人工智能原理-北京大学-9--PartVLearningChapter9Perspective-(9

SchoolofElectronicandComputerEngineeringPekingUniversityWangWenminWhyDifferentPerspectivesArtificialIntelligence269.PerspectivesaboutMachineLearning9.1.WhatisMachineLearning9.2.HistoryofMachineLearning9.3.WhyDifferentPerspectives9.4.ThreePerspectivesonMachineLearning9.5.ApplicationsandTerminologiesContents:ArtificialIntelligence::Learning::Perspectives27Whichalgorithmshouldchoose应该选择哪种算法Supposewehaveanapplicationthatmachinelearningmightbegoodfor,soweneedanappropriatealgorithmforlearningfromdata.假设我们有一个应用程序，机器学习会有帮助，因此需要一个适当的从数据中学习的算法。Theproblemwefacedishowtochooseoneofmachinelearningalgorithms.我们面临的问题是如何选择一个机器学习算法。Howmanylearningalgorithms有多少种算法Somanyalgorithmsformachinelearning.机器学习的算法如此之多。Literallythousandsavailable,andhundredsmorepublishedeachyear.大概有数千种，每年又会发表数百种。DifficultyinUnderstandingMachineLearning理解机器学习的难点9.3.WhyDifferentPerspectivesArtificialIntelligence::Learning::Perspectives28Whatisthedifficulty难点是什么Withoutacategoryofmachinelearning,howtodeterminewhichalgorithmcouldbeused?没有机器学习的分类法，如何确定哪种算法适用？Thecategorizationrelatesourperspectiveonmachineleaning.这种分类关系到我们观察机器学习的视点。Isoneperspectiveenough一个视点够吗Tooutlookonmostofmachinelearningalgorithms,oneperspectiveissohard.要了解大多数机器学习算法，仅有一个视点是不够的。Weshouldlookfrommultipleperspectivestohaveafullviewofmachinelearning.我们应该从多个视点来观察，使之对机器学习有一个完整的把握。DifficultyinUnderstandingMachineLearning理解机器学习的难点9.3.WhyDifferentPerspectivesArtificialIntelligence::Learning::Perspectives29Itusesexperienceorinteractswithenvironmenttoimproveperformance,ormakesaccuratepredictions.使用经验或与环境交互来改善性能，或做出精确预测。HowMachineLearningWorks机器学习如何工作9.3.WhyDifferentPerspectivesTrainingDataUnseenDataOutputFeedbackf(X)LearningAlgorithmh(X)HypothesisArtificialIntelligence::Learning::Perspectives30WhyThreePerspectives为什么有三个视点9.3.WhyDifferentPerspectivesProblems问题Scenarios场景Approaches手段LearningTasks学习任务LearningParadigms学习范式LearningModels学习模型WhattoDo做什么WhatSituations什么形式HowtoDo如何做ArtificialIntelligence::Learning::Perspectives31DefinitionoftheThreePerspectives三个视点的定义9.3.WhyDifferentPerspectivesTypes类型Description描述LearningTasks学习任务Denotingthegeneralproblemsthatcanbesolvedbymachinelearning.表示可以用机器学习解决的基本问题。LearningParadigms学习范式Denotingthetypicalscenariosthatarehappenedinmachinelearning.表示机器学习中发生的典型场景。LearningModels学习模型Denotingtheapproachesthatcanhandletofulfilalearningtask.表示可以处理完成一个学习任务的方法。