Re: typo in lecture 24?
29 views
Skip to first unread message
Ryan Tibshirani
Dec 3, 2012, 1:40:58 PM12/3/12
to 1072...@googlegroups.com
No, that is correct. By weak duality, c^Tx - (-b^Ty) = c^Tx + b^T >=
0. This is: primal objective >= dual objective. Strong duality is =
0, or sense we always have weak duality, you can think of strong
duality as <= 0. The relaxed strong duality just makes this <=
lambda.
Ryan
On Mon, Dec 3, 2012 at 9:27 AM, swordsnow <robin...@gmail.com> wrote:
> is there a typo on slides 4, for strong dual and relaxed dual (last two
> equation for KKT), are they both be greater than the RHS?I am scribing.
> Thanks!
>
> --
>
>
0. This is: primal objective >= dual objective. Strong duality is =
0, or sense we always have weak duality, you can think of strong
duality as <= 0. The relaxed strong duality just makes this <=
lambda.
Ryan
On Mon, Dec 3, 2012 at 9:27 AM, swordsnow <robin...@gmail.com> wrote:
> is there a typo on slides 4, for strong dual and relaxed dual (last two
> equation for KKT), are they both be greater than the RHS?I am scribing.
> Thanks!
>
> --
>
>
Lu Xie
Dec 5, 2012, 6:49:13 PM12/5/12
to 1072...@googlegroups.com
There might be a typo on slide Page 9. I think \Delta s = A \Delta x, without +b term.
Lu
Lu
--
Geoff Gordon
Dec 6, 2012, 2:07:37 PM12/6/12
to 1072...@googlegroups.com
Yes, that's right, Delta s should be A Delta x. Thanks for catching this!
-Geoff.
-Geoff.
Reply all
Reply to author
Forward
0 new messages