Yapay Zeka’ya giris Yapay sinir aglari ve bulanik mantik Uzay CETIN Université Pierre Marie Curie (Paris VI), Master 2 Recherche, Agents Intelligents, Apprentissage et Décision (AIAD) November 18, 2008 Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 1 / 44 soft computing Icindekiler 1 soft computing 2 fuzzy logic 3 Neurofuzzy Modelling YSA genel olarak yapay sinir aglari 4 Bulanik Mantik bulanik. akil yurutme bulanik. akil yurutme Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 2 / 44 soft computing Soft computing, parallels the remarkable ability of human mind to reason and learn in an environment of uncertainty and imprecision. Figure: A neural character recogniser and a knowladge base cooperation Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 3 / 44 soft computing Figure: A fuzzy inference system Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 4 / 44 soft computing Soft computing is a discipline situated at the combination of several relatively new and distinct mathematical techniques: fuzzy logic neural networks probabilistic reasoning (PR) which include genetic algorithms, chaos theory, belief nets and learning theory. Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 5 / 44 fuzzy logic Icindekiler 1 soft computing 2 fuzzy logic 3 Neurofuzzy Modelling YSA genel olarak yapay sinir aglari 4 Bulanik Mantik bulanik. akil yurutme bulanik. akil yurutme Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 6 / 44 fuzzy logic It is based on the idea that sets are not crisp but some are fuzzy, and these can be modeled in linguistic human terms such as large, small and medium. what is the advantage? This results in fewer rules and lower computer resources. Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 7 / 44 fuzzy logic It is based on the idea that sets are not crisp but some are fuzzy, and these can be modeled in linguistic human terms such as large, small and medium. what is the advantage? This results in fewer rules and lower computer resources. Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 7 / 44 fuzzy logic Modeling of the human operators behavior. In fuzzy systems, rules can be formulated that use these linguistic expressions and apply them to the human behavioral problem. Fuzzy inference systems are an extension of classical AI techniques that incorporate human knowledge and perform uncertain reasoning. Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 8 / 44 fuzzy logic In the supervised learning process input-output pairs of a process are used for training. In some cases genetic or evolutionary algorithms which are derivative-free optimization techniques Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 9 / 44 Neurofuzzy Modelling Icindekiler 1 soft computing 2 fuzzy logic 3 Neurofuzzy Modelling YSA genel olarak yapay sinir aglari 4 Bulanik Mantik bulanik. akil yurutme bulanik. akil yurutme Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 10 / 44 Neurofuzzy Modelling Why combination A neural network A neural network can approximate a function, with adjusted weights, but it is impossible to interpret the result in terms of natural language. A fuzzy system a fuzzy rule base consists of readable if-then statements that are almost natural language, but it cannot learn the rules itself. So combination Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 11 / 44 Neurofuzzy Modelling Extracting rules from data The goal is to reduce the complexity in a problem, or to reduce the amount of data associated with a problem. The inference mechanism If f (e1 is A1 , e2 is A2 , . . . , ek is Ak ) then y = g (e1 , e2 , . . . , ek ) (1) Here f is a logical function that connects the sentences forming the condition, y is the output, and g is a function of the inputs ei . Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 12 / 44 Neurofuzzy Modelling Feature determination In general, data analysis (Zimmermann, 1993) concerns objects which are described by features. The features form axes of an abstract feature space in which each object is represented by a point. For instance, in the four-dimensional coordinate system spanned by the axes top speed, colour, air resistance, and weight. So a vehicle V1 can be represented by the point (x1 , x2 , x3 , x4) = (220, red, 0.30, 1300). Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 13 / 44 Neurofuzzy Modelling The latter option is necessary in case a large number of features needs to be reduced to a smaller number of features. Objects are fuzzy when one or more features are described in fuzzy terms. An example is a vehicle with a very fast car engine, rather than top speed equal to some crisp number. Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 14 / 44 Neurofuzzy Modelling Figure: bulanik iliski Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 15 / 44 Neurofuzzy Modelling Figure: bulanik iliski Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 16 / 44 Neurofuzzy Modelling (two clusters) Each point belongs to one or the other class, so the clusters are crisp. Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 17 / 44 Neurofuzzy Modelling Bulanik kume ornekleri IF height is tall THEN weight is heavy. Figure: bulanik iliski Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 18 / 44 Neurofuzzy Modelling Bulanik kume ornekleri kucukler ve buyukler, iki tane bulanik kume olsun U = V = {1, 2, 3, 4} kucukler = 1.0/1 + 0.6/2 + 0.1/3 + 0.0/4 buyukler = 0.0/1 + 0.1/2 + 0.6/3 + 1.0/4 5’e yakin olan sayilar, S5 ile gosterilsin, S5 = 0.0/2 + 0.1/3 + 0.6/4 + 1.0/5 + 0.6/6 + 0.1/7 + 0.0/8 burdaki toplama isareti cebirsel toplama isareti degil sadece toplu gosterim amacini tasiyor. Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 19 / 44 Neurofuzzy Modelling bulanik. akil yurutme Icindekiler 1 soft computing 2 fuzzy logic 3 Neurofuzzy Modelling YSA genel olarak yapay sinir aglari 4 Bulanik Mantik bulanik. akil yurutme bulanik. akil yurutme Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 20 / 44 Neurofuzzy Modelling bulanik. akil yurutme Kural yapisi IF (x1 is A1 ) AND (y1 is B1 ) THEN (z1 is C1 ) µA∪B (x) = max[µA (x), µB (x)] µA∩B (x) = min[µA (x), µB (x)] Figure: bulaniklastirma islemi Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 21 / 44 Neurofuzzy Modelling bulanik. akil yurutme Avantaji nedir? Cok sayidaki gercel degerler, birkac tane bulanik degiskenle ifade edilebiliyor. Az sayidaki kuralla akil yurutme yapilabiliyor. kompakt!!! Figure: bulaniklastirma islemi Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 22 / 44 Neurofuzzy Modelling bulanik. akil yurutme Icindekiler 1 soft computing 2 fuzzy logic 3 Neurofuzzy Modelling YSA genel olarak yapay sinir aglari 4 Bulanik Mantik bulanik. akil yurutme bulanik. akil yurutme Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 23 / 44 Neurofuzzy Modelling bulanik. akil yurutme Kural yapisi Figure: bulanik cikarim Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 24 / 44 Neurofuzzy Modelling bulanik. akil yurutme Kural yapisi burda kurallarin ciktilarinin tumunu ayni potada eritiyoruz. Figure: bulanik birlestirme simdi geriye kalan islem durulastirma. bunun icin de cesitli yontemler var. onlardan biri agirlik noktasi bulma yontemi. bu ortaya cikan sekli iki esit agirliga X eksenine dik ir cizgi ile ayirir. Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 25 / 44 Neurofuzzy Modelling bulanik. akil yurutme Sugeno-style Figure: bulanik birlestirme Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 26 / 44 Neurofuzzy Modelling bulanik. akil yurutme Sugeno-style aggregation Figure: bulanik birlestirme Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 27 / 44 Neurofuzzy Modelling bulanik. akil yurutme Sugeno-style aggregation Figure: bulanik birlestirme Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 28 / 44 Ozet son Tesekkurler... bir sonraki derste, Bulanik mantik ve yapay sinir aglarini uygulamalarina giricez. takip edecegimiz kitap : C++ Neural Networks and Fuzzy Logic Valluru B. Rao Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 29 / 44 Appendix Referanslar Referanslar I Bernhard Nebel, Gnther Grz Yapay Zeka, Inkilap Kitabevi. wikipedia http://en.wikipedia.org/wiki/Monty Hall problem Uzay CETIN () Yapay Zeka’ya giris November 18, 2008 30 / 44