[Should do] - irina... ask the original WORD.doc => done - heinz... ask photo & cv => done [To viktor how] - seng via e-mail pages size 09) khalid 8 309K 04) colin 10 132K 11) rauf 6 552K 06) Diego 7 889K 10) Doudkin 5 678K 07) jj 13 1.06M 13) akira 10 - put in the web 03) kurosh 12 2.71M 02) Nechval 12 4.83M 05) manoonpong 12 5.44M - not-yet-sent 08) bob 10 01) heinz 12 12) gva 8 [Question to ask to the journal] - As for the journal format, like a title all capital who are responsible? [Title and Authors] => first plan => later changed ----------------------------------------------------------- (01) Heinz Muehlenbein ARTIFICIAL INTELLIGENCE AND NEURAL NETWORKS THE LEGACY OF ALAN TURING AND JOHN VON NEUMANN === application to industry === (02) Konstantin N. Nechval, Nicholas A. Nechval, Irina Bausova, Daina Skiltere, and Vladimir F. Strelchonok PREDICTION OF FATIGUE CRACK GROWTH PROCESS VIA ARTIFICIAL NEURAL NETWORK TECHNIQUE (03) Colin M. Frayn INDUSTRIAL APPLICATIONS OF COMPUTATIONAL INTELLIGENCE (04) Kurosh Madani, Matthieu Voiry, Veronique Amarger, Nadia Kanaoui, Amine Chohra, Francois Houbre COMPUTER AIDED DIAGNOSIS USING SOFT-COMPUTING TECHNIQUES AND IMAGE'S ISSUED REPRESENTATION: APPLICATION TO BIOMEDICAL AND INDUSTRIAL FIELDS === NN and its learning === (05) Jean-Jacques MARIAGE HOLISTIC SELF-REPROGRAMMING OF NEURAL NETWORKS: BETWEEN SELF-ORGANIZATION AND SELF-OBSERVATION (06) Diego Ariel Bendersky, Juan Miguel Santos LEARNING FROM THE ENVIRONMENT WITH A UNIVERSAL REINFORCEMENT FUNCTION (07) Poramate Manoonpong, Frank Pasemann, Hubert Roth A MODULAR NEUROCONTROLLER FOR A SENSOR - DRIVEN REACTIVE BEHAVIOR OF BIOLOGICALLY INSPIRED WALKING MACHINES (08) Robert E. Hiromoto, and Milos Manic INFORMATION-BASED ALGORITHMIC DESIGH => INFORMATION-BASED ALGORITHMIC DESIGN OF A NEURAL NETWORK CLASSIFIER === application of NN === (09) Marek Tabedzki, Khalid Saeed. HANDWRITTEN SCRIPT AND WORD RECOGNITION -- A VIEW-BASED APPROACH (10) Alexander Doudkin, Alexander Inyutin THE DEFECT AND PROJECT RULES INSPECTION ON PCB LAYOUT IMAGE (11) Rauf Kh.Sadykhov, and Denis V. Lamovsky ESTIMATION OF THE CROSS CORRELATION BASED OPTICAL FLOW FOR VIDEO SURVEILLANCE. === network intrusion detection === (12) Vladimir Golovko, and Leanid Vaitsekhovich NEURAL NETWORK APPROACHES FOR INTRUSION DETECTION AND RECOGNITION (13) Akira Imada HOW MANY PARACHUTISTS WILL BE NEEDED TO FIND A NEEDLE IN A PASTORAL --- WHO IS A LUCKEY ONE? ------------------- In real FINAL order ------------------- The papers with the final appearing order in the issue (1) Heinz Muehlenbein ARTIFICIAL INTELLIGENCE AND NEURAL NETWORKS THE LEGACY OF ALAN TURING AND JOHN VON NEUMANN (02) Konstantin N. Nechval, Nicholas A. Nechval, Irina Bausova, Daina Skiltere, and Vladimir F. Strelchonok PREDICTION OF FATIGUE CRACK GROWTH PROCESS VIA ARTIFICIAL NEURAL NETWORK TECHNIQUE (04) Kurosh Madani, Matthieu Voiry, Veronique Amarger, Nadia Kanaoui, Amine Chohra, Francois Houbre COMPUTER AIDED DIAGNOSIS USING SOFT-COMPUTING TECHNIQUES AND IMAGE'S ISSUED REPRESENTATION: APPLICATION TO BIOMEDICAL AND INDUSTRIAL FIELDS (14) newly added on 14 December Volodymyr Turchenko, Viktor Demchuk, Anatoly Sachenko SIMULATION MODELING OF INTERPLANETARY SHOCKS ARRIVAL TIME PREDICTION ON HISTORICAL DATA SET (03) Colin M. Frayn INDUSTRIAL APPLICATIONS OF COMPUTATIONAL INTELLIGENCE (07) Poramate Manoonpong, Frank Pasemann, Hubert Roth A MODULAR NEUROCONTROLLER FOR A SENSOR - DRIVEN REACTIVE BEHAVIOR OF BIOLOGICALLY INSPIRED WALKING MACHINES (06) Diego Ariel Bendersky, Juan Miguel Santos LEARNING FROM THE ENVIRONMENT WITH A UNIVERSAL REINFORCEMENT FUNCTION (05) Jean-Jacques MARIAGE HOLISTIC SELF-REPROGRAMMING OF NEURAL NETWORKS: BETWEEN SELF-ORGANIZATION AND SELF-OBSERVATION (08) Robert E. Hiromoto, and Milos Manic INFORMATION-BASED ALGORITHMIC DESIGH (09) Marek Tabedzki, Khalid Saeed. HANDWRITTEN SCRIPT AND WORD RECOGNITION -- A VIEW-BASED APPROACH (10) Alexander Doudkin, Alexander Inyutin THE DEFECT AND PROJECT RULES INSPECTION ON PCB LAYOUT IMAGE (11) Rauf Kh.Sadykhov, and Denis V. Lamovsky ESTIMATION OF THE CROSS CORRELATION BASED OPTICAL FLOW FOR VIDEO SURVEILLANCE (12) Vladimir Golovko, and Leanid Vaitsekhovich NEURAL NETWORK APPROACHES FOR INTRUSION DETECTION AND RECOGNITION (13) Akira Imada HOW MANY PARACHUTISTS WILL BE NEEDED TO FIND A NEEDLE IN A PASTORAL --- WHO IS A LUCKEY ONE? ================================================================================================== [MEMO FOR DEBUG] Konstantin N. Nechval et al. PREDICTION OF FATIGUE CRACK GROWTH PROCESS VIA ARTIFICIAL NEURAL NETWORK TECHNIQUE === letter to the authors === For the editorial purpose they require originl word document, so good chance, though its just a set of small - 1142 => 1,142 - respectively. => respectively; - can be obtain => can be obtained - ... (FCGR) is given => are given - stress intensity solution (67) => (13) ? - class.cation=> classification (maybe this is when translated to pdf) - Pairs formula => Paris formula (several times) -------------------------------------------------------------------------------------------------- Colin M. Frayn INDUSTRIAL APPLICATIONS OF COMPUTATIONAL INTELLIGENCE bug - variable in text should be italic - indent - empty line + indent - empty line + noindent - (after eq (4)) Where here => ? - L10m - L50Bn - tat => that - why left hand side of p 10 is empty? - a false positive is far worse than a false negative. => very end as for selection not understandable - Evolutionary algorithms will always find a way to exploit any poorly-considered fitness function in order to fulfil it to the letter, but with a potential solution tat does not achieve what you actually desire.??? === debug again === title 1) => delete whole ' ' p.2 L In "Travelling ... variability." N and so on => IT i, n delta m L ', => ,' (also others though of => thought of) p.3 L N-point empty line befor "These steps... R eq => IT p.4 L For example => => empty line OK but indent -> no Pi Si=> IT The fact that => empty line OK but indent -> no p.4 R Cn, i Cni Lmi => IT The updated => remove empty line and inden Where here => Here => no indent i, n p5. L Cm n, i ==================== OK befor 3. Stock Fi ==================== give up -------------------------------------------------------------------------------------------------- Kurosh Madani, Matthieu Voiry, Veronique Amarger, Nadia Kanaoui, Amine Chohra, Francois Houbre "Computer aided diagnosis using soft-computing techniques and image's issued representation: application to biomedical and industrial fields." -------------------------------------------------------------------------------------------------- Diego Ariel Bendersky, Juan Miguel Santos LEARNING FROM THE ENVIRONMENT WITH A UNIVERSAL REINFORCEMENT FUNCTION bug - abstract - and ach new task => each - consist => consists => DONE - the the => the => DONE - find a whole => find a hole => DONE - ?). => ?) - :a) -> : a) - 2.a 2.b 2.c => The authors start the paper with their description "... environment affects not only how animals perform particular tasks, but also what skills an animal will develop during its life. Then paper describes how a robot learn skills like avoiding obstacles, finding a hole in the wall to pass it, traversing a corrior, etc. -------------------------------------------------------------------------------------------------- Alexander Doudkin, Alexander Inyutin THE DEFECT AND PROJECT RULES INSPECTION ON PCB LAYOUT IMAGE - A writer talks of things that everyone knows but does not know they know. -------------------- A proposal by using image processing to find defects on printed circuit board such as unexpected pin holes or absence/displacement of elements which are supposed to be printed. (it seems not yet applied to manufacturing stage.) - layoutcan -> layout can => done - dependinging => done - logycal => logical - less then => less than (3 times) - fig. => Fig others - and classification => , classification => done - (caption) Fig.1 - (all) => not touched - Fig.2 ; => : page 3 - itemize: additinal - => done - indent The... and others => done - Rmin... centering => not changed - ,B => and B -------------------------------------------------------------------------------------------------- Rauf Kh.Sadykhov, and Denis V. Lamovsky ESTIMATION OF THE CROSS CORRELATION BASED OPTICAL FLOW FOR VIDEO SURVEILLANCE. In the topic of computer vision, how each point of an image moves from one scene to the next, e.g. direction and velocity is an important issue and called Optical Flow. The paper proposes a method using cross correlation to determin how similar between two optical flows which author wrote is robust but computationally expensive, and proposition is to use a parallel processing. (strange enough all bold font) Proposal of an algorithm which is exploting prallel processors Although since authors wrore, "All calculations ... were performed ... on a processor ... Pentium 4 ...,"it was a simulation of the algorithm on multi-processors -------------------------------------------------------------------------------------------------- Jean-Jacques MARIAGE HOLISTIC SELF-REPROGRAMMING OF NEURAL NETWORKS: BETWEEN SELF-ORGANIZATION AND SELF-OBSERVATION - pre-surgery - 6 months - overlap - Right elbow - Left elbow send letter to author - reference: [27] ICNNAI-03 => 2003 not 2004 - here after => hereafter - More over => moreover - dosen't => does not - cooperative / competi-tive => cooperative/competi-tive - extern => external - no the least => not the least - avail-able => available - postsynaptic => post-synaptic - eg => e.g. strange when grasp"HOLISTIC SELF-REPROGRAMMING OF NEURAL NETWORKS: BETWEEN SELF-ORGANIZATION AND SELF-OBSERVATION Jean-Jacques MARIAGE" it appears "general conception of learning and training dynamic character of learning structures and have considered to design the mechanisms of --- ask JJ what tricks about hiphenation => done -------------------------------------------------------------------------------------------------- Marek Tabedzki, Khalid Saeed. HANDWRITTEN SCRIPT AND WORD RECOGNITION -- A VIEW-BASED APPROACH p.3 R -where [H] is ... => shouldn't indent The authors applies their already successful method of recognizing who spoke what words to what is written just by authors call View-based method also successfully. -------------------------------------------------------------------------------------------------- Kurosh Madani, Matthieu Voiry, Veronique Amarger, Nadia Kanaoui, Amine Chohra, Francois Houbre COMPUTER AIDED DIAGNOSIS USING SOFT-COMPUTING TECHNIQUES AND IMAGE'S ISSUED REPRESENTATION: APPLICATION TO BIOMEDICAL AND INDUSTRIAL FIELDS {\bf Can ANN monitor human health in hospital in the same way as monitoring machine in industry?} The authors try this challangging riddle by attempting to diagnose auditory pathway of real human as well as to diagnose defects of high-tech optical device by using soft computing techniques including ANN and Fuzzy logic. --- wondering why black thin line top and botom of every pages are missed in you paper also indent is very small - 2.1 BEAP or BAEP => also in the second line from the last line in page 3. - see Fig.2. => Fig. 6.? - processfs in page 5 - 3-th => 3rd in page 6 - gcleaning mark-mikehin p.7 - two kind -> two kinds in p.7 - Clering? in p.8 - 8-by 8 (3times) - "p" in potential defect in page 7 has an accessary - non italic valuavle simbol like m-squeare --- one option is I correct => done (he made) -------------------------------------------------------------------------------------------------- Poramate Manoonpong, Frank Pasemann, Hubert Roth A MODULAR NEUROCONTROLLER FOR A SENSOR - DRIVEN REACTIVE BEHAVIOR OF BIOLOGICALLY INSPIRED WALKING MACHINES well organized and well written -------------------------------------------------------------------------------------------------- Vladimir Golovko, and Leanid Vaitsekhovich NEURAL NETWORK APPROACHES FOR INTRUSION DETECTION AND RECOGNITION Authors propose a network intrusion system by combining principal component analysis neural network to reduce the dimension of imput data and multilayer percerptron to recognize attacks from normal transactions. The target is KDD-cup-99 dataset, which has been, and still is going to be, a frequantly used common benchmark for evaluation of intrusion detection techniques. -------------------------------------------------------------------------------------------------- HOW MANY PARACHUTISTS WILL BE NEEDED TO FIND A NEEDLE IN A PASTORAL --- WHO IS A LUCKEY ONE? Akira Imada p.1 - needed => necessary - --- -> dash - keyword = Network intrusion detection, ... - -- -> dash - 2^20 p.2 - \it x3 - subtitle not capital - "Iris dataset is made up of - } - two -> normal and one -> abnormal - $n$ - is => no indent - cite p.3 - it - this -> this reduced dataset - footnote 8, 9 p.3 L - -- -> dash x2 => done - Fig. 5 -> to bottom - 3.2 & 3.4 -> not with capital => done p5. R - -- -> dash => done p.7 L - eq -> centering => done - (1) -> ... => done - 1.315*10^139 -> x => done p.7 R ... almost bottom - system -> the system => done p.8 L - to learn -> add . => done - after footnote 13 -> two . => done - problem 4 - -- -> dosh => done - cl -> cancel => done - {\it => done - The two -> These two => done - footnote 13: space after . => done p.10 ... after Concluding Rmarks - 1999 => 99 => done - take it acount => take it into acount => done - perelsen => perelson => done [my-CV] - Artificial human-like ignorance -------------------------------------------------------------------------------------------------- Heinz Muehlenbein ARTIFICIAL INTELLIGENCE AND NEURAL NETWORKS THE LEGACY OF ALAN TURING AND JOHN VON NEUMANN [bug] P,1 L do their exists solutions today? => do their solutions exist today? => done p.2 R footnote no and . or vice versa? => OK p2. L been been => already done p3 R A learning mechanism => indent => OK p5 R phi -> \phi => already done von Neuman later => indent? => maybe not => ok indent p6. R Ther seem to be => indent => already done p6. L * => x 3 times => already done read text => texts => done sensitive-they => done " but => not changed whatever-two => done p.7 R valiant [20] in 1984 => done in 1999 (8) => [8] in 1999. => done p8. R This algorithm has ... whetger there outputs -> their => done p.8 L n -> Probabilistic => indent => already done p.9 L each TM M a bit-string < M >.? => done === debug again === ... all citations itlicized incert empty line both end of itemized term items should end with ; or . n, w -> === debug further again === title -> cr befor legacy => OK whole ... p.2 L - "what will ... vame?" => '...' => Done - 10^9 => Done - why how why => ... and why => Done p. 2 R - say at birth => say, at birth => Done - footnote 1 $ => Done p3.R - Turing notes => not indent => Done - "There is => cr and italicize => Done - 3 items end with . => Done - "We normally ... to it" => IT => Done - "The use => IT => Done - "Again => CR => Done p3. L - 2^N -> IT => Done - ... => ..., => Done - 1.2. end with . => Done p4. R - P,Q,W,S => P, Q, W, S => Done - 1-pq =>1 - pq => Done - "Diciplin ... machines." => IT => Done - "A great positime ... powers in" => IT p.4 L - Turing remarks "The learnign ... feasible." => IT => Done - Natural => with "" => Done - there have been => not CR => Done p.5 R - "The logic ... respect. => not IT remove " - Von neumann tried => no indent - This new => "" at the both ends p.5 L - "The "functioning" of ... system." => IT and '' => Done p.5 R - Ther is => not CR => OK not change - "One can stat ... => IT => Done - Now it is ... => "" => Done - The forgoing ... => "" => Done p.6 L - equation => insert space speciall aroung + => Done p6.R - themachine => => Done p.7 L - an estimate => dash - Note that => no indent - More precisely => bf - alos variable, eq => IT p.8 L - '' => " => Done - This design => IT => Done p.9 L - 2^n => IT p.9 R - If M is a TM which on input w outputs string x, then the concatenated string < M >; w is a description of x. => If M is a TM in which input of a description of x outputs string x, then the concatenated string < M >. - 2^n 2^(n-1) => IT - that is => that is, => Done - 2n => 2^n => Done p.10 R - 2^-n => IT => Done - The bottom => to L by CR if necessary => OK no need p.10 L - 3 itmes => end with . - footnote add " after ... [letter to heinz] - correct a clear slip of finger like "been been" on the left side of p.2 - all the citation from von Neuman and Turing changed italicized as you did in the right side of p.2. And added quatation mark like "..." since your own texts sometimes are italicized. - If M is a TM which on input w outputs string x, then the concatenated string < M >; w is a description of x. - when you quoted from von Neuman, Turing and others, I found 1) in the text usually you - italisized and quotation mark "..." but sometimes - not italisized but with "..." - italisized but no "... " 2) as an independent paragraph all italisized but without "..." or with "..." I have had no time to ask but I edited all the quatation <>, whichever they appear in a text or as an independent paragrap. I want your paper [13] [13] H. Muhlenbein and R. Hons. The estimation of distributions and the minimum relative entropy principle. Evolutionary Computation. Vol. 13(1). 2005. pp. 1-27. => already got via google -------------------------------------------------------------------------------------------------- GVA two affiliations => one abstract combination => combination of P1. k => IT IDS => of IDS p2. type of atack => of attack type is proposed => we propose for recognizing +> to recognize and => or Per se this => This integrated knowledge per se where no indent xi ... => xi is input to the i-th unit for performing => to perform Combining => By combining what => which -------------------------------------------------------------------------------------------------- INFORMATION-BASED ALGORITHMIC DESIGH => INFORMATION-BASED ALGORITHMIC DESIGN OF A NEURAL NETWORK CLASSIFIER As I feel like I'm trying "a leture to Buddah" p.4 L triangle up side down => not possible for me to correct p.2 L - \lambda_{I}s represents -> represent ... OK p.3 L - always clear. . => . p.3. R Training by design => design by training? - \alphasuggests => alpha suggest (space necessaly) Fig. 2 - should both outputs from two osc-layers for x and y enter to the one AND unit (in the figure output from x are to rectangle box and the one from y to circle. Or if you agree I can change p.5 L The n x m output => n x m => italicized p.6 R - ''stripped\" => "stripped" -> two times also ''grow as needed" - The canonical ... rotations adds units ... => add p.9 L (x,y} => italicize reference - (2005) AGAIN design by training p.5 R if the choice of delta is chosen appropriately => if the value of delta is chosen appropriately p.6 equation I-p = ... put period at the end p.7 L d of d-ortohgonal => italicized 45(symbol of degree) was changed by WORD --- p.2 where p.9 However delete space The As --------------------------------------------------------------------------------------------------- Volodymyr Turchenko, Viktor Demchuk, Anatoly Sachenko SIMULATION MODELING OF INTERPLANETARY SHOCKS ARRIVAL TIME PREDICTION ON HISTORICAL DATA SET p.1 L - heath => health p.1 R - et al => et al. p.2 L - 0:03 => 0.03 p2. R - day => days - of 2000 => in 2000 - except from => except for p.5 L - equal 5 => equal to 5 not corrected but jornal format indicate, e.g., 2005 instead of (2005) and p. 989-993 is pp.9890993 in the format Text to be sent to Journal ================================================================================================= [Preface] <> - the period 2007-2009 is not (and I underline the "not" here) going to be a period when ... - When I underlined his words from the commission's meeting during the TV program ... - a problem that is now sitting on the EU's lap as a "hot potato." => a problem that is now sitting on my lap as a hot potato is ... - A writer talks of things that everyone knows but does not know they know. - "I nearly fell from my chair -> might make someone fell from the chair. - The reason to be shocked might be ... -> The reason to be exited is - Hope you enjoy reading. Guest editors. - According to a report in The Times of London in November, he traveled ... - plagiarism - you will be find the answer "yes" when you read the article "..." by - swan in the snow or craw in the sky at night - Opera fiasco raises question: Does booing have its place? - --- => hyphen - 30th May to 1st June - were all => were found to be all - enable => enabled - the => that - Brest, Belarus - true positive selection ... false negative one - krosh's title -> IT - how a robot learn - human-like-waking=> - senser- => hyphen - proposed => proposes <> --------------------------------------------------------------------------------------------------- (00) After a series of three events in 1999, 2001, 2003, gradually evolving to be better than before, the 4th of this conference --- International Conference on Neural Network and Artificial Intelligence was held in Brest Belarus from May 30 to to June 1, 2006 (ICNNAI'2006) A preparation of a conference is sometimes very dramatic. Let us introduce one episode happend during our selection procedure of the submissions. One guy tried to submitt three papers. Then those three submissions were all just the result of cut and paste operations from other already published paper from public domain web sites. This is not a citation, but all perfect copies of other authors. It is much worse than a plagerism and totally illegitimate. Fortunately, thanks to our Programm committy we avoided this situation of accepting his papers blindly. In this speciall issue, we select 14 out of 33 papers from our proceedings. This selection procedure was not so easy, partly due to the space. We faced a dilemma how our selection to be true positive risking a false negative selection. In the meantime Then we had unexpectedly a kindly offer of additional space to publish the papers submitted to ICNNAI'2006 in the one issue advance of this speciall issue from the editor in chief of the journal, which enable us to apply a somehow strange but a pragmatic formula one-author-apeares-only-once, not so intelligent one though. The papers published in the previous issue are as follows. Colin M. Frayn (UK) "A review of industrial applications of comuptaional intelligence." Marcin Adamski and Khalid Saeed (Poland) "Heuristic techniques for handwritten signature classification." Kurosh Madani (France) "Modular and self-organizing connectionist systems: toword higher level intelligent functions." Nataliya Kussul, Serhiy Skakun, Olga Kussul (Ukraine) " Comparative analysis of neural networks and statistical approaches to remote sensing image classification. Nabil M. Hewahi (Palestine) "Soft computing as a solution ot time/cost distributor." Iryna Turchenko, Volodymyr Kochan, and Anatoly Sachenko (Ukraine) "Simulation modelling of neural control system for coal mine ventilation." Aleksej Otwagin, and Alexander Doudkin (Belarus) "A framework for parallel processing of image dataflow in industrial applications." {\bf What is intelligence?} (01) The conference started with the plenary talk by H. Muehlenbein (Germany). This issue also starts with his paper. Agence France-Presse reported that Deep Fritz, the world's leading chess computer, beat its human counterpart, the Russian world chess champion Vladimir Kramnik, in a six-game encounter. Of the six games, Deep Fritz won two and four ended in draws. Then, can we conclud Deep Fritz has a human-like intelligence? There have been a fair amount of {\it definitions of life} but still we do not have a clear one as Chris Langton -- the founder of scientific discipline of Artificial Life --- claimed, "every time we succeed in synthetically satisfying the definition of life, the definition is lengthened or changed." On the other hand, Turing Test has never succeeded in gaining an inch despite of huge amount of challanges. As Muehlenbein wrote "there is no system in sight which comes near to passing the Turing test." Muehlenbein contribute to this issue with the paper entitled "Artificial intelligence and neural networks the legacy of alan turing and john von neumann." This article is not a nostalgic history of computer science, but rather a challenge to AI community, as the author starts the article with the Von Neumann's doubts that creating artificial intelligence will be possible with teacher-based-learning, since our concern is more or less on creaiting artificial intelligence by learning." He shows still open issues or unsolved provlems. For example, he introduced McCarthy's idea of designing human or animal like intelligence starting with an appropriate mental structure instead of a sheet of blank paper like Turings, as a problem nobody seems working at this moment. Just one thing I want to add here is a variation of Turing's imitation game with intelligent guy who pretend to be ignalant versus ignorant guy who think himself is intelligent. --------------------------------------------------------------------------------------------------- {\bf Is current ANN technology acutually being used in Industry?} (2) The answer is yes. We now take a look at few examples presented in the confrerence. The first one is an application to aerospace industy: "Prediction of fatigue crack growth process via artificial neural network technique" by K. N. Nechval, N. A. Nechval, I. Bausova, D. Skiltere, and V. F. Strelchonok (Latvia). Authors started abstract with "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." Yes, indeed. So, in the paper, as authors wrote "the artificial neural network technique for the data processing of on-line fatigue crack growth monitoring is proposed after analyzing the general technique for fatigue crack growth data." After well organized survey, authors wrote, "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." Even though people who make $150,000 are considerably happier than those who make $40,000, it's not clear why, --------------------------------------------------------------------------------------------------- {\bf Does money make you happier?} "Does being happier in the first place allow you to earn more money later, maybe by way of greater creativity or energy? Or does some other factor produce both money and happiness?" asks psychologist R. E. Lucas of Michigan State University. (3) The third paper is "Industrial applications of computational intelligence" by C. M. Frayn (UK). The topic is data analysis for financial applications. He pose the question, {\it "Are data mining methods learning real knowledge that depends on human trading behaviour, or are they merely learning to predict patterns in arbitrary time series?} Recent United Nation's report says, "The world's richest 2% own more than half of global wealth, while half the world's population owns only 1 %." Have they struggle to be that super rich? I guess not. Some times, we know, "When Doing Nothing Is Better"\footnote{The headline of article of New York Times on 16 December 2006.} The author also mentioned about, "No free lunch vs random investment instead of using artificial intelligent." What is the point? Well, read the paper. The author also explore a root optimization when we must salt in order for the road not to freeze --- much more complicated and practical variant of TSP. --------------------------------------------------------------------------------------------------- (05) Yet another contribution is a possibility of application in space industry. Volodymyr Turchenko, Viktor Demchuk, Anatoly Sachenko (Ukraine) contributes with their paper: "Simulation modeling of interplanetary shocks arrival time prediction on hisotical data set." High energy particles and solar wind to the earth from the planetary space which is called interplanetary shocks influence on spacecrafts, sattelites, radio communication, radors, and among others on human health. The paper proposes a tequnique to predict a future arrivla of these shocks from the data obtained by NASA's spacecraft. -------------------------------------------------------------------------------------------------- {\bf How ANN learns?} Recently Reuters reported "P. Hof and E. Van der Gucht of the Department of Neuroscience at Mount Sinai School of Medicine in New York studied the brains of humpback whales and discovered a type of cell called a spindle neuron in the cortex, in areas comparable to where they are seen in humans and great apes." Or Reuters also reported "Researchers at the Los Alamos National Laboratory in New Mexico said they trained honeybees to stick out their proboscis -- the tube they use to feed on nectar -- when they smell explosives in anything from cars and roadside bombs to belts similar to those used by suicide bombers." Then why not approach mor biologically. (05) J. J. Mariage gives a consideration from such aspect in his paper "Holistic self-reprogramming of neural networks: Between self-organization and self-observation." - Throughout this paper, we non-exhaustively reviewed some essential properties observed in the development of encoding structures in nature. Studying the process by which new structures can emerge in artificial unsupervised NNs models amounts to incorporate a control dynamic of their evolution to the learning algorithms. - Our main research concern is to model and simulate the dynamic character of learning struchures and processes, and their evolution ... Neural darwinism proposed by G. edleman are reffered by jj as - "Brain develops by a Darwinian selection process that takes place at the neuron groups' level instead of the individual neuron. - It would henceforth become possible to develop really autonomous tools that automatically initialize, learn permanently and forget when necessary, while accordingly adapting their structure. - Another point of interest is to look if the uncommitted cells in SOM, which may correspond to intermediary steps of organization, could be used as a SCs reserve and whether they could constitute a pre-learned basis from which further learning could start instead of starting tabula rasa. --------------------------------------------------------------------------------------------------- {\bf How did bats learn to fly back home?} Princeton University batologists used radio telemetry aboard a small aircraft to track big brown bats (Eptesicus fuscus) that were released 20 kilometers (12 miles) north of their home. (06) D. A. Bendersky, J. M. Santos "Learning from the environment with a universal reinforcement function." (07) {\bf Can a machine walk like human?} According to the recent report by Agence France-Presse, "The first walking android will make its debut within two to three years, said So Byung-Rok at the Korea Institute of Industrial Technology, quated as saying, "We are now working to improve the motion and upgrade intelligence so that next-generation androids can walk like a human, engage in more sophisticated conversations and have a wider range of facial expressions." So design human-like-walking machine is still not so easy task. The paper by P. Manoonpong, F. Pasemann, and H. Roth titled, "A modular neurocontroller for a sensor - Driven reactive behavior of biologically inspired walking machines" challenge this topic. -------------------------------------------------------------------------------------------------- (08) Robert E. Hiromoto, and Milos Manic Information-based algorithmic design of a neural network classifier (new title) To simply put, input layer performs an orthogonal search while the next layer performs a rotational search which reduces the search space - Random treatment of problem solutions has proved to provide a convenient approach in surveying landscapes for optimization problems where the solutions space is vast and appears to follows no predetermined schedule or route. - In cases where the patterns are sparsely distributed, the computational search time for the initial space can be dramatically reduced if the choice of value is chosen appropriately. - The algorithm developed above represents a canonical formulation of a clustering technique; however, it can also be used as a preconditioning search algorithm regardless of the dimensionality of the search space. - For this reason, the orthogonal search may be very effective for detecting patterns, rather than clusters. === application of NN === Yet another series of applications but more specific purpose of (09) M. Tabedzki and K. Saeed. "Handwritten script and word recognition -- A view-based approach." (10) A. Doudkin, and A. Inyutin "The defect and project rules inspection on PCB layout image." A proposal by using image processing to find defects on printed circuit board such as unexpected pin holes or absence/displacement of elements which are supposed to be printed. (it seems not yet applied to manufacturing stage.) (11) R. Kh.Sadykhov, and D. V. Lamovsky "Estimation of the cross correlation based optical flow for video surveillance." In the topic of computer vision, how each point of an image moves from one scene to the next, e.g. direction and velocity is an important issue and called Optical Flow. The paper proposes a method using cross correlation to determin how similar between two optical flows which author wrote is robust but computationally expensive, and proposition is to use a parallel processing. - Although since authors wrore, "All calculations ... were performed ... on a processor ... Pentium 4 ...," it was a simulation of the algorithm on multi-processors --------------------------------------------------------------------------------------------------- {\bf Can ANN detect network intrusion after learning?} The final topic of this issue is network intrusion detection. Here we have two papers: The one is optimistic to the topic while the other pesimistic. (12) Vladimir Golovko, and Leanid Vaitsekhovich "Neural network approaches for intrusion detection and recognition." (13) Akira Imada "How many parachutists will be needed to find a needle in a pastoral -- who is a luckey one?" The paper might shock readers with somewhat of a negative attitude of the paper. That is, author suggests there is no panacea to detect intrusion in a real sense, claiming some of the report of successes is like a powder of suger like placebo. Author wrote real dangeourous outlier is the ones which we can successfully classified, but it hides behind normal transactions. Hence, it is like a search for a needle in a haystack. --------------------------------------------------------------------------------------------------- {\fe A needle in a haystack can be reached by a search?} Or would it be difficult like digging a well with a needle? Although A. Imada shows some example of failures to find an efficient such algorithm, both Mulenbein and Hiromoto suggest a hint of possibilities. Muhlenbein wrote, "The problem of the exponential explosion has been solved in the 80's. For singly connected Bayesian networks exact inference is possible in one sweep of Pearl's belief propagation algorithm. Hiromoto colin also pointed out jj - needle in a haystack context - Adaptation thus provides a means to take self-organizing opportunities acting as attractors, to drive random variation towards effi-cient forms of organization. This avoids the needle in a haystack problem and enables natural selection to run across the hills and valleys of a varied land-scape to find basins of attractors, where it is easy to fall. Indeed, self-organizing processes are not very brittle. Their wide range of diversity, their spontane-ous arising, the structured emergence of their or-ganization states, as well as the broad spectrum of the initial conditions for their triggering, seem to in-dicate that attraction basins constitute wide and nu-merous areas in the state space. =================================================================================================== [Original memo for preface] lovely Turkish saying, to dig a well with a needle" to describe what writing is. === memo === As for English for two and more conclete description of problem, if any, of each paper. Conclusion of each papers? Please read them. --- as for diego's ------------------------------------------------- HEAD-1129: FEATURE (Wednesday, November 29, 2006) US scientists say trained bees can sniff bombs PHOENIX - Reuters Scientists at a U.S. weapons laboratory say they have trained bees to sniff out explosives in a project they say could have far-reaching applications for U.S. homeland security and the Iraq war. Researchers at the Los Alamos National Laboratory in New Mexico said they trained honeybees to stick out their proboscis -- the tube they use to feed on nectar -- when they smell explosives in anything from cars and roadside bombs to belts similar to those used by suicide bombers. Researchers in the program, dubbed the Stealthy Insect Sensor Project, published their findings on Monday. By exposing the insects to the odor of explosives followed by a sugar water reward, researchers said they trained bees to recognize substances ranging from dynamite and C-4 plastic explosives to the Howitzer propellant grains used in improvised explosive devices in Iraq. "When bees detect the presence of explosives, they simply stick their proboscis out," research scientist Tim Haarmann told Reuters in a telephone interview. "You don't have to be an expert in animal behavior to understand it as there is no ambiguity." The findings followed 18 months of research at the U.S. Energy Department's Los Alamos facility, the nation's leading nuclear weapons laboratory. "We are very excited at the success of our research as it could have far-reaching implications for both defense and homeland security," Haarmann said. While scientists have trained wasps to respond to the trace of explosives, Haarmann said research with bees appeared to show more promise. --- as for JJ's ------------------------------------------------- fascinating review of real biologikal ... by JJ HEAD-1128: FEATURE (Tuesday, November 28, 2006) Humpback whales have human brain cells WASHINGTON - Reuters Humpback whales have a type of brain cell seen only in humans, the great apes, and other cetaceans such as dolphins, U.S. researchers reported on Monday.c:\documents and Settings ...... Patrick Hof and Estel Van der Gucht of the Department of Neuroscience at Mount Sinai School of Medicine in New York studied the brains of humpback whales and discovered a type of cell called a spindle neuron in the cortex, in areas comparable to where they are seen in humans and great apes. Although the function of spindle neurons is not well understood, they may be involved in cognition -- learning, remembering and recognizing the world around oneself. Spindle cells may be affected by Alzheimer's disease and other debilitating brain disorders such as autism and schizophrenia. Although the function of spindle neurons is not well understood, they may be involved in cognition -- learning, remembering and recognizing the world around oneself. Spindle cells may be affected by Alzheimer's disease and other debilitating brain disorders such as autism and schizophrenia. ... Spindle neurons probably first appeared in the common ancestor of hominids, humans and great apes about 15 million years ago, the researchers said -- they are not seen in lesser apes or monkeys. In cetaceans they would have evolved earlier, possibly as early as 30 million years ago, the researchers said. Either the spindle neurons were only kept in the animals with the largest brains or they evolved several times independently, the researchers said. _____________________________________________ HEAD-1201: FEATURE (Friday, December 1, 2006) South Korean robot will walk the walk as well as talk the talk SEOUL - AFP South Korean scientists are working on a new-generation robot resembling a human that will be able to walk the walk as well as talk the talk, one of the team said Thursday. The first walking "android" will make its debut within two to three years, said So Byung-Rok, one of the team of researchers at the Korea Institute of Industrial Technology. Androids present particular technological challenges in cramming complicated modules, motors and actuators into a life-size body. The team has already developed two android prototypes designed to look like a Korean woman in her early 20s, which can hold a conversation, make eye contact and express anger, sorrow and joy. The latest version, named EveR2-Muse, was unveiled last month at a robot exhibition in Seoul. She made her debut billed as "the world's first android entertainer" singing a new Korean ballad "I Will Close My Eyes for You." [eng] (general) The main debate, he said, is how strong the effect is. "Standing like a human, she can sing a song and move her arms, hips and knees to the rhythm although she cannot lift her feet or walk yet," So told AFP in reference to EveR2-Muse. "We are now working to improve the motion and upgrade intelligence so that next-generation androids can walk like a human, engage in more sophisticated conversations and have a wider range of facial expressions." Researchers at Japan's Osaka University unveiled their own female android capable of speaking and making gestures but So said it was only mobile from the waist up. "We are using electric motors built underneath the silicon skin but a Japanese android uses air pressure to move. The Japanese android moves faster from the waist up than ours does but it needs a giant air tank attached to it," he said. [editorial] emphasizing the second part of the sentence in bold letters. AFP -- Agence France-Presse. heinz the CHP has succeeded in not gaining an inch despite the disappointment with the governing party. why? read his article. What if he insists on becoming the next president of Turkey? Before speculating on such a possibility, we should wait for his response against all these odds that ... google definition of life before mine He shocked everybody with his unique ability to reverse an enraged public feeling against him and against the Western Christian world into a feeling of trust. After a series of three events in 1999, 2001, 2003, gradually evolving to be a better one before the 4th of this conference --- ICNNAI-2006 was held in Brest Belarus from 30 May to 1 June. it was held from Nov. 28 to Dec. 1. also dramatic beside selecting submissions. one guy tried to submitt three papers but all of them were just the result of cut and paste operations from totally other already published paper -- much worse than a plagerism. Fortunately, thanks to our Programm committy we were able to avoid his blindly accepted his papers. illegitimate actions copied materials not cited, --- one author apeare only once he contributed the conference by submitting multiple papers for example ... but due to the space and more ... we one author apeare only once see also the previous issue different 5 papers wer published thanks to the c --- false negative ... 14 are selected out of 33, and finaly 12 papers are here in this issue 2 from 3 planary talks and 10 from Thanks to Anatoly Sachenko we were allowed to yet another space to publish additional 7 papers. Those are Kurosh et al from Paris, France, Khalid Saeed et al.... Let me write about ... this is the previous issue but not this issue. The conference started with the plenary talk by H. Muehlenbein. This issue also start with his paper. (01) H. Muehlenbein "Artificial intelligence and neural networks the legacy of alan turing and john von neumann." This article is not a nostalgic history of computer science, but rather a challenge to AI community, since the author start the article with the Von Neumann's doubts that creating artificial intelligence will be possible with teacher-based-learning, since our concern is more or less on creaiting artificial intelligence by learning. Further, Muelenbein went on as writing that the problemble mo creating artificial intelligence is [still] far from being solved. I'm proposing a variation of Turing's imitation game with intelligent guy who pretend to be ignalant versus ignorant guy who think himself is intelligent. Typical pieces of knowledge represented in the database are lrEvery tree is a plantly and lnPlants die eventuallyln. Cyc project reminds me of Sibmi needle The problem of the exponential explosion has been solved in the 80's. For singly connected Bayesian net- works exact inference is possible in one sweep of Pearl's belief propagation algorithm (15). A very interesting extension for incomplete data is done by the maximum entropy principle (3). But there is no system in sight which comes near to passing the Turing test. add on 04/12 ==== heinz By showing still open issues or unsolved provlems. For example, he introduced McCarthy's idea of designing human or animal like intelligence starting with an appropriate mental structure instead of a sheet of blank paper like Turings, as a problem nobody seems working at this moment. === application to industry === (02) K. N. Nechval, N. A. Nechval, I. Bausova, D. Skiltere, and V. F. Strelchonok Prediction of fatigue crack growth process via artificial neural network technique Authors started abstract with "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." Yes, indeed. So, in the paper, as authors wrote "the artificial neural network technique for the data processing of on-line fatigue crack growth monitoring is proposed after analyzing the general technique for fatigue crack growth data." 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. after well organized survey, ... (03) C. M. Frayn Industrial applications of computational intelligence 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. TSP is famous but the problem posed here is much more realistic are data mining methods learning real knowledge that depends on human trading behaviour, or are they merely learning to predict patterns in arbitrary time series? root optimization when we must salt in order for the road not to freeze One of the drawbacks of GP is that, because of its great flexibility and ability to model such a complex family of functions, it is unfortunately a very slow algorithm, attempting to cover an enormous, heavily-exponential search space. It is also very susceptible to suboptimal local maxima, becoming stuck in areas of solution space that are not close to optimal globally. In order to deal with this issue, and thanks to the eno free lunchf theorem[14,15], we can gain speed and reliability only at the cost of losing one of GPs virtues, namely its extraordinary flexibility. One such compromise is achieved by using Evolutionary Conjunctive Rules (ECR). no free lunch vs random investment instead of usin artificial intelligent (for colin) Even though people who make $150,000 are considerably happier than those who make $40,000, it's not clear why, says psychologist Richard E. Lucas of Michigan State University. Does money make you happier? Or does being happier in the first place allow you to earn more money later, maybe by way of greater creativity or energy? Or does some other factor produce both money and happiness? There's evidence for all three interpretations, Lucas says. (04) K. Madani, M. Voiry, V. Amarger, N. Kanaoui, A. Chohra, and F. Houbre Computed aided diagnosis using soft-computing techniques and image's issued representation: application to biomedical and industrial fields. (14) Volodymyr Turchenko, Viktor Demchuk, Anatoly Sachenko SIMULATION MODELING OF INTERPLANETARY SHOCKS ARRIVAL TIME PREDICTION ON HISTORICAL DATA SET === NN and its learning === (05) Jean-Jacques MARIAGE "Holistic self-reprogramming of neural networks: Between self-organization and self-observation." - Throughout this paper, we non-exhaustively reviewed some essential properties observed in the development of encoding structures in nature. Studying the process by which new structures can emerge in artificial unsupervised NNs models amounts to incorporate a control dynamic of their evolution to the learning algorithms. - Our main research concern is to model and simulate the dynamic character of learning struchures and processes, and their evolution ... Neural darwinism proposed by G. edleman are reffered by jj as - "Brain develops by a Darwinian selection process that takes place at the neuron groups' level instead of the individual neuron. - It would henceforth become possible to develop really autonomous tools that automatically initialize, learn permanently and forget when necessary, while accordingly adapting their structure. - Another point of interest is to look if the uncommitted cells in SOM, which may correspond to intermediary steps of organization, could be used as a SCs reserve and whether they could constitute a pre-learned basis from which further learning could start instead of starting tabula rasa. - needle in a haystack context - Adaptation thus provides a means to take self-organizing opportunities acting as attractors, to drive random variation towards effi-cient forms of organization. This avoids the needle in a haystack problem and enables natural selection to run across the hills and valleys of a varied land-scape to find basins of attractors, where it is easy to fall. Indeed, self-organizing processes are not very brittle. Their wide range of diversity, their spontane-ous arising, the structured emergence of their or-ganization states, as well as the broad spectrum of the initial conditions for their triggering, seem to in-dicate that attraction basins constitute wide and nu-merous areas in the state space. (06) D. A. Bendersky, J. M. Santos "Learning from the environment with a universal reinforcement function." (07) Poramate Manoonpong, Frank Pasemann, Hubert Roth "A modular neurocontroller for a sensor - Driven reactive behavior of biologically inspired walking machines." Accoding to the recent report by AFP, (08) Robert E. Hiromoto, and Milos Manic "Information-based algorithmic design." INFORMATION-BASED ALGORITHMIC DESIGH => INFORMATION-BASED ALGORITHMIC DESIGN OF A NEURAL NETWORK CLASSIFIER === application of NN === Yet another series of applications but more specific purpose of (09) M. Tabedzki and K. Saeed. "Handwritten script and word recognition -- A view-based approach." (10) A. Doudkin, and A. Inyutin "The defect and project rules inspection on PCB layout image." (11) R. Kh.Sadykhov, and D. V. Lamovsky "Estimation of the cross correlation based optical flow for video surveillance." - Proposal of an algorithm which is exploting prallel processors - Although since authors wrore, "All calculations ... were performed ... on a processor ... Pentium 4 ...," it was a simulation of the algorithm on multi-processors === network intrusion detection === The final topic of this issue is network intrusion detection. Here we have two papers: The one is optimistic to the topic while the other pesimistic. (12) Vladimir Golovko, and Leanid Vaitsekhovich "Neural network approaches for intrusion detection and recognition." (13) Akira Imada "How many parachutists will be needed to find a needle in a pastoral -- who is a luckey one?" There is no panacea to cure the troubles of the Middle East. Its metabolism is self-poisoning. Everyone in the community know there is no panacea to detect The author wrote quating an newspaper article "most dangerous..." author asume autlier's atack data is just among the normal data like a needle in a haystack. And we have no efficient algorithm to locat a needle in a hay more effectivly than a random/exahaustive searcn needle colin alos pointed out heinz and bob suggeste a hint hienz The problem of the exponential explosion has been solved in the 80's. For singly connected Bayesian networks exact inference is possible in one sweep of Pearl's belief propagation algorithm maybe before jj's _______________________________________________ HEAD-1207: FEATURE (Thursday, December 7, 2006) Home and away: Bat uses magnetic compass for long flights PARIS - AFP Scientists believe a species of bat has an inbuilt magnetic compass to find its way home over long distances, in addition to its famous echolocation, which guides it around its neighborhood. Princeton University batologists used radio telemetry aboard a small aircraft to track big brown bats (Eptesicus fuscus) that were released 20 kilometers (12 miles) north of their home. They first tested a "control" group of bats, which headed due south towards the roost without a problem. Two other groups of bats were then exposed to a false magnetic field for 90 minutes, comprising 45 minutes before and 45 minutes after sunset. One field was 90 degrees clockwise and the other was 90 degrees anticlockwise from magnetic north. The point of this was to see whether the bats used Earth's magnetic field as a guide and, if so, to see whether the bats used sunset or the stars as an additional cue. The "clockwise" group of bats flew due east, while the "anticlockwise" group went due west, suggesting that they had been using a magnetic compass that may have been calibrated by the sunset. Some of the bats, though, corrected their course and arrived home safely, which implies that, like homing pigeons, they can make a fix if a navigational glitch occurs. The research appears today in Nature, the weekly British science journal. Meanwhile, another study, also published by Nature, sheds light on a remarkable bat species whose tongues are one and half times longer than their bodies. The nectar bat (Anoura fistulata) lives in the cloud forests of the Ecuadorean Andes, feeding on flowers whose nectar is hidden at the end of long funnels. The bat not only depends on these plants, called corollas, for its survival, it also pollinates them -- an example of convergent evolution whereby two unrelated organisms independently evolve similar traits. _____________________________________________ HEAD-1201: FEATURE (Friday, December 1, 2006) South Korean robot will walk the walk as well as talk the talk SEOUL - AFP South Korean scientists are working on a new-generation robot resembling a human that will be able to walk the walk as well as talk the talk, one of the team said Thursday. The first walking "android" will make its debut within two to three years, said So Byung-Rok, one of the team of researchers at the Korea Institute of Industrial Technology. Androids present particular technological challenges in cramming complicated modules, motors and actuators into a life-size body. The team has already developed two android prototypes designed to look like a Korean woman in her early 20s, which can hold a conversation, make eye contact and express anger, sorrow and joy. The latest version, named EveR2-Muse, was unveiled last month at a robot exhibition in Seoul. She made her debut billed as "the world's first android entertainer" singing a new Korean ballad "I Will Close My Eyes for You." [eng] (general) The main debate, he said, is how strong the effect is. "Standing like a human, she can sing a song and move her arms, hips and knees to the rhythm although she cannot lift her feet or walk yet," So told AFP in reference to EveR2-Muse. "We are now working to improve the motion and upgrade intelligence so that next-generation androids can walk like a human, engage in more sophisticated conversations and have a wider range of facial expressions." Researchers at Japan's Osaka University unveiled their own female android capable of speaking and making gestures but So said it was only mobile from the waist up. "We are using electric motors built underneath the silicon skin but a Japanese android uses air pressure to move. The Japanese android moves faster from the waist up than ours does but it needs a giant air tank attached to it," he said. before heinz's turing test _______________________________________________ HEAD-1207: FEATURE (Thursday, December 7, 2006) Deep Fritz chess computer outplays human world champion BERLIN - AFP Deep Fritz, the world's leading chess computer, on Tuesday roundly beat its human counterpart, the Russian world chess champion Vladimir Kramnik, in a six-game encounter. Deep Fritz won by four points to two, after taking the last game in 47 moves in an epic battle lasting almost five hours. Of the six games, Deep Fritz won two and four ended in draws. The 31-year-old Kramnik, who received $500,000 (370,500 euros) for playing the mighty machine, could have walked away with double that sum if he had won. But the world champion had acknowledged even before the opening moves that Deep Fritz was "the clear favorite." In October 2002, Kramnik had held Deep Fritz to a draw after eight games, but the chess software has since been updated, becoming vastly more powerful. Chess experts gathered at the match venue in Bonn, western Germany, agreed that Kramnitz performed well against the machine, made by German software company Chessbase, capable of analyzing eight to 10 million positions per second. But in the second game Kramnitz had a surprising lapse of attention in the 35th move, allowing Deep Fritz to take that game, before wearing the world champion down again in the marathon final encounter. Against a computer, "you can't afford to make the smallest tactical error, because tactically, the computer is unbeatable. You pay for the tiniest error immediately," Kramnitz said. Putting on a brave face, the Russian said he felt he could still put up a good fight "against this incredibly strong opponent." But experts say rapid technical progress means that humans will be completely outclasssed by chess computers before long, making encounters between man and machine so one-sided that they will soon become a thing of the past. material-5 maybe before jj's _______________________________________________ HEAD-1207: FEATURE (Thursday, December 7, 2006) Home and away: Bat uses magnetic compass for long flights PARIS - AFP Scientists believe a species of bat has an inbuilt magnetic compass to find its way home over long distances, in addition to its famous echolocation, which guides it around its neighborhood. Princeton University batologists used radio telemetry aboard a small aircraft to track big brown bats (Eptesicus fuscus) that were released 20 kilometers (12 miles) north of their home. They first tested a "control" group of bats, which headed due south towards the roost without a problem. Two other groups of bats were then exposed to a false magnetic field for 90 minutes, comprising 45 minutes before and 45 minutes after sunset. One field was 90 degrees clockwise and the other was 90 degrees anticlockwise from magnetic north. The point of this was to see whether the bats used Earth's magnetic field as a guide and, if so, to see whether the bats used sunset or the stars as an additional cue. The "clockwise" group of bats flew due east, while the "anticlockwise" group went due west, suggesting that they had been using a magnetic compass that may have been calibrated by the sunset. Some of the bats, though, corrected their course and arrived home safely, which implies that, like homing pigeons, they can make a fix if a navigational glitch occurs. The research appears today in Nature, the weekly British science journal. Meanwhile, another study, also published by Nature, sheds light on a remarkable bat species whose tongues are one and half times longer than their bodies. The nectar bat (Anoura fistulata) lives in the cloud forests of the Ecuadorean Andes, feeding on flowers whose nectar is hidden at the end of long funnels. The bat not only depends on these plants, called corollas, for its survival, it also pollinates them -- an example of convergent evolution whereby two unrelated organisms independently evolve similar traits. <> HEAD-1207: BUSINESS REPORT (Thursday, December 7, 2006) World's richest 2 pct own half of global wealth ISTANBUL - TDN with wire dispatches The world's richest 2 percent of adults own more than half of global household wealth, while half the world's population owns only 1 percent, a U.N. report published on Tuesday showed. <> NYT-1216: EDITORIALS When Doing Nothing Is Better Almost nothing of consequence happened during this week's special session of New York's Legislature, which in this case was real progress.