人工智能原理人工智能原理 (53).pdf

《人工智能原理人工智能原理 (53).pdf》由会员分享，可在线阅读，更多相关《人工智能原理人工智能原理 (53).pdf（16页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、Applications and TerminologiesArtificial Intelligence519.Perspectives about Machine Learning 9.1.What is Machine Learning 9.2.History of Machine Learning 9.3.Why Different Perspectives 9.4.Three Perspectives on Machine Learning 9.5.Applications and TerminologiesContents:Artificial Intelligence:Learn

2、ing:Perspectives52Machine PerceptionComputer VisionVideo AnalysisPattern RecognitionFace/Speech/Fingerprint RecognitionOptical Character Recognition(OCR)Handwriting RecognitionGame PlayingNatural Language ProcessingInformation RetrievalApplication Fields of Machine Learning 机器学习的应用领域9.5.Applications

3、 and Terminologies机器感知计算机视觉视频分析模式识别人脸/语音/指纹识别光学字符识别(OCR)手写体识别玩游戏自然语言处理信息检索Artificial Intelligence:Learning:Perspectives53Text or Document Classification(e.g.Spam Email Detection)Recommender SystemsAd PlacementCredit ScoringFraud DetectionStock TradingDrug DesignMedical DiagnosisRoboticsApplication F

4、ields of Machine Learning 机器学习的应用领域9.5.Applications and Terminologies文本与文档分类(例如垃圾邮件检测)推荐系统广告配置信用评分欺诈检测股票交易新药设计医学诊断机器人学Artificial Intelligence:Learning:Perspectives54 Samples 样本 Items or instances of data used for learning or evaluation.用于学习或评估的数据项或实例。Features 特征 The set of attributes,often represent

5、ed as a vector associated to a sample:属性集，通常表示为与样本相关的向量：Handcrafted features:手工式特征e.g.,SIFT,HOG,SURF,LBP,GLOH,LESH,CENTRIST.例如，SIFT、HOG、SURF、LBP、GLOH、LESH、CENTRIST。Learned features:学习式特征e.g.,by convolutional neural network.例如，通过卷积神经网络。Some Terminologies in Machine Learning 机器学习中的一些术语9.5.Applications

6、 and TerminologiesArtificial Intelligence:Learning:Perspectives55 Handcrafted Features 手工式特征 HOG(Histogram of Oriented Gradients)HOG(定向梯度直方图)Similar to SIFT(Scale-Invariant Feature Transform),but improved accuracy.与SIFT(尺度不变特征变换)类似，但改善了精度。By distribution of intensity gradients or edge directions.按照强

7、度梯度或边缘方向分布。64128 detection window.64128个检测窗口。9.5.Applications and TerminologiesR-HOGC-HOGSome Terminologies in Machine Learning 机器学习中的一些术语Artificial Intelligence:Learning:Perspectives56 Learned Features 学习式特征 Humans can learn to see efficiently.Because brains are deep,with many layers of processing.

8、人类可以有效地学会观察。因为大脑是深度的，具有许多处理层次。Some algorithms for such deep architectures,can produce features from raw data for visual recognition.具有这种深度架构的算法，能从原始数据中生成视觉认知的特征。Feature learning also be called representation learning.特征学习也被称为表示学习。Understanding deep learning will enable us to build more intelligent m

9、achines for visual recognition.理解深度学习将使我们能够构建更智能的视觉认知机器。Some Terminologies in Machine Learning 机器学习中的一些术语9.5.Applications and TerminologiesObject modelPixelEdgesObject partsArtificial Intelligence:Learning:Perspectives57 Labels 标记 Values or categories assigned to samples.在样本上指定的值或类别。In classificatio

10、n problems,samples are assigned specific categories.分类问题中，样本被指定特定的类别。In regression problems,items are assigned real-valued labels.回归问题中，项被指定为实值的标记。Training sample 训练样本 Samples used for training learning algorithm.用于训练学习算法的样本。In spam problem,the training sample consist of a set of email samples along

11、 with their associated labels.对于垃圾邮件问题，训练样本由一组邮件样本以及相关标签组成。Some Terminologies in Machine Learning 机器学习中的一些术语9.5.Applications and TerminologiesArtificial Intelligence:Learning:Perspectives58 Validation sample 验证样本 Samples used to tune the parameters of a learning algorithm when working with labeled d

12、ata.用于在使用标记数据时调整学习算法参数的样本。Learning algorithms typically have one or more free parameters,and validation sample is used to select appropriate values for these model parameters.学习算法通常具有一个或多个自由参数，因而验证样本用于为这些模型参数选择适当的值。Test sample 测试样本 Samples used to evaluate the performance of a learning algorithm.用于评

13、估学习算法性能的样本。These predictions are then compared with the labels of the test sample to measure the performance of the algorithm.然后将这些预测与测试样本的标签进行比较，以衡量算法的性能。Some Terminologies in Machine Learning 机器学习中的一些术语9.5.Applications and TerminologiesArtificial Intelligence:Learning:Perspectives59 Hypothesis set

14、 假设集 A set of functions mapping features to the set of labels Y.将特征映射为标签Y的函数集。For example,the following are a set of functions mapping email features to:例如，映射电子邮件特征的函数集如下：Loss function 损失函数 To measure the difference,or loss,between a predicted label and a true label.用于度量预测标签和真实标签之间差异或损失。Denoting the

15、 set of all labels as Yand the set of possible predictions as Y,a loss function L is a mapping:将所有的标签集表示为Y、并且可能的预测集为Y，则损失函数L为映射：Some Terminologies in Machine Learning 机器学习中的一些术语9.5.Applications and TerminologiesY=spam,nonspam.L:Y Y +Artificial Intelligence:Learning:Perspectives60 Abstraction 抽象 It i

16、nvolves the translation of data into broader representations.其含义是将数据转化为更广泛的表示。Generalization 泛化 It describes the process of turning abstracted knowledge into a form that can be utilized for action.It is also the ability of a learning algorithm to perform accurately on unseen samples after having exp

17、erienced a learning data set.它形容将抽象知识转化为可用于动作形式的过程。它也是学习算法具有学习数据集的经验后，可以对未知样本正确地进行处理的能力。Some Terminologies in Machine Learning 机器学习中的一些术语9.5.Applications and TerminologiesDataAbstractionGeneralizationArtificial Intelligence:Learning:Perspectives61Some Notations in This Course 本课程的一些符号9.5.Application

18、s and TerminologiesNotation 符号Description 说明Set of real numbers实数集+Set of non-negative real numbers非负实数集nSet of n-dimensional real-valued vectorsn维实值向量集a,bClosed interval between a and ba和b之间的闭区间(a,b)Open interval between a and ba和b之间的开区间Set of natural numbers,i.e.,0,1,.自然数集，即：0,1,.YAn arbitrary set

19、任意集合YInput space输入空间YTarget space目标空间Hhypothesis set假设集Artificial Intelligence62Summary 本章小结9.Perspectives about Machine Learning Machine learning is to study some algorithms that can learn from and make predictions on data.机器学习是研究一些可以从数据中学习、并对数据进行预测的算法。The different perspectives are aimed to try to

20、 have a taxonomy on the algorithms of machine learning,for being easy to understand machine learning.几个不同视角旨在尝试对机器学习的算法进行分类，以便于理解机器学习。Three perspectives on machine learning are proposed in this chapter,those are learning tasks,learning paradigms and learning models.本章提出了机器学习的三个视角，他们是：学习任务、学习范例以及学习模型

21、。Artificial Intelligence:Learning:Perspectives63The Three Perspectives 三个视角9.Perspectives about Machine LearningMachine LearningDimensionality reductionDensity estimationRankingClusteringRegressionClassificationOptimizationLearning ParadigmsLearning TasksLearning ModelsProbabilisticNetworkedLogicalGeometricReinforcementUnsupervisedSupervisedArtificial Intelligence:Learning:Perspectives64The Three Perspectives 三个视角9.Perspectives about Machine LearningMaybe“Blind Men and an Elephant”