[Csdp] hinge loss with slack variables
Nicolas Bock
nicolasbock at gmail.com
Thu Aug 8 17:38:20 EDT 2013
Hi,
Suppose I would like to maximize
max_{X} [ Tr(C X) ]_{+}
where [z]_{+} = max(z, 0), the hinge loss function. I can replace the hinge
loss with a slack variable and get
max_{X} \xi
s.t.
\xi >= Tr(C X)
But turning the inequality constraint into an equality constraint would
require another slack variable, i.e.
max_{X} \xi
s.t.
\xi - \eta - Tr(C X) = 0
Is that the right approach? The fact that I now have two slack variables,
and use only one in the cost function strikes me as potentially problematic.
Thanks already,
nick
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/csdp/attachments/20130808/9d6aa82f/attachment.html>
More information about the Csdp
mailing list