【大学学习】统计机器学习 全8讲

 【大学学习】统计机器学习 全8讲

索引： Outline(00:00:08)  Challenging problems(00:00:19)  Data Mining(00:00:53)  Machine Learning(00:02:15)  Application in PR(00:03:14)  Difference(00:03:28)  Biometrics(00:04:04)  Bioinformatics(00:04:39)  ISI(00:05:08)  Confusion(00:05:34)  统计机器学习基础研究(00:06:00)  Machine learning community(00:06:31)  学习(00:06:55)  Performance(00:08:15)  学习(00:08:19)  Performance(00:08:22)  More(00:08:53)  Theoretical Analysis(00:09:11)  Ian Hacking(00:09:44)  Statistical learning(00:10:28)  Andreas Buja(00:10:46)  Interpretation of Algorithms(00:11:22)  统计学习(00:11:58)  Main references(00:13:18)  Main kinds of theory(00:13:39)  Definition of Classifications(00:14:02)  统计学习(00:14:23)  Main kinds of theory(00:15:21)  Definition of Classifications(00:15:22)  Definition of regression(00:15:50)  Several well-known algorithms(00:16:27)  Framework of algorithms(00:17:02)  Designation of algorithms(00:17:58)  统计决策理论(00:18:39)  Bayesian：classification(00:19:26)  统计决策理论(00:20:10)  Bayesian：classification(00:20:13)  Bayesian: regression(00:20:18)  统计决策理论(00:20:55)  Bayesian：classification(00:21:00)  Bayesian: regression(00:21:17)  Estimating densities(00:21:25)  KNN(00:22:45)  Interpretation：KNN(00:23:20)  高维空间(00:24:15)  维数灾难(00:25:01)  维数灾难(00:25:50)  维数灾难：其它体现(00:26:45)  LMS(00:27:33)  Interpretation: LMS(00:29:57)  维数灾难(00:30:57)  KNN(00:30:58)  Designation of algorithms(00:30:59)  Designation of algorithms(00:31:00)  统计决策理论(00:31:01)  Estimating densities(00:31:18)  高维空间(00:31:19)  维数灾难：其它体现(00:31:20)  Interpretation: LMS(00:31:21)  Fisher Discriminant Analysis(00:31:40)  Interpretation: FDA(00:32:35)  FDA and LMS(00:33:04)  FDA: a novel interpretation(00:33:38)  FDA: parameters(00:34:24)  FDA: framework of algorithms(00:35:09)  Disadvantage(00:35:59)  Bias and variance analysis(00:36:44)  Bias-Variance Decomposition(00:37:17)  Bias-Variance Tradeoff(00:38:46)  Bias-Variance Decomposition(00:38:52)  Bias-Variance Tradeoff(00:39:05)  Interpretation: KNN(00:40:29)  Ridge regression(00:41:35)  Interpretation: ridge regression(00:42:03)  Ridge regression(00:42:43)  Interpretation: ridge regression(00:43:05)  Interpretation: parameter(00:43:28)  Interpretation: ridge regression(00:43:35)  Interpretation: parameter(00:43:37)  A note(00:44:32)  Other loss functions(00:45:39)  Interpretation: boosting(00:46:35)  Boosting方法的由来(00:47:22)  Boosting方法流程（AdaBoost）(00:48:18)  Interpretation: margin(00:48:47)  Interpretation: SVM(00:49:43)  SVM: experimental analysis(00:50:48)  Interpretation: base learners(00:51:57)  Disadvantage(00:52:38)  Generalization bound(00:53:15)  PAC Frame(00:54:16)  VC Theory and PAC Bounds(00:54:44)  PAC Bounds for Classification(00:55:38)  VC Dimension(00:55:51)  PAC Bounds for Classification(00:55:52)  VC Dimension(00:56:27)  A consistency problems(00:57:39)  Remarks on PAC+VC Bounds(00:58:33)  SVM: Linearly separable(00:59:21)  SVM: soft Margin(01:00:28)  SVM: Linearly separable(01:01:12)  SVM: soft Margin(01:01:22)  SVM: algorithms(01:01:59)  泛化能力的界(01:03:01)  Bound: VC Dimension(01:04:04)  Bound: VC dimension+errors(01:04:45)  Disadvantages of SRM(01:05:52)  Disadvantage: PAC+VC bound(01:06:52)  Several concepts(01:07:51)  Disadvantage: PAC+VC bound(01:08:00)  Several concepts(01:08:02)  Generalization Bound: margin(01:08:35)  Importance of Margin(01:09:48)  Generalization Bound: margin(01:10:29)  Importance of Margin(01:10:34)  Vapnik’s three periods(01:10:35)  Neural networks(01:11:51)  Interpretation: neural networks(01:12:55)  BP Algorithms(01:14:17)  Disadvantage(01:15:42)  The End(01:16:32)