recitation-SVM

# recitation-SVM - Recitation 10-701/15-781, Fall 2008 SVM By...

This preview shows pages 1–9. Sign up to view the full content.

Recitation 10-701/15-781, Fall 2008 SVM By Hanghang Tong (htong@cs) Some Parts of the slides are from previous years’ recitation and lecture notes 1

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Roadmap Hinge Loss How to derive? What is the intuition SVM vs. LR vs. AdaBoost SMO How does it work Implications of SMO Limitations of SMO SVR 2
Hinge Loss (from HW2) We know (soft-margin linear SVM) Want to show 3 ?

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Hinge Loss (from HW2) We know (soft-margin linear SVM) Want to show 4 max(0,1 ( )) t i i i y w x b  But… ( )) t i i i y w x b  i=
Hinge Loss: What is the intuition? 5 ii yf Loss function ( ) ( ) i t y f w x b H x sign f  : true label

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Hinge Loss: What is the intuition? 6 ii yf Loss function ( ) ( ) i t y f w x b H x sign f  : true label Ideal Case Wrong Correct but less confident Correct & Confident
Hinge Loss: What is the intuition? 7 ii yf Loss function ( ) ( ) i t y f w x b H x sign f  : true label But, non-convex Wrong Correct but less confident Correct & Confident

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Hinge Loss: What is the intuition?
This is the end of the preview. Sign up to access the rest of the document.

## This note was uploaded on 01/26/2010 for the course MACHINE LE 10701 taught by Professor Ericp.xing during the Fall '08 term at Carnegie Mellon.

### Page1 / 23

recitation-SVM - Recitation 10-701/15-781, Fall 2008 SVM By...

This preview shows document pages 1 - 9. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online