《机器学习简介》PPT课件.ppt

《《机器学习简介》PPT课件.ppt》由会员分享，可在线阅读，更多相关《《机器学习简介》PPT课件.ppt（44页珍藏版）》请在三一办公上搜索。

1、Machine Learning：An Overview,石立臣,Outline,What is machine learning(ML)Types of machine learningWork flowPopular modelsApplicationsFutures,What is machine learning,Training set(labels known),Test set(labels unknown),f()=“apple”f()=“tomato”f()=“cow”,What is machine learning,DefinitionMachine learning r

2、efers to a system capable of the autonomous acquisition and integration of knowledgeMachine learning is programming computers to optimize a performance criterion using example data or past experience,Computer,Data,Algorithm,Program,Knowledge,Knowledge(new),What is machine learning,Every machine lear

3、ning algorithm has three componentsRepresentationModel(rules,statistics,instance;logic,KNN,SVM,DNN,)EvaluationPerformance(accuracy,mse,energy,entropy,)OptimizationParameters Combinatorial optimizationConvex optimizationConstrained optimization,Types of machine learning,Supervised learningTraining da

4、ta includes desired outputsUnsupervised learningTraining data does not include desired outputsSemi-supervised learningTraining data includes a few desired outputsReinforcement learningRewards from sequence of actions,Types of machine learning,Supervised learningClassification:discrete outputRegressi

5、on:continuous output,Bias-variance,Training and Validation Data,Full Data Set,Training Data,Validation Data,Idea:train eachmodel on the“training data”and then testeach modelsaccuracy onthe validation data,Underfitting&Overfitting,PredictiveError,Model Complexity,Error on Training Data,Error on Test

6、Data,Ideal Rangefor Model Complexity,Overfitting,Underfitting,Types of machine learning,Unsupervised learningClusteringDimensionality reductionFactor analysis,Types of machine learning,Semi-supervised learningClustering or classification,Types of machine learning,Reinforcement learningRobot&control,

7、Work flow,Prediction,Training Labels,Training,Training,Image Features,Image Features,Testing,Test Image,Learned model,Learned model,Slide credit:D.Hoiem and L.Lazebnik,Work flow,Features,Work flow,ModelsLogic,RulesStatistical,Black box modelStatic,dynamic modelOnline learningEnsemble learning,Work f

8、low,Architecture,Model,Feature,Hardware,Popular models,Linear model:logistic regression,linear discriminant analysis,linear regression(with basis function),Popular models,Nearest neighborFeature&distance,Popular models,Support vector machine,Popular models,Artificial neural network,Popular models,De

9、cision tree,Popular models,Collaborative filtering,Popular models,Hierarchical clusteringK-meansSpectral clusteringManifold learning,Popular models,Hidden markov modelConditional random fields,Applications,Applications,Applications,Applications,Applications,Applications,Applications,Applications,App

10、lications,Attention,Applications,Image classification,Applications,Applications,Brain machine interface,Applications,Applications,Applications,Applications,Applications,Indirect illuminationRegression,Applications,Indirect illuminationkd-tree,Applications,The core is the data set!Othersfeaturesmodel&optimization,Futures,DecisionControlKnowledgePrediction,