimental results show that the generalization capability of neural network is better in comparison with the multi-model.

Alexander Petrovsky, Denis Likhachev, Wanggen Wan: ”An anthropomorphic speech processing based on the cochlear model and its application for coding tasks” presents the new mathematical model of cochlea, which is transformed into a digital form using bilinear transformation. The model looks much simpler structure and comes to be a typical bandpass filter. The amplitude frequency response of the model is quite consistent with the experimental data.

Qiangfu Zhao: “Learning and understanding based on neural network trees”, describes a hybrid learning model called neural network tree (NNTree). An NNTree is a decision tree (DT) with each non-terminal node containing an expert neural network (ENN). The obtained results have confirmed that the NNTrees are suitable both for incremental learning and for understanding.

Vladimir Golovko, Jury Savitsky: “Computing of Lyapunov exponents techniques using neural networks”, discusses the use of neural networks for computing of Lyapunov spectrum using observations from unknown dynamical system. Such an approach is based on applying of multilayer perceptron (MLP) for forecasting the next state of dynamical system from the previous one. It allows for evaluating the Lyapunov spectrum of unknown dynamical system accurately and efficiently only by using scalar time series. The results of experiments are discussed.

Leonid Makhnist, Nikolaj Maniakov, Vladimir Rubanov: “Some methods of adaptive multiplayer neural networks training”, proposes the two approaches for training of multiplayer perceptron. It is based on the gradient descent method. As a result of applying this method the equations for computing of the adaptive training step were obtained.

Finally, Ivan Reznik and Rauf Sadykhov in “The steganographic system for hidden transfer of the colour images” present the approach for embedding hidden image in the container image. It is based on the modified model of the two-dimensional spatial correlator. The problems of efficiency, robustness, accuracy and performance of the proposed approach are considered.