【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)
