View Single Post
Old 08-29-2012, 01:26 PM
AdrienL AdrienL is offline
Join Date: Jul 2012
Posts: 13
Default Re: Calculating w from just support vectors--numerically risky?

I was wondering why my results were sometimes weird... Indeed it seems that you have to use all the vectors even those who are "negligible" to obtain the best results.
I checked that the b was consistant among all vectors but only did it visuall once... I did not think of adding an automated test.

(I still had the right answers for the homework though !)
Reply With Quote