电子科技大学：《统计学习理论及应用 Statistical Learning Theory and Applications》课程教学资源（课件讲稿，英文版）Lecture 05 Support Vector Machine

Statistical Learning Theory and Applications Lecture 5 Support Vector Machine Instructor:Quan Wen SCSE@UESTC Fal,2021

Statistical Learning Theory and Applications Lecture 5 Support Vector Machine Instructor: Quan Wen SCSE@UESTC Fall, 2021

Outline (Level 1) 1A problem in perceptron 2 Linear SVM in LSC 3 SVM in LIC 4 Kernel Functions 5)Nonlinear SVM 6 SMO Algorithm 1/263

Outline (Level 1) 1 A problem in perceptron 2 Linear SVM in LSC 3 SVM in LIC 4 Kernel Functions 5 Nonlinear SVM 6 SMO Algorithm 1 / 263

Topics: Core ideas of support vector machine Basic ideas and concepts of convex optimization o Formula derivation of support vector machine Key points and difficulties: Key points:Core ideas of support vector machine;Basic ideas and concepts of convex optimization;Formula derivation of support vector machine o Difficulties:KKT (Karush-Kuhn-Tucker)condition 2/263

Topics: Core ideas of support vector machine Basic ideas and concepts of convex optimization Formula derivation of support vector machine Key points and difficulties: Key points: Core ideas of support vector machine; Basic ideas and concepts of convex optimization; Formula derivation of support vector machine Difficulties: KKT（Karush-Kuhn-Tucker）condition 2 / 263

Outline (Level 1) A problem in perceptron Linear SVM in LSC SVM in LIC Kernel Functions Nonlinear SVM SMO Algorithm 3/263

Outline (Level 1) 1 A problem in perceptron 2 Linear SVM in LSC 3 SVM in LIC 4 Kernel Functions 5 Nonlinear SVM 6 SMO Algorithm 3 / 263

1.A problem in perceptron Linear Classifiers X g yest ·denotes+1 denotes-1 How would you classify this data? 4/263

1. A problem in perceptron Linear Classifiers 4 / 263

Linear Classifiers X g yest ·denotes+1 。denotes-1 How would you classify this data? 5/263

Linear Classifiers 5 / 263

Linear Classifiers X g yest ·denotes+1 。denotes-1 How would you classify this data? 6/263

Linear Classifiers 6 / 263

Linear Classifiers X g yest ·denotes+1 denotes-1 How would you classify this data? 7/263

Linear Classifiers 7 / 263

Linear Classifiers X g yest ·denotes+1 denotes-1 Any of these would be fine.. .but which is best? 8/263

Linear Classifiers 8 / 263

Outline (Level 1) A problem in perceptron Linear SVM in LSC SVM in LIC Kernel Functions Nonlinear SVM SMO Algorithm 9/263

Outline (Level 1) 1 A problem in perceptron 2 Linear SVM in LSC 3 SVM in LIC 4 Kernel Functions 5 Nonlinear SVM 6 SMO Algorithm 9 / 263