Non-linear optimization with no derivatives
I'm trying to optimize a convex (or concave?) function which is a sum of max functions. Is there a function that does this?
For example, I'd like to minimize
max(r_0 + s_0, 0) + max(r_0 + s_1, r_1 + s_0, 0) + max(r_0 + s_2, r_1 + s_1, r_2 + s_0, 0) + max(r_1 + s_2, r_2 + s_1, 0) + max(r_2 + s_2, 0)
subject to some linear constraints in terms of r_i and s_i.
I saw minimize_constrained, but it required the the minimizing function to have a defined derivative, which is not built in.