Support Vector Machines

Week 7: Oct 6 – Oct 10

Weekly Objectives

We will introduce three main concept, the SVM under linear decision, slack variables and kernel SVM. We will leave the reproducing kernel concept and penalized formulation to the end of the semester.

• Understand how SVM formulates and solves the decision rule in the linearly separable case
• Understand how SVM use slack variables to handle linearly nonseparable case
• Understand how SVM incorporate the kernel trick to fit nonlinear decision rules
• Use the e1071 and kernlab packages to fit linear and nonlinear SVM
• Use the caret package to perform cross-validation of parameters

Lecture Notes and R Examples

• SVM [Source File]

Additional Readings

• Linearly Non-separable Details
• SVM as a penalized model

Homework

• [Homework 7] [Source File]
• [Local: Social Network Ads Data]
• [Kaggle: Social Network Ads Data]
• Due Oct 16, 11:59PM CT
• [Solution]