《計算機科學導論》課件Unit 15Artificial Intelligence

1、?計算機科學導論?課件Unit 15Artificial Intelligence2345History of artificial intelligence6The field of artificial intelligencedeems the human cognition primitive is symbols, and cognitive processes are symbol operation. The core problems of artificial intelligence are knowledge representation, knowledge reaso

2、ning, and application of knowledge. The Symbolism always tries to use mathematical logic to create a unified theory of artificial intelligence system, but always encounters the difficulties that some knowledge cannot be resolved.7The field of artificial intelligence8The field of artificial intellige

3、nce9The field of artificial intelligence1011Semantic networks圖圖12Semantic networks13Rule-based systems14Other representations15Extracting knowledge16Extracting knowledge17Typical contribution of the school of Symbolism18Typical contribution of the school of Symbolism19Typical contribution of the sch

4、ool of Symbolism2021Image and sound processing22Image and sound processing23Image and sound processing24Image and sound processing25Image and sound processing26Image and sound processing27Image and sound processing28圖圖15.2 Googles self-driving carImage and sound processing29Image and sound processin

5、g30圖圖15.3 Googles self-driving carImage and sound processing31Image and sound processing32Image and sound processing33圖圖15.4 Tubenet Transit being developed in Beijing, ChinaImage and sound processing34Image and sound processing35圖圖15.5 CyCab a small autonomous carImage and sound processing36Image a

6、nd sound processing37Image and sound processing38Image and sound processing39Image and sound processing40Image and sound processingFigure 15.6 A quadrotor robot aircraft ()41Image and sound processing424344454647484950515253Intelligent robotFigure 15.7 A quadrotor robot aircraft(Courtesy of :/ ted )

7、5455565758Intelligent robotFigure 15.8 The BigDog robot(Courtesy of :/ ted )59606162636465666768697071ABA BA BA BA BAFFFFTTTFTFTTFTFFTFFFTTTTTT72ABA BA BA BA BAFFFFTTTFTFTTFTFFTFFFTTTTTT73ABA BA BA BA BAFFFFTTTFTFTTFTFFTFFFTTTTTT74ABA BA BA BA BAFFFFTTTFTFTTFTFFTFFFTTTTTT7576The weather is fine toda

8、yPremise 1Premise 2Therefore, well go to play football.7778798081All men are mortalsx man(x) mortal(x) Socrates is a manman(Socrates)Therefore, Socrates is mortal)()(xmortalxmanx8283848586Data mining and machine learning87Data mining and machine learning8815-5 Nature Inspired Computation89Introducti

9、on90Introduction91IntroductionInspired fromModels or AlgorithmsBrain information processingArtificial Neural NetworkFuzzy way of thinkingFuzzy SystemBiological immune mechanismArtificial ImmuneSystemBiological evolutionary processEvolutionary Computation (EC)Table15.2 Summary of some models or algor

10、ithms of Natural Computations92Artificial neural networks93Artificial neural networksFigure 15.9 The inputs and output of a perceptron94Artificial neural networks95Artificial neural networks96Artificial neural networksFigure 15.10 Schematic diagram of multi-layer neural network97Artificial neural ne

11、tworks98Artificial neural networksFigure 15.11 Schematic diagram of multi-layer neural network99Bionic intelligence100Bionic intelligence101Bionic intelligence102Bionic intelligence103Evolutionary computation104Evolutionary computation105Evolutionary computation106Evolutionary computation107Evolutio

12、nary computationAlgorithmMain Ideas or ObjectiveGenetic Algorithms (GA)Introduced byJohn Holland, in 1970sTo use three operators (selection, crossover, and mutation) to process some data structures which are used for simulating biological gene to get the result of the problem.The largest application

13、 of techniques is in the domain of optimization, which are the mainstream algorithms of EC.Evolution Strategies (ES)Introduced byI. Recenberg, H. P. Schwefel, in 1960sTo solve parameter optimization problems.Table15.3 Summary of some evolutionary computations108Evolutionary computationAlgorithmMain

14、Ideas or ObjectiveEvolutionary Programming (EP)Introduced byI. Recenberg, H. P. Schwefel, in 1960sTo evolve Finite State Machines (FSM) to predict events on the basis of former observations.Genetic Programming (GP)Introduced byCramer, in 1985.J, Koza made it more perfect, in 1992To evolve the program