add option to add a hardcoded gradient of the action

[oameye]

Performance-wise, it would be very beneficial to add the option of having a hardcoded gradient of the action function S(x::Path) in the (g)MAM solvers.

[reykboerner]

Yes, I guess by gradient you mean the functional derivative of S here with respect to the path? This would also be very useful for langevinmcmc. Maybe this can also be done symbolically? We have derived explicit equations for a general drift b and covariance matrix.

[oameye]

Yes that is exactly what I mean. In the end, it will be something in term of b and derivative of it.