View Single Post
Old 04-15-2012, 08:03 PM
jcatanz jcatanz is offline
Join Date: Apr 2012
Posts: 41
Default Re: HW2.10 out-sample error on original(2D) or transformed (5D) X space?

Originally Posted by rukacity View Post
I may be asking a very silly question but I am stuck here - what are the out of sample points? And given a sample of 1000 points how do we get out of sample points? Is it picking random samples out of these 1000 points? If yes, then how many do we pick?
Here, the 'sample' refers to the training set, the original set of random points (x1,x2) which are used to solve for the the weights.

The 'out-of-sample' points are a new set of random points (x1,x2). This set is used to test how well your weights worked on random points outside the training set, i.e points that were not used to derive them.
Reply With Quote