This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: problem can be written as min w,ξ 1 2 k w k 2 + C m X i =1 ξ i (2a) s.t. y i h w,x i i ≥ 1ξ i for all i (2b) ξ i ≥ , (2c) Derive the dual of (2) and contrast it with ( ?? ). What changes to the SMO algorithm would you make to solve this dual? Problem 3 (4 pt) LIBSVM is a widely used solver for training a SVM. The aim of this problem is to familiarize yourselves with LibSVM and use it for our familiar CML/PAKDD spam detection problem. • Download and read the documentation about LibSVM. • Use LibSVM with three diﬀerent kernels to classify the ECML/PAKDD discovery challenge dataset that you used in the previous HW. 1 • Compare the performance of diﬀerent kernels for your problem. • Write a report (5 pages max) to document your ﬁndings. In particular, mention how you choose the kernel parameters. 2...
View
Full
Document
This note was uploaded on 02/23/2012 for the course STAT 598 taught by Professor Staff during the Spring '08 term at Purdue.
 Spring '08
 Staff

Click to edit the document details