inteligenta_artificiala_dzitac

I O A N D Z I A C

I N T E L I G E N A R T I F I C I A L

2008

Ioan Dziac, Inteligen artificial, 2008

II

Titlul crii: INTELIGEN ARTIFICIAL Book title: ARTIFICIAL INTELLIGENCE

Descrierea CIP a Bibliotecii Naionale a Romniei

DZIAC, IOAN Inteligen artificial / Ioan Dziac. - Arad : Editura Universitii Aurel Vlaicu, 2008 ISBN 978-973-752-292-4

004.42

Ioan Dziac

http://dzitac.rdsor.ro Domenii principale de interes:

Calcul paralel i distribuit, Inteligen artificial, Metode fuzzy i neuro-fuzzy

Ioan Dziac, Miu-Jan Manolescu,

Lotfi A. Zadeh, Florin Gheorghe Filip

Ioan DZIAC (n. 14.02.1953, Poienile de sub Munte, Maramure), este absolvent al Facultii de Matematic i Informatic al Universitii Babe-Bolyai din Cluj-Napoca (1977), unde a obinut i titlul doctor n informatic (2002) cu teza Procedee de calcul paralel i distribuit n rezolvarea unor ecuaii operatoriale. n perioada 1977-1991 a predat matematic n nvmntul preuniversitar, obinnd titlul de profesor evideniat n 1988 i gradul didactic I cu lucrarea Utilizarea calculatorului n predarea - nvarea matematicii (1990). n perioada 1991-2003 a ocupat prin concurs un post de lector universitar la Universitatea din Oradea, iar apoi cel de confereniar universitar (2003-2005). n perioada 2004 -2005 a fost directorul Departamentului de Matematic i Informatic al Universitii din Oradea. n prezent este confereniar universitar i directorul centrului de cercetare Tehnologii informatice avansate n management i inginerie la Universitatea Agora din Oradea, unde a fondat n 2006, alturi de acad. F.G. Filip i prof. Miu-Jan Manolescu, revista International Journal of Computers, Communications and Control, prima revist romneasc de Computer Science cotat ISI Web of Science (ncepnd cu numrul suplimentar din 2006).

A fondat conferina International Conference Computers, Communications and Control (ICCCC 2006), care la ediia ICCCC 2008 s-a bucurat de prezena printelui mulimilor i a logicii fuzzy, Lotfi A. Zadeh, alturi de care a editat cartea Lotfi A. Zadeh, Dan Tufis, Florin Gheorghe Filip, Ioan Dzitac (eds.), From Natural Language to Soft Computing: New Paradigms in Artificial Intelligence , Editura Academiei Romne, ISBN:978-973-27-1678-6, 2008 .


III

PREFA

nva de la toate... (Din lirica norvegian)

1. De ce n titlul crii apare Inteligen artificial i nu Inteligena artificial?

Rspunsul este simplu: a doua variant ar putea sugera o abordare

exhaustiv a Inteligenei Artificiale (IA), ceea ce ar fi prea ambiios i dificil. Aici ne vom limita doar la un studiu introductiv n IA. Vom face o incursiune superficial (i inegal), fr a intra n amnunte tehnice ale tuturor domeniilor i subdomeniilor IA. Vom insista mai mult pe aspectele care implic logica fuzzy, deoarece raionamentul bazat pe acest nou tip de logic (nuanat) a deschis noi perspective n soluionarea problemelor pentru care nu dispunem de modele matematice exacte. 2. Care este scopul acestei cri i cui se adreseaz?

Prezenta carte se adreseaz n primul rnd studenilor de la programul de studii de licen, anul III, de la Universitatea Aurel Vlaicu din Arad, Facultatea de tiine Exacte, servindu-le drept suport de curs, dar poate fi util oricrui cititor care dorete s se familiarizeze cu preocuprile i aplicaiile IA.

Cartea este scris cu scopul declarat de a trezi interesul studenilor, i nu numai, pentru studiul IA, ca studiu independent sau n cadrul unor programe licen, masterat i doctorat din nvmntul superior, motiv pentru care va semna pe alocuri mai mult cu o carte de popularizare a tiinei dect cu o lucrare tiinific propriu-zis. Aceast abordare are o dubl motivaie: prima este de natur didactic n spiritul principiului accesibilitii, iar a doua este de natur metodologic n spiritul familiarizrii cu metodele de lucru cu noiuni imprecise, specifice soft computing-ului. 3. Ce este IA?

Dac vei pune aceast ntrebare la 10 informaticieni, vei primi 10 rspunsuri diferite. Cert este c IA, ca ramur tiinific, se poate ncadra n clasa mai larg a informaticii/tehnologiei informaiei/tiinei calculatoarelor (nici ntre aceste 3 tiine nu exist o delimitare clar i uneori se folosesc ca sinonime). Dar, vom vedea c IA este o tiin interdisciplinar, chiar


IV

multidisciplinar, implicnd rezultate din alte tiine ca matematica, biologia, psihologia, filozofia etc. IA ncearc s imite inteligena natural.

Pentru a evita discuiile teoretice, de natur filozofic, asupra sensului sau nonsensului IA, ni se pare mai productiv s utilizm atributul artificial n coresponden strict cu atributul natural (Brbat, 2002), pentru a desemna o entitate elaborat de om care imit, mai mult sau mai puin, o entitate corespondent care exist i n natur. De exemplu, mierea produs de albine exist n natur, este un produs natural i o numim miere natural (miere de albine), dar exist i miere artificial, fabricat de om. Atributul artificial nu se utilizeaz pentru produse care nu exist i n natur. De exemplu, telefonul nu este un produs artificial, dei este produs de om, deoarece nu exist telefoane produse pe cale natural (ns telefonul poate avea n componena sa elemente inteligente).

Vom utiliza sintagma inteligen artificial pentru a desemna clasa unor produse concepute de om (sisteme, ageni, programe de calculator, maini, automate, roboi, etc.), care imit inteligena natural: creierul uman, reeaua neuronal, membrana celulei, inteligena comportamental emergent grupurilor sociale non-umane (colonii de furnici, roiuri de albine, stoluri de psri, bancuri de peti, haite de lupi etc.) i altele.

4. De unde am cules informaia?

Aceast carte este un studiu bazat pe alte cri i articole tiinifice, studiu trecut prin filtrul autorului i actualizat, pe ct posibil, fa de crile de referin. Este ceea ce se numete ndeobte un curs de autor. Dar, chiar n timp ce scriu aceast carte apar alte nouti n IA, ca urmare invit cititorii s-i actualizeze singuri noutile din tot mai multe surse web demne de ncredere.

Am folosit n documentare n mod special cartea [Br02], n limba romn, Sisteme inteligente orientate spre agent, scris de Boldur E. Brbat (Ed. Academiei, 2002), precum i cartea [RuN03], n limba englez, Artificial Intelligence: A Modern Approach, scris de Stuart Russell i Peter Norvig (ed. a II-a, Prentice Hall 2003), mprosptnd informaia cu alte surse, cri i reviste existente n biblioteci reale i virtuale (lista complet a surselor de informaie este precizat n bibliografie sau/i n notele de subsol).

Am lsat multe cuvinte i expresii netraduse din limba englez, fie c nu am putut gsit traducerea potrivit, fie c am considerat c cititorii ar trebui s cunoasc limba englez.

Recomand cu cldur cele dou cri i cititorilor care vor s afle mai multe despre IA, precum i celelalte surse bibliografice i webografice.

Ioan Dziac


V

CUPRINS

O INTRODUCERE SUMAR N IA 1. CE este IA? Testul Turing, cuvinte i sintagme cheie 2. DE CE studiem, cercetm i producem IA? 3. UNDE se studiaz, cerceteaz i promoveaz IA ? 4. CINE? Autori, conductori de doctorate, cercettori 5. CUM? Metode, modele, mijloace, tehnologii

1 2 3 4 12 26

CAPITOLUL 1. INTELIGENA ARTIFICIAL CA IMITAIE APROXIMATIV A INTELIGENEI NATURALE 1.1. Inteligena natural uman 1.1.1. Generaliti 1.1.2. Inteligene multiple 1.1.3. Inteligena cognitiv (IQ). Teste psihometrice pentru IQ 1.1.4. Inteligena emoional (EQ). Inteligena social 1.2. Inteligena natural emergent grupurilor sociale de animale (Swarm Intelligence) 1.3. Inteligena artificial (IA) 1.3.1. Definiii. Testul Turing 1.3.2. Obiectul de studiu al IA 1.3.3. Istoricul IA 1.3.4. Domeniile de cercetare i aplicaie ale IA

1.3.4.1. Raionamentul logic 1.3.4.2. Reprezentarea cunoaterii 1.3.4.3. Percepia 1.3.4.4. Calcul evolutiv. Algoritmi genetici 1.3.4.5. Reele neurale 1.3.4.6. Teoria jocurilor 1.3.4.7. nvarea automat (Machine learning) 1.3.4.8. Ageni inteligeni 1.3.4.9. Sisteme expert 1.3.4.10. Sisteme fuzzy

1.3.5. Sisteme i maini inteligente 1.3.5.1. Deep Blue 1.3.5.2. Sisteme de traducere automat 1.3.5.3. Optical Character Recognition 1.3.5.4. DENDRAL 1.3.5.5. Roboi

1.3.6. Computere SPRAY: praf inteligent, vopsea inteligent

31 31 31 33 34 36 39 42 42 45 47 51 51 51 52 52 53 54 54 55 56 57 57 57 58 58 58 58 59


VI

CAPITOLUL 2. SOFT COMPUTING 2.1. Raionament nuanat. Logic fuzzy 2.1.1. Generaliti 2.1.2. Logica fuzzy 2.1.3. Mulimi crisp. Mulimi fuzzy 2.1.4. Submulimi fuzzy 2.1.5. Relaii ntre mulimi fuzzy 2.1.6. Numere fuzzy 2.1.7. Operaii cu mulimi fuzzy 2.1.8. Implicaii pentru logica fuzzy 2.1.9. Operatori de compunere fuzzy 2.1.10. Fuzificarea i defuzificarea informaiei 2.1.11. Evaluarea i aprecierea. Scale liniare i neliniare 2.2. Calcul neural. Reele neurale. Reele neuro-fuzzy 2.2.1. Uniti funcionale (de procesare) ale reelelor neurale 2.2.2. Arhitectura reelelor neurale 2.2.3. Algoritmi de funcionare i nvare a unei reelele neurale 2.2.4. Reele neuro-fuzzy. Un exemplu de aplicaie neuro-fuzzy 2.2.4.1. Modelarea neuro-fuzzy 2.2.4.2. Utilizarea modelelor neuro-fuzzy la predicia unor

evenimente 2.2.4.3. Aplicaie privind predicia intervalelor de timp la

care pot s apar evenimente n cadrul unui sistem energetic 2.3. Calcul evolutiv. Algoritmi genetici 2.4. Raionament probabilist. Reele bayesiene 2.4.1. Metoda bayesian 2.4.2. Abordarea bayesian 2.4.3. Limitele metodei bayesiene 2.4.4. Comparaie ntre abordarea bayesian i factorii de

certitudine 2.4.5. Reele bayesiene 2.5. Teoria nvrii. Machine Learning 2.5.1. Noiuni de teoria nvrii 2.5.2. Despre date 2.5.3. Despre informaii 2.5.4. Despre cunotine 2.5.5. Inferena ca proces de cutare i constrngerile care apar

pentru satisfacerea condiiilor 2.5.6. Varietatea metodelor de raionare i rolul variabilelor 2.5.7. Metode din ingineria cunoaterii pentru nvarea

automatelor 2.5.8. nvarea inductiv. nvarea prin exemple

61 62 62 64 69 71 73 73 80 81 83 85 86 97 98 102 104 105 107

108 109 114 118 118 118 121 122

124 126 128 128 131 133 134

136 137

138 139


VII

2.5.9. Alte metode de nvare ale automatelor 2.6.Teoria haosului. Fractali 2.6.1. Despre teoria haosului 2.6.2. Fractali

140 140 140 142

CAPITOLUL 3. SISTEME BAZATE PE CUNOTINE. SISTEME EXPERT 3.1. Exemple de sisteme expert 3.2. Sisteme expert bazate pe reguli 3.2.1. Despre cunotine 3.2.2. Regulile ca i tehnic de reprezentare a cunotinelor 3.2.3. Structura unui sistem expert bazat pe reguli 3.2.4. Caracteristicile fundamentale ale unui sistem expert 3.2.5. Avantajele i dezavantajele unui sistem expert bazat pe

reguli 3.2.6. Managementul incertitudinii n sistemele expert bazate

pe reguli 3.3. Sisteme expert fuzzy 3.3.1. Generaliti 3.3.2. Reguli fuzzy 3.3.3. Inferene fuzzy 3.3.4. Construirea unui sistem expert fuzzy 3.4. Proiectarea sistemelor expert

155 155 159 159 160 161 164

166

167 169 169 170 172 172 175

BIBLIOGRAFIE 177


VIII

LIST DE ABREVIERI

AAAI - American Association for Artificial Intelligence ACM - Association for Computing Machinery AIAI - Artificial Intelligence Applications Institute AT&T - Bell Labs AUAI - The Association for Uncertainty in Artificial Intelligence () CLIPS - C Language Integrated Production System EC - Evolutionary Computation EQ - Inteligena emoional FL - Fuzzy Logic IA - Inteligen artificial IQ - Intelligence quotient IN Inteligen natural ISI - Institute of Scientific Information MIT - AI lab at Massachusetts Institute of Technology ML - Machine Learning NC - Neural Computing OCR - Optical Character Recognition OPS - Official Production System PR - Probabilistic Reasoning SC - Soft Computing SE - Sistem expert TI - Tehnologia Informaiei

O INTRODUCERE SUMAR N IA

n aceast lucrare INTELIGENA ARTIFICIAL (IA) se va considera

ca fiind o imitaie aproximativ a INTELIGENEI NATURALE (IN). n aceast introducere vom ncerca s descrierem succint problematica i

cteva aspecte eseniale ale IA n raport cu IN (v. fig.1 i 2).

Figura 1. Inteligena artificial - imitaie aproximativ a inteligenei naturale


2

Figura 2. Problematica IA 1. CE este IA? Testul Turing, cuvinte i sintagme cheie

IA este tiina care i propune s gseasc soluii la probleme complexe care erau apanajul inteligenei umane, cu ajutorul programelor de calculator sau a mainilor automate.

Dei IA ca tiin, este considerat, n general, ca o ramur a Informaticii, Tehnologiei Informaiei sau a tiinei Calculatoarelor, trebuie s evideniem legturile sale puternice cu alte tiine, cum ar fi Matematica (Logica, Teoria probabilitilor), Psihologia (Inteligena uman, Teoria nvrii), Medicina (Neurotiinele), Gnoseologia (Teoria cunoaterii), Biologia, Filozofia i multe altele. Abilitatea de a combina cunoaterea din toate aceste domenii duce la progres n crearea IA.

La nceput, crearea i cercetarea IA s-a desfurat pe terenul psihologiei, punndu-se accent pe inteligena lingvistic, ca de exemplu la testul Turing (v. fig. 3). Acest test const ntr-o conversaie n limbaj uman natural cu o main (computer, program) care a fost programat special pentru acest test. Exist un juriu uman care converseaz cu acest computer, dar i cu un om, prin cte un canal pur text (fr ca ei s se vad sau s se aud, dac interlocutor 1 este maina, atunci interlocutor 2 este omul i invers). n cazul n care juriul nu poate s-i dea seama care este computerul i care omul, atunci inteligena artificial (programul/calculatorul) a trecut testul. n acest mod, Turing a dat o definiie implicit a IA, evitnd disputele filozofice i rigorile unei definiii formale.


3

Figura 3. Schema testului Turing

Cuvinte i sintagme cheie n IA: inteligen artificial (artificial

intelligence), reele neurale (neural networks), reele semantice (semantic networks), reele bayesiene/reele de ncredere (Bayesian networks/belief networks), sisteme bazate pe cunotine (knowledge based systems), sisteme expert (expert systems), sisteme fuzzy (fuzzy systems), sisteme hibride, calcul inteligent (intelligent computing), calcul natural (natural computing), calcul evolutiv (evolutionary computing), algoritmi genetici (genetic algorithms), calcul sumar (soft computing), calcul emergent (swarm computing), ants computing (calcul de tip muuroi), calcul membranar (membrane computing), calcul n limbaj natural (natural language computing), agent, multi-agent, robot, colonie de roboi, nvare automat (machine learning), test Turing etc. Lsm deschis lista unor ntrebri de natur filozofic, unele chiar ocante: Exist IA? Este ea necesar? Nu este periculoas pentru omenire? 2. DE CE studiem, cercetm i producem IA?

Figura 4. Coperta Revistei Engineering Applications of Artificial Intelligence1

1 Sursa: http://www.sciencedirect.com/science/journal/09521976


4

Argument. Dintre cele mai cunoscute i utile aplicaii ale IA, care duc la progres (n tiin, tehnologie, economie i viaa social), putem enumera:

rezolvarea automat a problemelor, demonstrarea automat a teoremelor, traducerea automat a textelor, ageni inteligeni, recunoaterea formelor/tiparelor (pattern recognition), voce electronic, sisteme expert, robotic, asisten inteligent n sntate, nvare automat (machine lerning) instruire asistat de calculator, marketing i management asistat, telecomunicaii bazate pe computer, internet i aplicaii web, motoare de cutare, inginerie, proiectare asistat jocuri i altele.

3. UNDE se studiaz, cerceteaz i promoveaz IA ?

Programe de licen (ciclul I universitar) n Romnia: IA se studiaz mai nou, de cele mai multe ori ca disciplin obligatorie, n

cadrul a tot mai multe specializri universitare de licen din toate centrele universitare care au astfel de programe de studii: informatic, informatic economic, cibernetic, calculatoare, tehnologia informaiei, automatizri, robotic, energetic, etc.

Master n IA (ciclul II universitar, programe acreditate de ARACIS ncepnd cu 2008-2009 n Romnia):

1. Universitatea Babe-Bolyai, Cluj- Napoca: Sisteme inteligente; 2. Universitatea din Bucureti: Inteligen artificial; 3. Universitatea Tehnic Gheorghe Asachi Iai, Sisteme inteligente; 4. Universitatea din Craiova: Metode i modele n inteligen

artificial; 5. Universitatea Dunrea de Jos Galai: Sisteme inteligente; 6. Universitatea de Vest din Timioara: Inteligen artificial i calcul

distribuit (Informatic); Nano-Microsisteme inteligente (Fizic).


5

Cercetare n Romnia: 1) Institutul de Cercetri pentru Inteligen al Academiei Romne

http://www.racai.ro 2) Centrul de Cercetri Cognitive i Neuronale (Coneural, institut privat)

http://www.coneural.org

Forumuri n Romnia: 1) http://www.inteligenta-artificiala.ro/ 2) Clubul de inteligen artificial a Facultii de Informatic din Iai

(Universitatea A.I. Cuza) http://profs.info.uaic.ro/~alaiba/club-ai/index.php?title=Pagina_principal%C4%83

Reviste romneti cotate ISI, care public articole tiinifice din domeniul IA:

1) International Journal of Computers, Communications and Control (apare n Editura Universitii Agora, Oradea din 2006, fondatori Ioan Dziac (Editor executiv), Florin-Gheorghe Filip- Editor ef i M.-J. Manolescu - Managing editor, cu articole introduse n ISI Web of Science ncepnd cu Vol.1, numrul suplimentar din 2006), http://www.univagora.ro

2) Romanian Journal of Information Science and Technology (apare n Editura Academiei Romne, Bucureti din 1998, Editor ef - Dan Dasclu, Editor executiv - Gheorghe Pun, cu articole introduse n ISI Web of Science ncepnd cu Vol.10, Nr.1/2007), http://www.imt.ro/romjist/

Asociaii i institute internaionale de cercetare remarcabile: 1) AAAI: American Association for Artificial Intelligence http://www.aaai.org/home.html 2) ACM: the Association for Computing Machinery http://www.acm.org/ 3) AIAI: Artificial Intelligence Applications Institute http://www.aiai.ed.ac.uk/ 4)AT&T Bell Labs http://www.research.att.com/ 5) Carnegie Mellon University Artificial Intelligence Repository http://www.cs.cmu.edu/afs/cs.cmu.edu/project/ai-repository/ai/0.html 6) MIT: AI lab at Massachusetts Institute of Technology http://www.csail.mit.edu/ 7) IJCAI Home Page http://ijcai.org/ 8) The Association for Uncertainty in Artificial Intelligence (AUAI) http://www.auai.org/


6

Mari universiti din lume unde se studiaz IA: 1) MIT: Massachusetts Institute of Technology: http://web.mit.edu/ 2) University of California Berkeley: http://berkeley.edu/ 3) Stanford University: http://www.stanford.edu/ 4) Oxford University: http://www.robots.ox.ac.uk/~aisoc/ 5) University of Edinburgh: http://www.inf.ed.ac.uk/ 6) Yowa State University: http://www.cs.iastate.edu/~honavar/aigroup.html

Cele mai prestigioase reviste internaionale care public articole tiinifice n domeniul IA:

Sursa: http://www.thomsonscientific.com/cgi-bin/jrnlst/jlsubcatg.cgi?PC=D

Data accesrii: 02.12.2008

SCIENCE CITATION INDEX EXPANDED - COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE - JOURNAL LIST

Total journals: 100

1. ADAPTIVE BEHAVIOR 2. ADVANCED ENGINEERING INFORMATICS 3. AI COMMUNICATIONS 4. AI EDAM-ARTIFICIAL INTELLIGENCE FOR ENGINEERING DESIGN ANALYSIS AND MANUFACTURING 5. AI MAGAZINE 6. ANNALS OF MATHEMATICS AND ARTIFICIAL INTELLIGENCE 7. APPLIED ARTIFICIAL INTELLIGENCE 8. APPLIED INTELLIGENCE 9. APPLIED SOFT COMPUTING 10. ARTIFICIAL INTELLIGENCE 11. ARTIFICIAL INTELLIGENCE IN MEDICINE 12. ARTIFICIAL INTELLIGENCE REVIEW 13. ARTIFICIAL LIFE 14. AUTONOMOUS AGENTS AND MULTI-AGENT SYSTEMS 15. AUTONOMOUS ROBOTS 16. CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS 17. COGNITIVE SYSTEMS RESEARCH 18. COMPUTATIONAL INTELLIGENCE 19. COMPUTATIONAL LINGUISTICS 20. COMPUTER SPEECH AND LANGUAGE 21. COMPUTER VISION AND IMAGE UNDERSTANDING 22. COMPUTING AND INFORMATICS 23. CONNECTION SCIENCE 24. CONSTRAINTS 25. DATA & KNOWLEDGE ENGINEERING 26. DATA MINING AND KNOWLEDGE DISCOVERY 27. DECISION SUPPORT SYSTEMS


7

28. ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE 29. ENGINEERING INTELLIGENT SYSTEMS FOR ELECTRICAL ENGINEERING AND COMMUNICATIONS 30. EVOLUTIONARY COMPUTATION 31. EXPERT SYSTEMS 32. EXPERT SYSTEMS WITH APPLICATIONS 33. GENETIC PROGRAMMING AND EVOLVABLE MACHINES 34. IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE 35. IEEE INTELLIGENT SYSTEMS 36. IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION 37. IEEE TRANSACTIONS ON FUZZY SYSTEMS 38. IEEE TRANSACTIONS ON IMAGE PROCESSING 39. IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING 40. IEEE TRANSACTIONS ON NEURAL NETWORKS 41. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 42. IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART B-CYBERNETICS 43. IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART C-APPLICATIONS AND REVIEWS 44. IET COMPUTER VISION 45. IMAGE AND VISION COMPUTING 46. INFORMATION FUSION 47. INFORMATION TECHNOLOGY AND CONTROL 48. INTEGRATED COMPUTER-AIDED ENGINEERING 49. INTELLIGENT AUTOMATION AND SOFT COMPUTING 50. INTELLIGENT DATA ANALYSIS 51. INTERNATIONAL JOURNAL OF APPLIED MATHEMATICS AND COMPUTER SCIENCE 52. INTERNATIONAL JOURNAL OF APPROXIMATE REASONING 53. INTERNATIONAL JOURNAL OF COMPUTER VISION 54. INTERNATIONAL JOURNAL OF FUZZY SYSTEMS 55. INTERNATIONAL JOURNAL OF INFORMATION TECHNOLOGY & DECISION MAKING 56. INTERNATIONAL JOURNAL OF INNOVATIVE COMPUTING INFORMATION AND CONTROL 57. INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS 58. INTERNATIONAL JOURNAL OF NEURAL SYSTEMS 59. INTERNATIONAL JOURNAL OF PATTERN RECOGNITION AND ARTIFICIAL INTELLIGENCE 60. INTERNATIONAL JOURNAL OF SOFTWARE ENGINEERING AND KNOWLEDGE ENGINEERING 61. INTERNATIONAL JOURNAL OF UNCERTAINTY FUZZINESS AND KNOWLEDGE-BASED SYSTEMS 62. INTERNATIONAL JOURNAL ON ARTIFICIAL INTELLIGENCE TOOLS 63. INTERNATIONAL JOURNAL ON DOCUMENT ANALYSIS AND RECOGNITION 64. INTERNATIONAL JOURNAL ON SEMANTIC WEB AND INFORMATION SYSTEMS 65. JOURNAL OF ARTIFICIAL INTELLIGENCE RESEARCH 66. JOURNAL OF AUTOMATED REASONING 67. JOURNAL OF CHEMOMETRICS 68. JOURNAL OF COMPUTER AND SYSTEMS SCIENCES INTERNATIONAL


8

69. JOURNAL OF EXPERIMENTAL & THEORETICAL ARTIFICIAL INTELLIGENCE 70. JOURNAL OF HEURISTICS 71. JOURNAL OF INTELLIGENT & FUZZY SYSTEMS 72. JOURNAL OF INTELLIGENT & ROBOTIC SYSTEMS 73. JOURNAL OF INTELLIGENT INFORMATION SYSTEMS 74. JOURNAL OF INTELLIGENT MANUFACTURING 75. JOURNAL OF MACHINE LEARNING RESEARCH 76. JOURNAL OF MATHEMATICAL IMAGING AND VISION 77. JOURNAL OF MULTIPLE-VALUED LOGIC AND SOFT COMPUTING 78. JOURNAL OF WEB SEMANTICS 79. KNOWLEDGE AND INFORMATION SYSTEMS 80. KNOWLEDGE ENGINEERING REVIEW 81. KNOWLEDGE-BASED SYSTEMS 82. MACHINE LEARNING 83. MACHINE VISION AND APPLICATIONS 84. MECHATRONICS 85. MEDICAL IMAGE ANALYSIS 86. MINDS AND MACHINES 87. NETWORK-COMPUTATION IN NEURAL SYSTEMS 88. NEURAL COMPUTATION 89. NEURAL COMPUTING & APPLICATIONS 90. NEURAL NETWORK WORLD 91. NEURAL NETWORKS 92. NEURAL PROCESSING LETTERS 93. NEUROCOMPUTING 94. PATTERN ANALYSIS AND APPLICATIONS 95. PATTERN RECOGNITION 96. PATTERN RECOGNITION LETTERS 97. REAL-TIME IMAGING 98. ROBOTICS AND AUTONOMOUS SYSTEMS 99. SOFT COMPUTING 100. TRAITEMENT DU SIGNAL

SCIENCE CITATION INDEX EXPANDED - COMPUTER SCIENCE, CYBERNETICS - JOURNAL LIST Total journals: 19

1. ACM TRANSACTIONS ON COMPUTER-HUMAN INTERACTION 2. BEHAVIOUR & INFORMATION TECHNOLOGY 3. BIOLOGICAL CYBERNETICS 4. CONTROL AND CYBERNETICS 5. CYBERNETICS AND SYSTEMS 6. HUMAN-COMPUTER INTERACTION 7. IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART A-SYSTEMS AND HUMANS 8. IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART B-CYBERNETICS 9. IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART C-APPLICATIONS AND REVIEWS


9

10. INTERACTING WITH COMPUTERS 11. INTERNATIONAL JOURNAL OF HUMAN-COMPUTER INTERACTION 12. INTERNATIONAL JOURNAL OF HUMAN-COMPUTER STUDIES 13. JOURNAL OF COMPUTER AND SYSTEMS SCIENCES INTERNATIONAL 14. KYBERNETES 15. KYBERNETIKA 16. MACHINE VISION AND APPLICATIONS 17. MODELING IDENTIFICATION AND CONTROL 18. PRESENCE-TELEOPERATORS AND VIRTUAL ENVIRONMENTS 19. USER MODELING AND USER-ADAPTED INTERACTION

SCIENCE CITATION INDEX EXPANDED - COMPUTER SCIENCE, THEORY & METHODS - JOURNAL LIST Total journals: 90 (din acestea am selectat doar 23, care au i profil IA)

1. ACM TRANSACTIONS ON COMPUTATIONAL LOGIC 2. ARTIFICIAL LIFE 3. COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE 4. CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE 5. EVOLUTIONARY COMPUTATION 6. EXPERT SYSTEMS 7. FUZZY SETS AND SYSTEMS 8. GENETIC PROGRAMMING AND EVOLVABLE MACHINES 9. HUMAN-COMPUTER INTERACTION 10. IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION 11. IEEE TRANSACTIONS ON NEURAL NETWORKS 12. IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART A-

SYSTEMS AND HUMANS 13. IMAGE AND VISION COMPUTING 14. INTERNATIONAL JOURNAL OF GENERAL SYSTEMS 15. INTERNATIONAL JOURNAL OF QUANTUM INFORMATION 16. JOURNAL OF CELLULAR AUTOMATA 17. JOURNAL OF LOGIC AND COMPUTATION 18. JOURNAL OF MULTIPLE-VALUED LOGIC AND SOFT COMPUTING 19. JOURNAL OF SYMBOLIC COMPUTATION 20. MULTIDIMENSIONAL SYSTEMS AND SIGNAL PROCESSING 21. QUANTUM INFORMATION & COMPUTATION 22. ROMANIAN JOURNAL OF INFORMATION SCIENCE AND TECHNOLOGY 23. THEORY AND PRACTICE OF LOGIC PROGRAMMING

SCIENCE CITATION INDEX EXPANDED - ROBOTICS & AUTOMATIC CONTROL - JOURNAL LIST Total journals: 58

1. ANNUAL REVIEWS IN CONTROL 2. ASIAN JOURNAL OF CONTROL 3. ASSEMBLY AUTOMATION 4. AT-AUTOMATISIERUNGSTECHNIK 5. AUTOMATICA 6. AUTOMATION AND REMOTE CONTROL


10

7. AUTONOMOUS AGENTS AND MULTI-AGENT SYSTEMS 8. CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS 9. CONTROL & AUTOMATION 10. CONTROL AND CYBERNETICS 11. CONTROL ENGINEERING 12. CONTROL ENGINEERING PRACTICE 13. DISCRETE EVENT DYNAMIC SYSTEMS-THEORY AND APPLICATIONS 14. ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE 15. ESAIM-CONTROL OPTIMISATION AND CALCULUS OF VARIATIONS 16. EUROPEAN JOURNAL OF CONTROL 17. IEEE CONTROL SYSTEMS MAGAZINE 18. IEEE ROBOTICS & AUTOMATION MAGAZINE 19. IEEE TRANSACTIONS ON AUTOMATIC CONTROL 20. IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING 21. IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY 22. IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS 23. IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS 24. IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART B-CYBERNETICS 25. IEEE-ASME TRANSACTIONS ON MECHATRONICS 26. IET CONTROL THEORY AND APPLICATIONS 27. IMA JOURNAL OF MATHEMATICAL CONTROL AND INFORMATION 28. INFORMATION TECHNOLOGY AND CONTROL 29. INTELLIGENT AUTOMATION AND SOFT COMPUTING 30. INTERNATIONAL JOURNAL OF ADAPTIVE CONTROL AND SIGNAL PROCESSING 31. INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY 32. INTERNATIONAL JOURNAL OF APPLIED MATHEMATICS AND COMPUTER SCIENCE 33. INTERNATIONAL JOURNAL OF COMPUTERS COMMUNICATIONS & CONTROL 34. INTERNATIONAL JOURNAL OF CONTROL 35. INTERNATIONAL JOURNAL OF CONTROL AUTOMATION AND SYSTEMS 36. INTERNATIONAL JOURNAL OF FUZZY SYSTEMS 37. INTERNATIONAL JOURNAL OF INNOVATIVE COMPUTING INFORMATION AND CONTROL 38. INTERNATIONAL JOURNAL OF ROBOTICS & AUTOMATION 39. INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL 40. INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE 41. JOURNAL OF CHEMOMETRICS 42. JOURNAL OF DYNAMIC SYSTEMS MEASUREMENT AND CONTROL-TRANSACTIONS OF THE ASME 43. JOURNAL OF DYNAMICAL AND CONTROL SYSTEMS 44. JOURNAL OF MACHINE LEARNING RESEARCH 45. JOURNAL OF PROCESS CONTROL 46. JOURNAL OF SYSTEMS ENGINEERING AND ELECTRONICS 47. JOURNAL OF THE FRANKLIN INSTITUTE-ENGINEERING AND APPLIED MATHEMATICS 48. MATHEMATICS OF CONTROL SIGNALS AND SYSTEMS 49. MEASUREMENT & CONTROL 50. MECHATRONICS


11

51. MODELING IDENTIFICATION AND CONTROL 52. OPTIMAL CONTROL APPLICATIONS & METHODS 53. PROCEEDINGS OF THE INSTITUTION OF MECHANICAL ENGINEERS PART I-JOURNAL OF SYSTEMS AND CONTROL ENGINEERING 54. REVISTA IBEROAMERICANA DE AUTOMATICA E INFORMATICA INDUSTRIAL 55. ROBOTICS AND AUTONOMOUS SYSTEMS 56. SIAM JOURNAL ON CONTROL AND OPTIMIZATION 57. SYSTEMS & CONTROL LETTERS 58. TRANSACTIONS OF THE INSTITUTE OF MEASUREMENT AND CONTROL SCIENCE CITATION INDEX EXPANDED - COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS - JOURNAL LIST Total journals: 96 (din acestea am selectat doar 28, care au i profil IA)

1) AI EDAM-ARTIFICIAL INTELLIGENCE FOR ENGINEERING DESIGN

ANALYSIS AND MANUFACTURING 2) APPLIED SOFT COMPUTING 3) ARCHIVES OF COMPUTATIONAL METHODS IN ENGINEERING 4) BIOINFORMATICS 5) BIOMEDICAL ENGINEERING-APPLICATIONS BASIS COMMUNICATIONS 6) COMPUTATIONAL BIOLOGY AND CHEMISTRY 7) COMPUTATIONAL GEOSCIENCES 8) COMPUTATIONAL LINGUISTICS 9) COMPUTATIONAL STATISTICS & DATA ANALYSIS 10) COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE 11) COMPUTER METHODS IN BIOMECHANICS AND BIOMEDICAL

ENGINEERING 12) IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE 13) IEEE TRANSACTIONS ON MEDICAL IMAGING 14) IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART C-

APPLICATIONS AND REVIEWS 15) IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND

BIOINFORMATICS 16) INTERNATIONAL JOURNAL OF INFORMATION TECHNOLOGY & DECISION

MAKING 17) INTERNATIONAL JOURNAL ON ARTIFICIAL INTELLIGENCE TOOLS 18) JOURNAL OF BIOMEDICAL INFORMATICS 19) JOURNAL OF COMPUTATIONAL BIOLOGY 20) MEDICAL & BIOLOGICAL ENGINEERING & COMPUTING 21) MEDICAL IMAGE ANALYSIS 22) NEUROINFORMATICS 23) QSAR & COMBINATORIAL SCIENCE 24) QUEUEING SYSTEMS 25) ROBOTICS AND COMPUTER-INTEGRATED MANUFACTURING 26) SAR AND QSAR IN ENVIRONMENTAL RESEARCH 27) SOFT COMPUTING 28) SPEECH COMMUNICATION


12

SCIENCE CITATION INDEX EXPANDED - COMPUTER SCIENCE, INFORMATION SYSTEMS - JOURNAL LIST Total journals: 104 (din acestea am selectat doar 16, care au i profil IA)

1. ACM TRANSACTIONS ON COMPUTER-HUMAN INTERACTION 1. DATA & KNOWLEDGE ENGINEERING 2. DATA MINING AND KNOWLEDGE DISCOVERY 3. DECISION SUPPORT SYSTEMS 4. IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE 5. IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING 6. INFORMATION TECHNOLOGY AND CONTROL 7. INTERNATIONAL JOURNAL OF COMPUTERS COMMUNICATIONS &

CONTROL 8. INTERNATIONAL JOURNAL OF INFORMATION TECHNOLOGY & DECISION

MAKING 9. INTERNATIONAL JOURNAL OF MEDICAL INFORMATICS 10. INTERNATIONAL JOURNAL ON SEMANTIC WEB AND INFORMATION

SYSTEMS 11. JOURNAL OF SIGNAL PROCESSING SYSTEMS FOR SIGNAL IMAGE AND

VIDEO TECHNOLOGY 12. JOURNAL OF THE AMERICAN MEDICAL INFORMATICS ASSOCIATION 13. JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION 14. JOURNAL OF WEB SEMANTICS 15. KNOWLEDGE AND INFORMATION SYSTEMS 16. MEDICAL INFORMATICS AND THE INTERNET IN MEDICINE

Observaie. Dintr-o list de 584 de reviste de COMPUTER SCIENCE incluse n baza de date ISI de Thomson Reuters, la categoria SCIENCE CITATION INDEX EXPANDED, un numr de 244 de reviste public articole tiinifice despre IA, avnd denotaia sau conotaia chiar n titlul revistei. n realitate numrul revistelor care public articole tiinifice despre IA este cu mult mai mare. Titlurile revistelor de mai sus dovedesc marea varietate a subdomeniilor IA i ponderea mare pe care o ocup IA n cercetarea i aplicaiile actuale. 4. CINE? Autori, conductori de doctorate, cercettori Cri de referin din domeniul IA din Romnia:

1) N. ndreanu: -Introducere in programarea logica. Limbajul Prolog, Editura Intarf, 1994;

2) D. Dumitrescu, H.Costin: - Reele neuronale, Teorie i aplicaii, Ed. Teora, Bucureti,1996;

3) H. Luchian: - Clasificare evolutiv, Iasi, 1999; 4) D. Dumitrescu: -Principiile inteligenei artificiale, Ed. Albastr, Cluj-

Napoca, 1999;


13

5) N. ndreanu: - Sisteme Expert. Reprezentarea cunotintelor i inferena, Editura Universitaria, 2001, 344 p.;

6) R. Andonie, A. Caaron, Inteligen computaional, Editura Univ. Transilvania Braov, 2002;

7) B. E Brbat: - Sisteme inteligente orientate spre agent, Ed. Academiei Romne, 2002;

8) N. ndreanu: - Baze de cunostine, Editura Sitech, 2004; 9) H.-N. Teodorescu, M. Zbancioc, O. Voroneanu, Sisteme bazate pe

cunotinte. Aplicaii, Editura Performantica, Iasi, 2004; 10) T.I. Bjenescu, Performantele inteligentei artificiale - de la teorie la

aplicaii (ed. II), Ed. Albastr, 2008; 11) L.A. Zadeh, D. Tufi, F. G Filip., I. Dzitac I.(eds.), From Natural

Language to Soft Computing: New Paradigms in Artificial Intelligence , Editing House of Romanian Academy, 2008.

Rzvan Andonie (UT Braov)

Titu I. Bjenescu (AT Militar)

Boldur E. Brbat (ULB Sibiu)

Dan Dumitrescu (UBB Cluj)

Henri Luchian (UAIC Iai)

Nicolae ndreanu (U Craiova)

Figura 5. Civa dintre cercettori, profesori i autori de cri din domeniul IA din

Romnia


14

Doctorate n IA n Romnia:

Aproape n toate centrele universitare mari, unde exist coli doctorale n domeniul Informatic sau tiina calculatoarelor (i nu numai), se elaboreaz i se susin teze de doctorat pe teme de IA. De exemplu, exist conductorii de doctorat n IA n urmtoarele centre universitare:

1) Bucureti (A. C. Atanasiu, F.G. Filip, V. Mitrana, D. Tufi) 2) Braov (R. Andonie) 3) Cluj (D. Dumitrescu) 4) Craiova (N. ndreanu) 5) Iai (D. Cristea, D. Glea, H. Luchian, H. N. Teodorescu), 6) Timioara (B. Brbat, V. Negru, S. Preitl).

Cercettori romni importani n IA

Grigore C. Moisil

(1906-1973)

Solomon Marcus

n. 1935

Paul Dan Cristea

n. 1941

F.G. Filip n. 1947

Gheorghe Pun

n. 1950

H.-N. Teodorescu

n. 1951

Gheorghe Tecuci n. 1954

Dan Tufi n. 1954

Figura 6. Civa dintre cercettorii romni importani n IA

1. Grigore C. Moisil (1906-1973): Logici cu mai multe valori. Pionier al informaticii recunoscut de IEEE Computer Society, http://www.ici.ro/romania/ro/stiinta/moisil.html


15

2. Solomon Marcus (n. 1935): Lingvistic matematic, Gramatici i automate finite. Membru al Academiei Romne

http://www.imar.ro/~smarcus/ 3. Paul Dan Cristea, Reele neurale, Informatic medical. Membru

Corespondent al Academiei Romne, http://www.dsp.pub.ro/info/staff/pcristea.htm

4. Florin Gheorghe Filip: Sisteme suport pentru decizii cu cunotine combinate (modele numerice i elemente de inteligen artificial). Membru al Academiei Romne, http://www.ici.ro/ici/homepage/filipf.html

5. Gheorghe Pun (DNA Computing, Membrane computing, P Systems), Cercettor la Instititul de Matematic al Academiei Romne i asociat la Universitatea din Sevilla. Membru Corespondent al Academiei Romne, Membru al Academiei Europea .http://www.imar.ro/~gpaun/

6. Horia-Nicolai Teodorescu (Fuzzy szstems, Artificial live), Membru Corespondent al Academiei Romne, profesor la Universitatea Tehnic Gheorghe Asachi din Iai http://www.fict.ro/HNT.htm

7. Gheorghe Tecuci (Instructable agents, multistrategy learning, mixed -initiative reasoning, modeling and knowledge acquisition, ontologies, knowledge engineering), Membru al Academiei Romne, profesor la George Mason University, SUA,

http://lalab.gmu.edu/members/tecuci.htm 8. Dan Tufi (Corpus Linguistics, Intelligent Computer Aided Language

Learning, Machine Language Learning, Machine Translation, Natural Language Understanding, Natural Language Generation, Knowledge Representation.). Membru Corespondent al Academiei Romne, director la Istititutului de Cercetri n Inteligen Artificial al Academiei Romne. http://www.racai.ro/~tufis/

Cele mai populare cri de iniiere n IA pe plan mondial:

1) Stuart Russel, Peter Norvig, Artificial Intelligence: A Modern Approach (2nd Edition), Prentice Hall Series in Artificial Intelligence, Hardcover, p. 1112, 2003. 2) Michael Negnevitsky, Artificial Intelligence, A Guide to Intelligent Systems, Second Edition, Pearson Education Limited, 2002. 3) Toshinori Munakata, Fundamentals of the New Artificial Intelligence, Second Edition, Springer-Verlag London Limited, 2008 4) Ben Goertzel, Cassio Pennachin, Artificial General Intelligence, Springer-Verlag Berlin Heidelberg, 2007.


16

Fondatori i cercettori pe plan internaional n IA:

George Boole (1815-1864)

Alan Turing (192-1954)

John McCarthy

n. 1927

Marvin Minsky

n.1927

Lotfi A. Zadeh

n. 1921

Allen Newell (1937-1992)

Herbert Simon (1916-2001)

Seymour Papert

n. 1928

Ray Kurzweil

n.1948

Kevin Warwick

n.1954.

Figura 7. Civa dintre fondatorii de subdomenii i cercettori n IA pe plan internaional Ali cercettori n IA (redm textul n limba englez, n original, cf. sursei2) Cognitive Scientists

Ballard, Dana H. - "Animate vision". Computational theories of the brain with emphasis on human vision. (Univ. of Rochester, USA)

Barsalou, Lawrence W. - Perceptual bases of cognition, situated conceptualization, dynamic representations of concepts, frames, category learning, event memory.

Brooks, Rodney A. - Embodied cognition in autonomous robots. (MIT, USA).

Calvin, William H. - Theoretical neurophysiologist and popularizer. Author of "The Cerebral Code," "How Brains Think," "Conversations with Neil's Brain" and "A Brain for All Seasons," amongst other works.

Cave, Kyle R. - Cognitive psychology of visual cognition, including attention, imagery, and object recognition.

Coneural Center for Cognitive and Neural Studies - Research center for computational embodied neuroscience. Director: Razvan Florian.

2 http://www.dmoz.org/Computers/Artificial_Intelligence/ (02.12.2008): De menionat

c n lista de mai jos mai sunt menionate numele a trei cercettori romni: Rzvan Florian, Rzvan Andonie i Raul Murean.


17

Cotterill, Rodney M.J. - Neurophysiology of consciousness, considered as an active process. Implications of this view for understanding autism. (Danish Technical University, Denmark)

Cziko, Gary - Educational psychology. Darwinian approaches. (Univ. of Illinois, USA)

Dunbar, Kevin - Scientific discovery and reasoning (especially in molecular biology); visual, causal, abductive, and analogical reasoning; gender differences in reasoning; science education; mental models; invivo cognition; visualization. (Dartmouth College.)

Gregory, Richard - Influential researcher on perception and perceptual illusion. Site gives access to selected publications as well as his CV and animated demonstrations of significant illusions.

Jrvilehto, Timo - Finnish Psychologist with interests in psychophysiology, preception, education, and "the theory of the organism-environment system".

Kirsh, David - Representation in everyday activity, cognitive complexity.

Koch, Christoph - Neuroscience and consciousness. Kosslyn, Stephen M. - Mental imagery, perception. Kubovy, Michael - Perceptual grouping. Loftus, Elizabeth F. - Psychology of memory, false memory, eyewitness

testimony. Minsky, Marvin - One of the pioneers and most creative thinkers of

Artificial Intelligence research. Modestino, Ed - A graduate student in the Complex Systems and Brain

Sciences program at Florida Atlantic University, working on neuroimaging of complex systems and neural networks of cognition.

Moravec, Hans - Mobile robots and their psychology. Norman, Don - Author of The Design of Everyday Things, list of books

and articles in human-centered design. O'Regan, J. Kevin - Visual perception: "change blindness" (nice

animated demos), active perception, eye movements, consciousness, and "the world as external memory".

Pylyshyn, Zenon W. - Visual attention and preattention, critique of "pictorial" theories of mental imagery, foundational issues in the computational theory of the architecture of cognition.

Scaruffi, Piero - Information about Scaruffi's research and teaching activities in Cognitive Science, Psychology of Consciousness and Philosophy of Mind, and links to his papers, and to his annotated bibliography of cognitive science, artificial intelligence, neurobiology,


18

artificial life, linguistics, neural networks, connectionism, cognitive psychology, and consciousness.

Smolensky, Paul - A leading researcher in connectionism. Thomas, Nigel - The philosophical, scientific, and historical study of

imagination and mental imagery, and their role in consciousness and cognition [Cal State LA].

Turner, Mark - Professor and Dean of Arts and Sciences at Case Western Reserve University. Research focus is to study how human thought processes are different than other species with an emphasis on blending.

van Diepen, Paul - Scene perception, eye movements, chronometry of information processing (no longer active in cognitive research).

Zhang, Jiajie - External representations; human-computer interaction; human factors; medical informatics

Fuzzy Intelligence People Abonyi, Janos - Fuzzy modeling for system identification and data

mining De Baets, Bernard - Fuzzy Relations and Preference Modelling Fodor, Jnos - Basics of fuzzy sets and connectives Fortemps, Philippe - Preference modelling Klement, Peter - Triangular Norms Mencar, Corrado - Neuro-fuzzy classifiers Roubens, Marc - Multicriteria decision aid Yager, Ronald R. - Decision support Zadeh, Lotfi A. - inventor of Fuzzy Sets Zimmermann, Hans-Jurgen - Fuzzy O.R.

Neural Networks People

Adelson, Edward T. - Visual perception, machine vision, image processing.

Agakov, Felix - Probabilistic graphical modeling, statistical learning theory, pattern recognition, prediction, and causality.

Allan, Moray - Computer vision, probabilistic models for image sequences, invariant features.

Amari, Shun-ichi - Neural network learning, information geometry. Andonie, Razvan - Data structures for computational intelligence. Andrieu, Christophe - Particle filtering and Monte Carlo Markov Chain

methods. Anthony, Martin - Computational learning theory, discrete mathematics. Attias, Hagai - Graphical models, variational Bayes, independent factor

analysis.


19

Bach, Francis - Machine learning, kernel methods, kernel independent component analysis and graphical models

Ballard, Dana H. - Visual perception with neural networks. Bartlett, Marian Stewart - Image analysis with unsupervised learning,

face recognition, facial expression analysis. Beal, Matthew J. - Bayesian inference, variational methods, graphical

models, nonparametric Bayes. Becker, Sue - Neural network models of learning and memory,

computational neuroscience, unsupervised learning in perceptual systems.

Bengio, Samy - Torch machine learning library, including SVMTorch support vector machine program. Research on mixture models, hidden markov models, multimodal fusion, speaker verification.

Beveridge, Ross - Computer vision, model-based object recognition, face recognition.

Bishop, Chris - Graphical models, variational methods, pattern recognition.

Boutilier, Craig - Decision making and planning under uncertainty, reinforcement learning, game theory and economic models.

Brody, Carlos D. - Somatosensory working memory, computation with action potentials, design of complex stimuli for sensory neurophysiology.

Brown, Andrew - Machine learning of dynamic data, graphical models and Bayesian networks, neural networks.

Bulsari, A. - Neural networks and nonlinear modelling for process engineering.

Calvin, William H. - Theoretical neurophysiologist and author of The Cerebral Code, How Brains Think.

Caruana, Rich - Multitask learning. Cheung, Vincent - Machine learning and probabilistic graphical models

for computer vision and computational molecular biology. Chu, Selina - Artificial intelligence, machine learning, data mining. Coolen, Ton - Physics of disordered systems. Working on dynamic

replica theory for recurrent neural networks. Cottrell, Garrison W. - An artrificial intelligence researcher who is an

expert on neural networks. Dahlem, Markus A. - Neural network models of visual cortex to model

neurological symptoms of migraine. Dayan , Peter - Representation and learning in neural processing

systems, unsupervised learning, reinforcement learning.


20

de Freitas, Nando - Bayesian inference, Markov chain Monte Carlo simulation, machine learning.

de Garis, Hugo - Evolvable neural network models, neural networks for programmable hardware, large neural networks.

De vito, Saverio - Neural networks for sensor fusion, wireless sensor networks, software modeling, multimedia assets management architectures

De Wilde, Philippe - Brain inspired models of uncertainty, linguistic and fuzzy uncertainty, uncertainty in dynamic multi-user environments.

Dietterich, Thomas G. - Reinforcement learning, machine learning, supervised learning.

Dr Hooman Shadnia - Dedicated to artificial neural networks and their applications in medical research and computational chemistry. Offers a quick tutorial on theory on ANNs written in Persian.

Freeman, William T. - Bayesian perception, computer vision, image processing.

Frey, Brendan J. - Iterative decoding, unsupervised learning Friedman, Nir - Learning of probabilistic models, applications to

computational biology. Frohlich, Jochen - Overview of neural networks, and explanation of

Java classes that implement backpropagation, and Kohonen feature maps.

Garcia, Christophe - Computer vision, image analysis, neural networks. Ghahramani, Zoubin - Sensorimotor control, unsupervised learning,

probabilistic machine learning. Hansen, Lars Kai - Neural network ensembles, adaptive systems and

applications in neuroinformatics. Herbrich, Ralph - Statistical learning theory, support vector machines

and kernel methods. Heskes, Tom - Learning and generalization in neural networks. Hinton, Geoffrey E. - Unsupervised learning with rich sensory input.

Most noted for being a co-inventor of back-propagation. Honavar, Vasant - Constructive learning, computational learning theory,

spatial learning, cognitive modelling, incremental learning. Hughes, Nicholas - Automated Analysis of ECG. Jaakkola, Tommi S. - Graphical models, variational methods, kernel m. Jensen, Finn Verner - Graphical models, belief propagation. Jordan, Michael I. - Graphical models, variational methods, machine

learning, reasoning under uncertainty. Joshi, Prashant - Computational motor control, biologically realistic

circuits, humanoid robots, spiking neurons.


21

Kearns, Michael - Reinforcement learning, probabilistic reasoning, machine learning, spoken dialogue systems.

Koller, Daphne - Probabilistic models for complex uncertain domains. Lafferty, John D. - Statistical machine learning, text and natural

language processing, information retrieval, information theory. Lawrence, Neil - Probabilistic models, variational methods. Lawrence, Steve - Information dissemination and retrieval, machine

learning and neural networks. LeCun, Yann - Handwritten recognition, convolutional networks, image

compression. Noted for LeNet. Leen, Todd - Online learning, machine learning, learning dynamics. Leow, Wee Kheng - Computer vision, computational olfaction. Lerner, Uri N. - Hybrid and Bayesian networks. Li, Zhaoping - Non-linear neural dynamics, visual segmentation,

sensory processing. Maass, Wolfgang - Theory of computation, computation in spiking

neurons. MacKay, David - Bayesian theory and inference, error-correcting codes,

machine learning. Malchiodi, Dario - Machine learning, Learning from uncertain data. McCallum, Andrew - Machine learning, text and information retrieval

and extraction, reinforcement learning. Meila, Marina - Graphical models, learning in high dimensions, tree

networks. Minka, Thomas P. - Machine learning, computer vision, Bayesian

methods. Muresan, Raul C. - Neural Networks, Spiking Neural Nets, Retinotopic

Visual Architectures. Murphy, Kevin P. - Graphical models, machine learning, reinforcement

learning. Murray, Alan - Neural networks and VLSI hardware. Murray-Smith, Roderick - Gesture recognition, Gaussian Process priors,

control systems, probabilistic intelligent interfaces. Neal, Radford - Bayesian inference, Markov chain Monte Carlo

methods, evaluation of learning methods, data compression. Oja, Erkki - Unsupervised learning, PCA, ICA, SOM, statistical pattern

recognition, image and signal analysis. Olier, Ivan - Artificial intelligence, generative topographic map, missing

data.


22

Olshausen, Bruno - Visual coding, statistics of images, independent components analysis.

Opper, Manfred - Statistical physics, information theory and applied probability and applications to machine learning and complex systems.

Paccanaro, Alberto - Learning distributed representation of concepts from relational data.

Pearlmutter, Barak - Neural networks, machine learning, acoustic source separation and localisation, independent component analysis, brain imaging.

Prashant, Joshi - Computational neuroscientist, with main areas of research interest being computational motor control, computational models of olfaction, computation with spiking neurons, neurocomputational basis of working memory and decision making, learning in biologically realistic circuits.

Rao, Rajesh P. N. - Models of human and computer vision. Rasmussen, Carl Edward - Gaussian processes, non-linear Bayesian

inference, evaluation and comparison of network models. Revow, Michael - Hand-written character recognition. Roberts, Stephen - Machine learning and medical data analysis,

independent component analysis and information theory. Rovetta, Stefano - Research on Machine Learning/Neural

Networks/Clustering. Applications to DNA microarray data analysis/industrial automation/information retrieval. Teaching activities.

Roweis, Sam T. - Speech processing, auditory scene analysis, machine l. Russell, Stuart - Many aspects of probabilistic modelling, identity

uncertainty, expressive probability models. Rutkowski, Leszek - Neural networks, fuzzy systems, computational int. Saad, David - Neural computing, error-correcting codes and

cryptography using statistical and statistical mechanics techniques. Sahani, Maneesh - Statistical analysis of neural data, experimental

design in neuroscience. Sallans, Brian - Decision making under uncertainty, reinforcement

learning, unsupervised learning. Saul, Lawrence K. - Machine learning, pattern recognition, neural

networks, voice processing, auditory computation. Saund, Eric - Intermediate level structure in vision. Schein, Andrew I. - Machine learning approaches to data mining

focussing on text mining applications. Sejnowski, Terry - Sensory representation in visual cortex, memory

representation and adaptive organization of visuo-motor transformations.


23

Seung, Sebastian - Short-term memory, learning and memory in the brain, computational learning theory.

Shkolnik, Alexander - Neurally controlled robotics. Shuurmans, Dale - Computational learning, complex probability

modelling. Simard, Patrice - Machine learning and generalization. Smola, Alex J. - Kernel methods for prediction and data analysis. Storkey, Amos - Belief networks, dynamic trees, image models, image

processing, probabilistic methods in astronomy, scientific data mining, Gaussian processes and Hopfield neural networks.

Sutton, Richard S. - Reinforcement learning. Sykacek, Peter - Brain Computer Interface. Teh, Yee Whye - Learning and inference in complex probabilistic

models. Tipping, Mike - Varied machine learning and data analysis topics,

including Bayesian inference, relevance vector machine, probabilistic principal component analysis and visualisation methods.

Tishby, Naftali - Machine learning; applications to human-computer interaction, vision,neurophysiology, biology and cognitive science.

Versace, Massimiliano - Neural networks applied to visual perception and computational modeling of mental disorders.

Wainwright, Martin - Statistical signal and image processing, natural image modelling, graphical models.

Wallis, Guy - Object recognition, cognitive neuroscience, interaction between vision and motor movements.

Weiss, Yair - Vision, Bayesian methods, neural computation. Welling, Max - Unsupervised learning, probabilistic density estimation,

machine vision. Wiegerinck, Wim - Inference in graphical models, mean field and

variational approaches. Williams, Christopher K. I. - Gaussian processes, image interpretation,

graphical models, pattern recognition. Winther, Ole - Variational algorithms for Gaussian processes, neural

networks and support vector machines. Also work on belief propagation and protein structure prediction.

Wiskott, Laurenz - Face recognition, Invariances in learning and vision. Wu, Yingnian - Stochastic generative models for complex visual

phenomena. Xing, Eric - Statistical learning, machine learning approaches to

computational biology, pattern recognition and control.


24

Yedidia, Jonathan S. - Statistical methods for inference and learning. Zemel, Richard - Unsupervised learning, machine learning,

computational models of neural processing. Zhou, Zhi-Hua - Neural computing, data mining, evolutionary

computing, ensemble networks. Computers: Robotics: History: People

Asimov, Isaac - Asimov was one of the three grand masters of science fiction and the creator of the term "robotics". He is known in the field for his Four Laws of Robotics and the large body of fiction he wrote about robots. He envisioned a future in which robots were (usually) safe, well-controlled servants of man.

Brooks, Rodney A. - Director of the MIT CS and AI Lab. CTO of iRobot. Involved in a wide range of robotics research initiatives. Best known as the creator of Subsumption Architecture in which layers of simple behaviors create complex emergent behaviors.

Capek, Karel - Czech author and playwright who popularized the term robot in his 1920 play, RUR: Rossum's Universal Robots.

Moravec, Hans P. - Researcher at the CMU Robotics Institute who believes intelligent machines will be the descendants of the human race. Author of several books on the nature of evolving robot intelligence. Current work involves 3D mapping and stereoscopic vision.

Tesla, Nikola - Tesla demonstrated a robotic boat that he called a teleautomaton in 1898 that many consider to be the first robot.

Tesla's Race of Robots - A summary of Nikola Tesla's work in robotics done as part of a PBS documentary on his life. Includes information on his teleautomaton and his predication of mechanical, automated men.

Tilden, Mark - Creator of BEAM robotics, a controversial philosophy of robot building that advocates the use of "nervous nets" instead of conventional microprocessors as controllers, the use of recycled parts, and solar power.

Walter, W. Grey, Online Archive - Biography of one of the pioneers of robotics. Includes historical information and photos of many of the robots Walter created in the 1940s and 1950s.

Walter, W. Grey, Robotics Pioneer - A respected neurophysiologist who did early work on autonomous mobile robots in the 1940s.

Wiener, Norbert - Generally known as the "father of cyborgs". Wiener was the originator of the term cybernetics and part of a group of scientists who originated the field. He was also an MIT mathematician, science fiction author, and designer of an early chess-playing robot.


25


26

5. CUM? Metode, modele, mijloace, tehnologii

INFORMATIC

TEHNOLOGIA INFORMAIEI

TIINA CALCULA-TOARELOR

AUTOMA-TIC

CIBERNETIC

ELECTRONIC

NEUROTINELE Neuropsihologia

Neurologia Neuropsihiatria Neurobiologia

INTELIGEN ARTIFICIAL

(IA)

MATEMATIC Logic

Probabiliti Grafuri

Limbaje formale Combinatoric...

BIOLOGIA Teoria evoluiei

Teoria emergenei ...

PSIHOLOGIA

FILOZOFIA GNOSEOLOGIA ...

Figura 8. Principalele tiine implicate n IA

Nu exist o definiie unanim acceptat pentru IA nici ca domeniu tiinific. Foarte muli consider IA ca un subdomeniu al tiinei calculatoarelor, dar ea studiaz i este strns legat i ar putea fi revendicat ca subdomeniu i de Informatic, Tehnologia informaiei, Automatic i Cibernetic. Cert este c IA este un domeniul tiinific interdisciplinar care implic pe lng tiinele amintite (cele scrise cu alb n figura 8) i alte tiine (cele scrise cu negru n figura 8), cum ar fi: Psihologia, Biologia, Neurotiinele, Electronica, Matematica, Filozofia etc.

TIPURI DE PROBLEME

MODELARE REZOLVARE ASOCIEREA NTRE DATELE DE INTRARE (ipoteze, valori iniiale) I RSPUNSUL CORECT (concluzie, rezultat)

1. Probleme bine-puse (date certe, valori exacte)

Exist un model formal al problemei

determinist: HARD COMPUTING

Relaie funcional explicit

2. Probleme ru-puse (date incerte, valori aproximative, adevr parial)

Nu exist un model formal (rspunsuri doar n cazuri particulare)

aproximativ: SOFT COMPUTING

nvare (pe baz de exemple de asociere ntrebare-rspuns)

Figura 9. Modelare i metode n IA


27

Calcul

inteligent (pur simbolic)

Logica fuzzy

Reele neurale Raionament probabilist

Learning Machine

Calcul evolutiv Algoritmi genetici

Calcul convenional (pur numeric)

HARD COMPUTING

SOFT COMPUTING tolerant la imprecizie, nesiguran,

adevr parial i aproximare

HARD COMPUTING

Figura 10. Tipuri de calcul inteligent

Bazndu-ne doar pe logica cu dou valori de adevr (numit i

tradiional, clasic, binar, boolean, aristotelic), putem rezolva doar PROBLEMELE BINE PUSE, cu date certe, precise, exacte. Dar n practic ne ntlnim la tot pasul cu probleme care conin noiuni inexacte, n care intervin date incerte, mrimi imprecise, variabile lingvistice sau adevruri pariale, numite i PROBLEME RU-PUSE, probleme care nu pot fi rezolvate prin calculul tradiional. Rezolvarea unor astfel de probleme se face azi prin soft computing, care are la baz logica fuzzy i cteva instrumente ale IA.

Logica fuzzy (Fuzzy Logic, Zadeh3, 1973) este o logic nuanat, non-aristotelic, cu mai mult de dou valori de adevr. Logica fuzzy mpreun cu Raionamentul probabilist (Probabilistic Reasoning), prin utilizarea Calcului evolutiv (Algoritmilor Genetici4) pe Reelele neurale5 i Learning Machine, formeaz un nou stil de calcul, numit SOFT COMPUTING6, cu ajutorul cruia putem rezolva probleme ru-puse.

Soft Computing-ul, la care se adaug, n mod nedisjunctiv, Sistemele expert (din economie, comer, medicin, nvmnt etc.), Sistemele bazate pe cunotine, Roboii, Reele semantice, Reelele de ncredere .a., sunt cteva din subdomeniile importante ale Inteligenei Artificiale (v. i fig. 11).

3 http://www.cs.berkeley.edu/~zadeh/ 4 n 1975 John Holland, inspirat de teoria evoluiei a lui Darwin, a pus bazele ALGORITMILOR GENETICI moderni. Dar exist contribuii premergtoare n domeniu ale unor biologi sau informaticieni: Barriceli (1954), Bremermann (1960), Fraser i Burnell (1970), Crosby )1973) .a. De multe ori algoritmii genetici se bazeaz pe modele de tip SWARM INTELLIGENCE, de inspirate din inteligena natural emergent a unor grupuri de insecte, cum ar fi cea a COLONIILOR DE FURNICI. 5 Prin anii 40, Warren McCulloch i Walter Pitts, inspirai de teoria nvrii condiionate a lui Pavlov, au pus bazele REELELOR NEURONALE, capabile s nvee pe baza anumitor reguli simple. Ulterior au adus contibuii Hebb (spresftul anilor40, Rosenblat (PERCEPTRON, 1950), Minsky i Papert (PERCEPTRONI, 1969) .a. 6 ...the principal constituents of Soft Computing (SC) are Fuzzy Logic (FL), Neural Computing (NC), Evolutionary Computation (EC) Machine Learning (ML) and Probabilistic Reasoning (PR), http://www.soft-computing.de/def.html


28

Clase Subclase Modele de calcul imitate

Realizri

Calcul neural Reelele bioelectrice din creierul uman formate de neuroni i sinapsele acestora.

Reele neurale

Calcul evolutiv Principiile darwiniste de evoluie

Algoritmi genetici Cutare armonic

Calcul molecular (DNA Computing)

ADN

Calcul membranar (Membrane Computing)

Membrana celulei

Calcul cuantic (Quantum Computing)

Sistem fizic

Calcul natural

Calcul de tip swarm (Swarm computing)

Colonii de furnici Bancuri de peti Colonii de bacterii Grupuri de animale Stoluri de psri

Algoritmi de optimizare

Calcul fuzzy

Control fuzzy Reele neuro-fuzzy

Raionamentul uman comun

Sisteme fuzzy Sisteme neuro-fuzzy Sisteme expert Sisteme hibride

Figura 11. Domeniile i subdomeniile calcului inteligent

Ageni inteligeni Citm din Boldur Brbat, Sisteme inteligente orientate spre agent [Br02]:

- Ceva software care tie s fac lucruri pe care ai fi putut probabil s le

faci singur, dac ai fi avut timp (Selker, citat de Hermans, 1996). - Un obiect care gndete (Gentia Software, 1997). - O entitate cu scopuri, aciuni i cunotine ntr-un anumit domeniu, situat

ntr-un mediu (Stone i Veloso, 1997). - O component de software i/sau echipament capabil s acioneze

exigent n scopul ndeplinirii unor sarcini pentru utilizator (Nwana, 1996). (Nwana adaug c prefer s-l foloseasc drept metatermen pentru o gam de tipuri de ageni.)

n fine, iat rezumatul unei definiii de sintez, dat de Tecuci (1998), care cuprinde multe din aspectele de mai sus. Un agent inteligent este un sistem bazat pe cunotine care: i percepe mediul; raioneaz pentru a interpreta percepiile, infer, rezolv probleme i stabilete aciuni; acioneaz asupra mediului pentru a ndeplini o seam de scopuri sau sarcini pentru care a fost proiectat. [...]


29

Ce pretindem de la cineva pe care ni-l alegem ca agent? n esen, trei trsturi (Pallmann, 1999):

a) Competen (inteligen nsoit de cunotine n domeniu). b) Personalizare (chiar dac nu aparine persoanei al crei agent este,

trebuie s presteze serviciile cerute de acest client n stil croit pe msur, nu standard).

c) Reprezentare (s acioneze nu numai pentru, ci i n numele persoanei care l-a ales, evident, n limitele mputernicirii date).

n fine, adoptnd o perspectiv i mai vdit antropocentric, trstura de frunte a agentului ar putea fi credibilitatea. Aceasta are, ca i n cazul oamenilor, dou subdimensiuni:

a) calitatea de a fi demn de ncredere (trustwortiness); b) competena (expertise).

Limbaje de programare utilizate n IA

LISP. Denumirea LISP vine de la LISt Processing7. Printele su este John McCarthy de la MIT AI Lab, iar anul naterii este considerat anul 1958. LISP este un limbaj declarativ (nu este imperativ). Conceput iniial ca un formalism matematic menit s conduc la dezvoltarea unei teorii riguroase a programelor, astzi LISP-ul este un limbaj puternic, n jurul cruia s-a dezvoltat un veritabil mediu de programare. Se poate spune ca LISP-ul este limbajul care ar permite s vedem calculatorul nu doar ca pe o complicat main de efectuat calcule aritmetice, ci ca pe un adevrat creier electronic aflat ntr-o continu i exploziv evoluie spre inteligen.

PROLOG. Limbajul Prolog ( PROgrammation en LOGique) a fost creat

la Marsilia la nceputul anilor 70, inventatorii fiind Alain Colmeraurer i Philippe Roussel. Una din principalele idei ale programrii logice este aceea c un algoritm este constituit din doua elemente disjuncte: logic i control. Componenta logic corespunde definiiei problemei ce trebuie soluionat, n timp ce componenta control stabilete cum poate fi obinut soluia. Un programator trebuie sa descrie numai componenta logic a unui algoritm, lsnd controlul executrii s fie exercitat de sistemul de programare logic utilizat. Cu alte cuvinte, sarcina programatorului este specificarea problemei ce trebuie soluionat. Astfel, limbajul logic poate fi conceput simultan ca limbaj de descriere, specificare formal a problemei i ca un limbaj de programare a calculatoarelor.

7 http://en.wikipedia.org/wiki/Lisp_programming_language


30

Fundamentele teoretice se gsesc n metoda demonstrrii automate dezvoltat ncepnd cu 1965 dup metoda rezoluiei data de Julian Robinson. Un rol important l-a avut si Robert Kowalski8 prin demonstrarea faptului ca se poate programa folosind logica. n limbajul Prolog9 se poate face orice ca i n alte limbaje de programare. ns Prologul are avantaje distincte, precum i dezavantaje. Prelucrarea rapida a datelor numerice este un punct vulnerabil al lui. Prologul poate mnui numere, ns nu aa de eficient ca un limbaj special destinat acestui lucru. Prologul, nsa, puncteaz cnd e vorba de manipularea simbolurilor. Ori manipularea simbolurilor este inima a ceea ce a devenit azi cunoscut sub numele de IA.

CLIPS este un acronim pentru C Language Integrated Production System, un sistem expert dezvoltat de NASA n anii 1980. Sintaxa i numele au fost inspirate de OPS (sistem de producie oficial, n englez Official Production System) creat de Charles Forgy. Primele versiuni de CLIPS au fost dezvoltate ncepnd cu 1984 la NASA -Johnson Space Center (ca o alternativ la sistemul existent numit ART*Inference) pn la nceputul anilor 1990 cnd subvenia a ncetat din cauza problemelor bugetului Federal i a unui ordin conform cruia NASA trebuia s cumpere software comercial n loc s-l dezvolte.

CLIPS este probabil cel mai folosit sistem expert, deoarece este rapid, eficient i gratuit. Cu toate c acum face parte din domeniul public, este nc actualizat i susinut de autorul original, Gary Riley10.

CLIPS ncorporeaz un limbaj de programare orientat obiect numit COOL pentru a scrie sisteme expert. Cu toate c este scris n limbajul de programare C, interfaa sa seamn foarte mult cu cea a limbajului de programare LISP. Se pot scrie extensii n C, iar CLIPS poate fi chemat din C. Ca i alte sisteme expert, CLIPS are de-a face cu reguli i fapte. n timpul rulrii programelor, existena diferitelor fapte ntr-o baz de cunotine pot face ca o regul s fie aplicabil.

8 http://www.doc.ic.ac.uk/~rak/ 9 http://thor.info.uaic.ro/~georgie/prolog/introducere.html 10 http://clipsrules.sourceforge.net/


31

CAPITOLUL 1

INTELIGENA ARTIFICIAL CA IMITAIE APROXIMATIV A INTELIGENEI NATURALE

Rezumat. Paradoxul aproximrii inteligenei naturale prin inteligena artificial: Ceea ce omul face mai uor (de exemplu: recunoatere vizual, vorbire, mers, nvare) e mai greu de imitat prin inteligen artificial; iar ceea ce omul face mai greu (de exemplu: calcule matematice, calcule simbolice, memorare) e mai uor de imitat prin inteligen artificial. Cel mai bune rezultate se obin imitnd inteligena emergent a grupurilor de animale sociale.

1.1. Inteligena natural uman 1.1.1. Generaliti

ntr-o descriere aproximativ, inteligena uman ar fi capacitatea individului de a nva uor i bine, uurina de a soluiona probleme noi, de a se adapta la situaii noi, pe baza experienei acumulate.

William Stern a descris inteligena uman ([Ste12], [Ste38]), ca fiind aptitudinea general a individului de a-i adapta contient gndirea unor cerine noi: ea este capacitatea spiritual de adaptare general la noile cerine i condiii ale vieii.

n latin, intelligere nseamn a relaiona, a organiza, iar interlegere nsemna stabilirea de relaii ntre oameni. Probabil c aceste dou cuvinte au o legtur strns cu etimologia cuvntului inteligen. Cuvntul inteligent provine tot din limba latin de la intelligo, care n traducere nseamn detept, nelept, priceput, ager la minte. Pe scurt, inteligena ar putea fi definit ca fiind capacitatea minii de a stabili legturi ntre diferite date.

n [DEX98] gsim descrierea: INTELIGN, inteligene, s.f. Capacitatea de a nelege uor i bine, de a sesiza ceea ce este esenial, de a rezolva situaii sau probleme noi pe baza experienei acumulate anterior; deteptciune [...].

Inteligena uman este obiectul de studiu al psihologiei n colaborare cu neurotiinele (neurofiziologia, neuropsihologia, neuroanatomia, etc), psihofizica, psihologia comportamental, cibernetica biologic, biomecanica,


32

inteligena artificial, teoria controlului, management, informatica i matematicile aplicate, respectiv robotica, mecatronica i controlul automat.

Pentru msurarea inteligenei este vehiculat termenul de coeficient de inteligen sau IQ (intelligence quotient), care este un scor derivat din unul sau mai multe teste psihometrice standardizate de msurare a inteligenei.

Definirea noiunii de inteligen uman (delimitarea i clarificarea aspectelor sale cantitative i calitative) sunt probleme extrem de dificile i ele au fost abordate de muli oameni de tiin, n special, psihologi, dintre care se remarc:

Alfred Binet (18571911) - psiholog i psihometrician francez, autor al celebrului test de inteligen utilizat pentru msurarea inteligenei cognitive (IQ);

Thodore Simon (1872 - 1961) psiholog i psihometrician francez, care a contribuit la stabilirea faimoasei scale de inteligen, cunoscut sub numele de scala de inteligen Binet- Simon;

William Lewis Stern (1871-1938) - psiholog german, inventatorul coeficientului de inteligen pe baza cruia se face testul IQ;

Lewis Madison Terman (1877-1946) psiholog american, pionier al psihologiei cognitive, profesor la Universitatea Stanford, care a pus la punct, pe baza lucrrilor lui Binet, testul de inteligen cunoscut sub numele de testul IQ Stanford-Binet;

Edouard Claparde (1873-1940) - neurolog i psiholog elveian, care s-a ocupat de studiul psihologiei copilului i a tipurilor de memorie;

Jean Piaget (1896-1980) - filozof i om de tiin elveian care s-a ocupat de studiul etapelor dezvoltrii cognitive. Piaget a spus despre inteligen c este ceea ce foloseti cnd nu tii ce s faci i c inteligena nseamn a nelege i a inventa ;

Howard Gardner (n. 1943) - fondatorul teoriei inteligenelor multiple; Daniel Goleman (n. 1946) - psiholog american care s-a ocupat de

studiul inteligenei emoionale (EQ); i alii: cele mai recente studii legate de inteligena emoional se

datoreaz lui Steven J. Stein, Howard E. Book i Karl Albrecht.

Secolul XX a fost dominat de conceptul IQ, ca indicator extrem de important n anticiparea performanelor de care ar fi putut fi capabil un individ, despre care se spune c nu prea mai poate fi mbuntit pe parcursul vieii. IQ ia n considerare doar un singur tip de inteligen, inteligena cognitiv (mental). Ulterior ns, a fost scos n eviden un alt aspect al inteligenei umane, inteligena emoional, care are la baz teoria inteligenelor multiple a lui Gardner.


33

1.1.2. Inteligene multiple

Teoria inteligenelor multiple a fost formulat pentru prima oar de ctre psihologul american Howard Gardner [Gar83]. Teoria sa pornete de la ideea existenei unor inteligene diferite i autonome ce conduc la modaliti diverse de cunoatere, nelegere i nvare. El consider c inteligena nu este o nsuire pus n lumin prin fore standard, ci capacitatea de a rezolva probleme i de a dezvolta sau realiza produse n situaii concrete de via.

Astfel, capacitatea cognitiv a omului este descris printr-un set de abiliti, talente, deprinderi mentale pe care Gardner le numete ,,inteligene (iniial 7, apoi 8 i apoi a fost pomenit i a 9-a: [Gar06]). Toi indivizii normali posed fiecare din aceste inteligene ntr-o msur mai mare sau mai mic. Ceea ce-i deosebete este gradul lor de dezvoltare i natura unic a combinrii acestor inteligene la fiecare individ n parte. n acest sens Gardner subliniaz ideea c o inteligen trebuie s fie probat de existena unei zone de reprezentare pe creier i prin existena unui sistem propriu de expresie.

Howard Gardner a identificat pn n prezent (n [Gar83] i [Gar06]) urmtoarele tipuri de inteligen: 1. Inteligena lingvistic: capacitatea de a rezolva probleme i de a construi

produse cu ajutorul codului lingvistic (presupune abilitatea i plcerea de a citi, scrie, povesti sau a se juca cu cuvintele (exemple: rezolvarea cuvintelor ncruciate, practicarea jocului SCRABLE etc.);

2. Inteligena logico-matematic: presupune capacitatea de a descoperi legturi, modele, categorii i relaii (exemple: se manifest n jocuri logice, rezolvarea problemelor de aritmetic sau n jocurile de strategie);

3. Inteligena spaial: se refer la capabilitatea de a gndi n imagini i la facilitatea n rezolvarea unor probleme de tip geometrico-spaial (exemple: gsirea drumului ntr-un labirint, aptitudinea de a desena sau de a construi figuri din cuburi Lego sau jocuri pe calculator de tip spaial);

4. Inteligena corporal-kinestezic: implic o mare sensibilitate n identificarea i prelucrarea senzaiilor fizice, de exemplu, a simi ritmul unui dans; inteligena la nivelul corpului i al minilor ne permite s controlm i s interpretm micrile corpului, s manevrm obiecte, s realizm coordonarea (armonia) dintre trup i spirit. Acest tip de inteligen nu se regsete numai la atlei, acrobai sau dansatori, ci poate fi ntlnit n micrile, n practicarea unor meserii riscante care presupun o mare finee pentru reuit (o operaie pe creier, dezamorsarea unei bombe, pilotarea unei maini de curse etc.). Acest tip de inteligen include deprinderi fizice speciale precum coordonarea, echilibrul, dexteritatea, fora, flexibilitatea, viteza, precum i deprinderi la nivelul proprioceptorilor, la nivel tactil i cutanat;


34

5. Inteligena muzical: presupune existena urechii muzicale, adic a posibilitii de a percepe i distinge sunete care par la fel altor persoane;

6. Inteligena interpersonal: presupune empatie (uurin de a te pune n situaia altcuiva, adic existena capacitii de a nelege sentimentele altora), este prezent mai ales la cei cu spirit de conductor;

7. Inteligena intrapersonal: reflect o bun cunoatere a propriilor sentimente i posibiliti, capacitate de introspecie i autoanaliz non subiectiv.

8. Inteligena naturalist (capacitatea de a rezolva probleme i de a dezvolta produse cu ajutorul clasificrilor i reprezentrilor din mediul nconjurtor).

9. Inteligena existenial Gardner este convins c este o modalitate de cunoatere a lumii care i caracterizeaz pe filozofi, pe cei care pun ntrebri despre sensul fericirii, nceputul universului etc. Probabil c i spiritualitatea aparine acestui tip de inteligen. Gardner ns nu a stabilit localizarea pe creier. De aceea vorbete despre aceasta ca despre o jumtate de inteligen. Din punct de vedere biologic, inteligenele sunt independente, n funcie

de zonele corticale care le guverneaz. La nivel individual ele apar n combinaii, fiecare individ fiind de fapt , o colecie de inteligene.

Un studiu recent elaborat de ctre Karl Albrecht [Alb06] menioneaz necesitatea rearanjrii modelului inteligenelor multiple al lui Howard Gardner, n alte ase categorii primare:

1. Inteligena abstract: raionamente simbolice; 2. Inteligena social: legturile interumane; 3. Inteligena practic: organizarea activitilor; 4. Inteligena emoional: contiina de sine i auto-controlul; 5. Inteligena estetic: sensul formelor, desenul, muzica,arta i literatura; 6. Inteligena kinestezic: capacitile fizice, cum ar fi cele sportive,

dansul, muzica etc. 1.1.3. Inteligena cognitiv (IQ). Teste psihometrice pentru IQ

Se zice c IQ este un coeficient care nu poate fi ameliorat semnificativ pe parcursul vieii. Testele de inteligen au fost folosite pentru a anticipa succesul educaional. Astfel c persoanele cu un IQ sczut sunt uneori orientate ctre un program de educaie pentru persoane cu nevoi speciale, n timp ce un IQ ridicat recomand aceste persoane pentru un program educaional avansat.

Preocuparea pentru realizarea unor teste ct mai exacte dateaz nc din anul 1905 i se datoreaz psihologului francez Alfred Binet, care a publicat primul test modern de evaluare a inteligenei, cunoscut azi sub numele de scala de inteligen Binet-Simon. Principalul scop al acestui test era de a identifica


35

studenii care aveau nevoie de asisten special n parcurgerea planului de nvmnt. n colaborare cu Theodore Simon, Binet a adus optimizri, publicnd noi versiuni ale acestui test de evaluare a inteligentei n 1908 i 1911.

Abrevierea termenului de coeficient de inteligen sau IQ (traducerea din german a Intelligenz-Quotient) a fost atribuit n 1912 unui psiholog german, William Stern, care definea IQ ca fiind coeficientul de msurare a nivelului de inteligen al unei persoane.

n 1916, Lewis M. Terman, profesor la Universitatea Stanford, aplic teoria lui Stern pentru o versiune rafinat a Scalei Binet-Simon i elaboreaz un test numit scala de inteligen Stanford-Binet. Testul lui Terman a pus bazele unuia dintre cele mai moderne teste de evaluare a inteligentei folosit pn n prezent.

IQ se calcula folosind formula IQ= 100 X vrsta mintal/vrsta biologic,

rezultnd c pentru o persoan n vrst de 10 ani care era evaluat la nivelul unei persoane cu vrsta de 13 ani s aib un IQ de 130 (100 X 13/10).

Primul test de inteligen special creat pentru aduli a fost publicat de abia n 1939 de David Wechsler i era numit scala de inteligen pentru aduli a lui Weschler (WAIS), ceea ce a determinat i apariia WISC (scala de inteligen pentru minori a lui Weschler), standardiznd astfel coeficientul de inteligen, care nu se mai baza pe vrst.

Astfel, pe baza rezultatelor aplicrii acestui test, persoanele testate sunt arondate unei anumite categorii de inteligen (v. fig. 1.1.1).

IQ CLASIFICARE >140 Geniu 120-140 Deosebit de inteligent 110-119 Foarte inteligent 90-109 Inteligen medie (Normal) 80- 89 Inteligen mediocr 70- 79 La limita deficienei mintale 50-69 napoiat mintal (Cretin) 20-49 Imbecil < 20 Idiot

Figura 1.1.1. Clasificarea inteligenei umane dup IQ

Faptul ca IQ msoar doar inteligena nnscut i nu poate fi ameliorat

semnificativ pe parcursul vieii, dar i ncercrile de a explica situaiile n care persoane cu un IQ mediu sau sczut au cunoscut un mare succes n societate (de exemplu, Shakespeare, Darwin, Spilberg, Picasso, Ghandi, Einstein, Mozart, Freud etc.), au deschis perspectivele teoriei privind importana i influena


36

inteligenei emoionale n dezvoltarea personalitii umane. Deci, inteligena este de dou tipuri: cea cognitiv (analitic, logic) i cea emoional. Prima este strategic i acioneaz pe termen lung, iar cea de-a doua poate oferi rspunsuri la probleme curente n viaa de zi cu zi. 1.1.4. Inteligena emoional (EQ). Inteligena social

Cartea lui Daniel Goleman, Inteligena emoional, cheia succesului n via (v. [Gol04] i www.DanielGoleman.info), a marcat o revoluie uluitoare n psihologie prin analiza importanei covritoare a emoiilor n dezvoltarea personalitii umane.

Studiul su ne explic cum, atunci cnd ne nelegem sentimentele, situaia n care ne aflm devine mai limpede. Descoperim chiar un nou mod de a privi cauzele bolilor care ne macin familia i societatea. Prelund rezultatele cercetrilor asupra creierului i comportamentului, autorul propune extinderea conceptului de inteligen. Autorul a deschis calea unei psihologii care acord un interes egal i inteligenei sentimentelor.

Inteligena emoional (EQ) presupune, n primul rnd, contientizare de sine, autodisciplin i empatie. Ea d seama de felul n care ne controlm impulsurile i sentimentele.

Vestea bun este faptul c inteligena emoional poate fi mbuntit. Dei copilria este extrem de important n punerea unor baze solide pentru dezvoltarea inteligenei emoionale, ea poate fi mbuntit i cultivat inclusiv la vrsta adult.

La zece ani de la apariia primei ediii n limba englez, studiul inteligenei emoionale a cptat proporiile unui domeniu tiinific autonom n slujba cruia lucreaz un numr impresionant de cercettori folosind cele mai avansate metode tehnologice. Astzi, inteligena emoional se pred n coli i universiti, competenele sale au devenit criterii de angajare sau de promovare n carier, iar programele de educaie pe baza sa au devenit punctul de plecare n politicile sociale de prevenire a mbolnvirilor psihice sau criminalitii.

Mai mult, se vorbete azi deja i de o inteligen social. Omul este un animal social, dup cum a spus nc Aristotel. Omul triete i acioneaz ntr-un mediu social. De aceea, eficiena sa att n rezolvarea problemelor din viaa de zi cu zi, ct si a celor ntlnite n activitatea profesional, nu depinde numai de aptitudinea intelectual, ci i de capacitatea de a construi i dezvolta relaii interpersonale pozitive i armonioase, care s permit ndeplinirea elurilor propuse.

O alt carte a lui Daniel Goleman, Inteligena social. Noua tiin a relaiilor umane [Gol07], ne atrage atenia asupra unui nou tip de inteligen, diferit de inteligena emoional, nu ca parte a acesteia, ci ca dimensiune de sine stttoare.


37

Reaciile pe care le avem fa de ceilali, ca i ale lor fa de noi, au un impact biologic mult mai extins dect ne nchipuim. Este vorba de declanarea unor cascade de hormoni care ne regleaz ntregul organism, de la inim pn la sistemul imunitar, determinnd relaiile bune s acioneze ca nite vitamine, iar pe cele proaste ca nite otrvuri. Ne putem molipsi de emoiile celorlali oameni tot aa cum lum o grip, iar, pe de alt parte, consecinele izolrii sau cele ale unui stres social intens ne pot scurta viaa.

n Inteligena social, Goleman explic surprinztoarea corectitudine a primelor impresii, fundamentul carismei i fora emoional, complexitatea atraciei sexuale i sesizarea minciunilor. El descrie partea ntunecat a inteligenei sociale, de la narcisism la machiavelism i psihopatie. Ne mai vorbete despre uimitoarea noastr capacitate de a fi vizionari, ca i despre tragedia celor care, asemenea copiilor autiti, au un acces redus la raiune. Iar mesajul distinct al acestei cri este urmtorul: noi, oamenii, avem o predilecie nnscut ctre empatie, cooperare i altruism, astfel nct putem dezvolta o inteligen social prin care s ne hrnim aceste caliti nepreuite.

Suntem n era n care tot mai multe voci susin necesitatea dezvoltrii acelei laturi a naturii umane care este inteligena emoional, cea care ne asigur succesul profesional i n plan personal.

Modelul practic al competenei EQ elaborat de Daniel Goleman identific urmtoarele cinci dimensiuni:

1. Contiina de sine; 2. Auto-cenzurarea; 3. Motivaia; 4. Empatia; 5. Relaiile.

Steven J. Stein i Howard E. Book [StB08] se refer la organizarea EQ sub

forma unor domenii: 1. Domeniul intrapersonal: contiina emoional de sine, caracterul

asertiv, independena, respectul de sine, mplinirea de sine; 2. Domeniul interpersonal: empatia, responsabilitatea social, relaiile

interpersonale; 3. Domeniul adaptabilitii: testarea realitii, flexibilitatea, soluionarea

problemelor; 4. Domeniul administrrii stresului: tolerana la stres, controlul

impulsurilor; 5. Domeniul strii generale: optimismul, fericirea.


38

Modelul inteligenei sociale elaborat de ctre K. Albrecht [Alb06] este alctuit din cinci dimensiuni distincte:

1. Contiina situaional: radar social sau abilitatea de a studia situaiile i de a interpreta comportamentul oamenilor;

2. Prezena: inuta sau ntreaga colecie de semnale pe care ceilali le proceseaz ntr-o impresie evaluatoare a unei persoane.

3. Autenticitatea: radarele sociale ale altora asupra comportamentului nostru.

4. Claritatea: capacitatea noastr de ai face pe alii sa coopereze cu noi. 5. Empatia: dar empatia mprtit ntre dou persoane, ca stare a

legturilor cu alt persoan care s creeze bazele pentru o interaciune pozitiv i cooperant.


39

1.2. Inteligena natural emergent a grupurilor sociale de animale (swarm intelligence)

n natur ntlnim comportamente inteligente emergente la colonii de furnici sau termite, roiuri de albine, stoluri de psri, turme de animale, haite de lupi, etc.

n general, vom prefera s numim acest tip de inteligen Swarm Intelligence, denumire utilizat n limba englez pentru formele de inteligen artificial bazate pe comportamentul colectiv al sistemelor distribuite, auto-organizate din natur, ca cele amintite mai sus. Aceast expresie a fost introdus de ctre Gerardo Beni i Jing Wang n 1989 n contextul sistemelor celulare de roboi.

O alt definiie pentru Swarm Intelligence, care a fost dat de ctre Bonabeau, Dorigo, Theraulaz, este: orice ncercare de a proiecta algoritmi sau echipamente distribuite inspirate din comportamentul colectiv al coloniilor de insecte sociale sau alte societi de animale, n [BDT01], [Ca+01], [HRR05]11.

Unul dintre cele mai simple i mai relevante exemple de comportament emergent inteligent l constituie, indiscu

inteligenta_artificiala_dzitac

Documents