Methods to determine parameters for Support Vector Machines with RBF kernel other than grid search and cross validation.

I am using python scikit learn to work with SVM with an RBF kernel.

Is there a good way to solve for the Complexity cost (C value) and gamma when using an RBF kernel other than grid search then looking for the best score or cross validation? The reason why I am asking is because grid search seems kind of brute force and memory intensive.

Also when you have a large data sets you should scale everything (either between 0 and 1 or 1 and -1) I guess this is to prevent skewing of results. Is there a preference on which scale 0,1 or 1,-1? (I use 0,1 but I am just wondering.)