Module: Linear Classifiers

This module looks at basic linear classification approaches. First is the natural extension of regression, the Fisher Discriminant. Second is the maximum entropy approach, Logistic Regression. Finally, the maximum margin method, Support Vector Machines.

Dates: Mon, Jul 12 - Fri, Jul 16

Learning Outcomes

Linear Classifiers

Recognize different fundamental approaches to Linear Classification
Understand how to formulate the Fisher Discriminant (not necessary for Spr 23)
Understand the maximum entropy principle, Logistic Regression as an example
Understand the margin approach, Support Vector Machine as an example

Readings

Fisher Discriminant

Lin Reg with categorical labels interpreted as numbers

Max Entropy: Logistic Regression

Max Entropy approach to modeling classes

Max margin: Support Vector Machine

Max margin approach: Support Vector Machines

Assessments

Linear Classifiers

Assessment

Outcome(s) assessed:

16 Jul 09:00 AM