1. A. Imada, V. Golovko. Editorial. - p. 7-8.
2. L. Labiod, N. Grozavu, Y. Bennani. Simultaneous Topological Categorical Data Clustering and Cluster Characterization. - p. 9-23.
3. N. Rogovschi, M. Lebbah, Y. Bennani. A Self-Organizing Map for Mixed Continuous and Categorical Data. - p. 24-32.
4. V.G. Red’ko. Approaches to Modeling of Cognitive Evolution. - p. 33-41.
5. W. Pietruszkiewicz. Unbiasedness of Feature Selection by Hybrid Filtering. - p. 42-49.
6. A. Borovyi, V. Kochan, T. Laopoulos, A. Sachenko. Improved Sorting Methodology of Data-Processing Instructions. - p. 50-55.
7. T. Wang, F. Gautero, C. Sabourin, K. Madani. A Neural Fuzzy Inference Based Adaptive Controller for Nonholonomic Robots. - p. 56-65.
8. A. Imada. Is Artificial Neural Network Intelligent? - p. 66-76.

Editorial
Neural Networks and Artificial Intelligence

Akira Imada 1), Vladimir Golovko 2)

1) Brest State Technical University
Moskowskaja Street, 267
Brest 224017 Belarus
akira@bstu.by
2) Laboratory of Artificial Neural Networks
Intelligent Information Technologies Department
Brest State Technical University
Moskowskaja Street. 267,
Brest 224017 Belarus
gva@bstu.by

The 5th International conference on Neural Networks and Artificial Intelligence (ICNNAI’2010) was held during 1-4 June 2010 in Brest, Belarus.
We had started preparing for it about one and half year before the conference. Preparing a conference is sometimes more like a drama. For example, some do not hesitate to try to submit a paper created just by drag and paste from papers already published by other authors, that is, a perfect plagiarism. Some try to eat a free lunch, that is, try to submit a paper without any intention of showing up at the conference site, and later would sent an e-mail asking, which page of the proceedings their paper is on. Some reply to the call-for-paper saying, “if as a key-noter, otherwise no.” Among others, most interesting singularity this time is, some ten’s of emails, asking how to get official invitation letter just claiming the sender is a scientist. The most interesting one was the email query saying, “I am very interesting in your conference theme - Artificial Intelligence - since I am working in my country as an intelligent agent. So I eager to join your conference and sincerely asking you to send me an official invitation letter.” Well, we had eventually overcome these strange intentions.
It would be good if a conference has lots of repeaters and has an atmosphere like a meeting of birds of a feather. In that case, participants’ motivation is, something like, “I have a conference to attend then which topic?” while originally motivation should be like, “I have a topic then which conference?” However, it is more or less difficult to attract the latter attitude of participants. So a dilemma.
Anyway we were happy to have the both type of wonderful 26 participants other than two world-topclass key-note talks by Shun-ichi Amari and Xin Yao.
Now, owing to Professor Anatoly Sachenko, editor-in-chief of this journal, we have an opportunity to publish some of the submitted papers to the conference in this journal as a special issue. Firstly, we picked up 10 candidate authors for this issue. Then, our policy of final selection from them for this special issue was to select those succeeded in making their extended versions elaborate much more than their originals submitted to the conference. If they were similar or identical, there would be no necessity of such a special issue.
So Call-for-Paper read, “Usually, whenever we are writing an article, we have to stop elaborating the manuscript because of deadline. So this would be a good opportunity to further elaborate our already published papers. Hence, this time, our policy of selecting papers for the special issue will be, how the paper has extended elaborately, rather than its quality per ce. Of course quality is one of the most important factors, but if the version already published in the proceedings has no room to be elaborated, there would be no reason to re-publish the same version even if it was a high quality one.”
Some had not responded, some ran out of time, some rejected and thus we have 8 papers here in this special issue as follows.
The first two papers are those discussed in the special session “Incremental Topological Learning Models and Dimensional Reduction.”
(1) “Simultaneous Topological Categorical Data Clustering and Cluster Characterization” By Lazhar Labiod, Nistor Grozavu, and Younes Bennani. The authors propose a method to cluster data specifically categorical data. The method uses a visualization called Relational Topological Clustering, that is, it visualizes the clustering result on a 2-D grid. Then detects relevant features which detect the most important variables. The approach was validated on some real dataset and suggested its effectiveness.
(2) “A Self-organizing Map for mixed continuous and categorical data” By Nicoleta Rogovschi, Mustapha Lebbah, Younes Bennani. This is Also a proposition of clustering data, but here, they are made up of both numerical and categorical variables, by making a self organizing map learn those topological maps. The method is tested using the data sets taken form a public data repository, and resulted in a good quality of clustering.
The next topic is from sort of ethology, for a change.
(3) “Approaches to Modeling of Cognitive Evolution” By Vladimir G. Red’ko. The author overviews adaptive behavior of organisms in their environment, and then explores a modeling of how animal cognitive abilities evolve. Then a future direction of this study is argued.
Now let’s go back a practical topic.
(4) “Unbiasedness of Feature Selection by Hybrid Filtering” By Wieslaw Pietruszkiewicz. The topic is feature selection. The author proposes a robust hybrid method of it, aiming an unbiased selection. The method is verified using a personal bankruptcy dataset, and it shows the method creates effective and unbiased features quickly.
(5) “Improved Sorting Methodology of Dataprocessing Instructions” By Andrii Borovyi, Volodymyr Kochan, Theodore Laopoulos, Anatoly Sachenko. With target being designing a set of instructions of a CPU to process data for sorting purpose, authors experiment a neural network which estimates power consumption, and discuss how to create appropriate training sets. It is shown that the proposed sorting method provides a high accuracy estimation of power consumption.
Then robotics.
(6) “A Neural Fuzzy Inference Based Adaptive Controller for Nonholonomic Robots” By Ting Wang, Fabien Gautero, Christophe Sabourin, and Kurosh Madani. Being based on an adaptive neural fuzzy inference, the authors proposed a control strategy for a nonnholonomic robot. The strategy is tested with a path-finding problem using virtual robot at the beginning, and then a validation is tried on a concept of real robot.
However, are those agent, robot, method, whatever the authors mentioned above really intelligent? At least they do not have a human like intelligence. This conference title includes Artificial Intelligence. So we try a round table discussion under a title “Is artificial intelligence by neural network intelligent?” The next paper is the paper to break the ice of the discussion.
(7) “Is Artificial Neural Network Intelligent?” By Akira Imada. The author speculates that one of characteristics behavior by human is, if it requires an intelligence, is a spontaneous behavior even in a similar situation as before. And the author shows a couple of experiments using spiking neurons and a thought-experiment using quantum computation. While those experiment can simulate a spontaneity, they are far from human like intelligence since it looked to be made randomly rather than consciously.
That’s all. We hope you enjoy those topics.

Top

SIMULTANEOUS TOPOLOGICAL CATEGORICAL DATA CLUSTERING
AND CLUSTER CHARACTERIZATION

Lazhar Labiod, Nistor Grozavu and Younès Bennani

LIPN – UMR CNRS 7030 Institut Galilée – University Paris-Nord
99, avenue Jean-Baptiste Clément 93430 Villetaneuse, France
{lazhar.labiod, nistor.grozavu, younes.bennani}@lipn.univ-paris13.fr

In this paper we propose a new automatic learning model which allows the simultaneously topological clustering and feature selection for quantitative datasets. We explore a new topological organization algorithm for categorical data clustering and visualization named RTC (Relational Topological Clustering). Generally, it is more difficult to perform clustering on categorical data than on numerical data due to the absence of the ordered property in the data. The proposed approach is based on the self-organization principle of the Kohonen’s model and uses the Relational Analysis formalism by optimizing a cost function defined as a modified Condorcet criterion. We propose an iterative algorithm, which deals linearly with large datasets, provides a natural clusters identification and allows a visualization of the clustering result on a two dimensional grid. Thereafter, the statistical ScreeTest is used to detect relevant and correlated features (or modalities) for each prototype. This test allows to detect the most important variables in an automatic way without setting any parameters. The proposed approach was validated on variant real datasets and the experimental results show the effectiveness of the proposed procedure.

Top

A SELF-ORGANIZING MAP FOR MIXED CONTINUOUS
AND CATEGORICAL DATA

Nicoleta Rogovschi 1), Mustapha Lebbah 2), Younès Bennani 2)

1) LIPADE – University Paris Descartes
45, rue des Saints Pères
75270 Paris Cedex 06, France
nicoleta.rogovschi@parisdescartes.fr
2) LIPN – UMR CNRS 7030 Institut Galilée – University Paris-Nord
99, avenue Jean-Baptiste Clément 93430 Villetaneuse, France
{mustapha.lebbah, younes.bennani}@lipn.univ-paris13.fr

Most traditional clustering algorithms are limited to handle data sets that contain either continuous or categorical variables. However data sets with mixed types of variables are commonly used in data mining field. In this paper we introduce a weighted self-organizing map for clustering, analysis and visualization mixed data (continuous/binary). The learning of weights and prototypes is done in a simultaneous manner assuring an optimized data clustering. More variables has a high weight, more the clustering algorithm will take into account the informations transmitted by these variables. The learning of these topological maps is combined with a weighting process of different variables by computing weights which influence the quality of clustering. We illustrate the power of this method with data sets taken from a public data set repository: a handwritten digit data set, Zoo data set and other three mixed data sets. The results show a good quality of the topological ordering and homogenous clustering.

Top

APPROACHES TO MODELING OF COGNITIVE EVOLUTION

Vladimir G. Red’ko

Scientific Research Institute for System Analysis,
Russian Academy of Science,
44/2, Vavilova Str., Moscow, 119333, Russia,
vgredko@gmail.com, http://www.niisi.ru/iont/staff/rvg/index_eng.php

Approaches to modeling of cognitive evolution that is evolution of animal cognitive abilities are proposed and discussed. Backgrounds of models of cognitive evolution, that are developed in an area of researches “Adaptive behavior”, in which modeled “organisms” adapting to variable environment are studied, are outlined. Initial steps of modeling of cognitive evolution are characterized. The sketch program for future investigations of cognitive evolution is proposed.

Top

UNBIASEDNESS OF FEATURE SELECTION BY HYBRID FILTERING

Wiesław Pietruszkiewicz

West Pomeranian University of Technology
49, Żołnierska Str., 71-210 Szczecin, Poland,
wpietruszkiewicz@wi.zut.edu.pl

In this article we examine characteristics of feature selection algorithms by introducing their aspects important in practice. We will focus on the unbiasedness, analyse it and investigate a robust hybrid method of feature selection, being a composition of several feature filters, that could ensure unbiased results of selection. Using parallel multi-measures and voting, we reduce the risk of selecting non-optimal features, a common situation when we select attributes using single evaluation based on one evaluation criterion. To test this method we selected a personal bankruptcy dataset, containing various types of attributes and one of the popular machine learning benchmarks. By the performed experiments we will demonstrate that an approach of multi-evaluation used for features filtering may lead to the creation of effective and fast methods of features selection with an unbiased outcome.

Top

IMPROVED SORTING METHODOLOGY OF
DATA-PROCESSING INSTRUCTIONS

Andrii Borovyi 1), Volodymyr Kochan 2), Theodore Laopoulos 1), Anatoly Sachenko 2)

1) Electronics Lab, Physics Dept., Aristotle University, 54124, Thessaloniki, Greece
e-mail: aborovyi@physics.auth.gr, laopoulos@physics.auth.gr
2) Research Institute for Intelligent Computer Systems, Ternopil National Economic University
3, Peremoha Sq., 46020 Ternopil, Ukraine
e-mail: oko@tneu.edu.ua, as@tneu.edu.ua

An improved classification methodology for sorting data-processing instructions for ARM7TDMI CPU core is presented in this paper. Main discussion here is related to the process of creating appropriate training sets for neural network (NN) based estimation of power consumption. We have proposed instructions’ sorting methodology according to the binary instruction representation and the resources being used for the overall system model. Thus separate instructions groups are obtained for NN-based estimation of power consumption. Experimental results of the proposed method confirm successful usage of this sorting methodology for providing higher accuracy estimation of power consumption.

Top

A NEURAL FUZZY INFERENCE BASED ADAPTIVE CONTROLLER
FOR NONHOLONOMIC ROBOTS

Ting Wang, Fabien Gautero, Christophe Sabourin and Kurosh Madani

Images, Signals, and Intelligent Systems Laboratory (LISSI / EA 3956),
University PARIS-EST Creteil (UPEC),
Senart Institute of Technology, Avenue Pierre Point,
77127 Lieusaint, France,
{sabourin, madani}@univ-paris12.fr, http://lissi.univ-paris12.fr

In this paper, we propose a control strategy for a nonholonomic robot which is based on an Adaptive Neural Fuzzy Inference System. The neuro-controller makes it possible the robot track a desired reference trajectory. After a short reminder about Adaptive Neural Fuzzy Inference System, we describe the control strategy which is used on our virtual nonholonomic robot. And finally, we give the simulations’ results where the robot have to pass into a narrow path as well as the first validation results concerning the implementation of the proposed concepts on real robot.

Top

IS ARTIFICIAL NEURAL NETWORK INTELLIGENT?

Akira Imada

Brest State Technical University
267, Moskowskaja str., Brest 224017 Belarus
akira@bstu.by

This article was originally written for the purpose of breaking the ice in the round table discussion held in the conference. Since the name of the conference is ‘Neural Network and Artificial Intelligence’ the topic of this article is, “What is intelligence?” when we talk about artificial intelligence in general, and artificial neural network in particular. In the history of the field of artificial intelligence, we have had many arguments claiming that artificial intelligence was not intelligent enough yet, or would not be possible to be intelligent even in the future. We take a brief look at such arguments in the history, and then try a speculation concerning if a machine intelligence is as flexible as human intelligence or not. Some experiments of path-finding, with spiking neurons, from this point of view are shown. These were discussed in the roundtable discussion. Here, in this special issue, additionally one thought-experiment using a quantum random walk is discussed. Then a further consideration on a role of consciousness for a machine to be intelligent is followed.

Top