Here is a draft showing how one can calculate solutions with an approximation guarantee for the whole range of regularization parameters for support vector machines, multiple kernel learning, and other parameterized optimization problems.
A conference version of this work just appeared as Approximating Parameterized Convex Optimization Problems in ESA 2010. This is joint work with Joachim Giesen and Sören Laue.
Update [November 2011]:
A journal version of this paper has been accepted at the ACM Transactions on Algorithms, and is scheduled to appear soon.
Update [March 2011]:
Below are the slides of my seminar talk on this topic, generalizing this approach also for matrix problems and semidefinite optimization (e.g. regularization paths for matrix factorizations).
