In an RBF network where the basis functions are identified through clustering, the typical approach to setting the spread on the basis functions is to tune a single spread parameter and then use that parameter on all of the basis function nodes. If we wanted to set individual spreads, what is the best way to go about doing that?

A. Tune each spread parameter individually using a grid search.

B. There isn't really a best way. Ultimately, it will depend on the problem.

C. Apply random spread values.

D. Tune each spread parameter individually and find the best combination of values across all of the basis functions.