LFD Book Forum (http://book.caltech.edu/bookforum/index.php)
-   Chapter 2 - Training versus Testing (http://book.caltech.edu/bookforum/forumdisplay.php?f=109)
-   -   Practically finding break point or VC dimension (http://book.caltech.edu/bookforum/showthread.php?t=4114)

 udaykamath 03-16-2013 08:33 PM

Practically finding break point or VC dimension

Dear Prof Yaser
Greetings! I understand the proofs and the theoretical argument. The point that you underline is "give me the breakpoint" and i will give the error bound in terms of examples N etc. Now practically for an algorithm how do we find a breakpoint? If someone comes up with an algorithm f(x), how is breakpoint or VC dimension actually computed? Are there any formal steps etc ?
Thanks in anticipation!
Uday Kamath
PhD Candidate
GMU

 yaser 04-08-2013 12:46 PM

Re: Practically finding break point or VC dimension

Quote:
 Originally Posted by udaykamath (Post 9961) I understand the proofs and the theoretical argument. The point that you underline is "give me the breakpoint" and i will give the error bound in terms of examples N etc. Now practically for an algorithm how do we find a breakpoint? If someone comes up with an algorithm f(x), how is breakpoint or VC dimension actually computed? Are there any formal steps etc ?
There is no general, systematic way for finding the break point of a learning model, but for many of the popular models, the break point has already been estimated (approximately if not exactly). In one of the homeworks in the online course, this problem is addressed in a geometric case.

 All times are GMT -7. The time now is 12:06 AM.