|Date: 2014-09-18 22:42|
Sender: Peter Eastman
Can you suggest an algorithm to use for such an integrator? The Verlet and Langevin integrators estimate the error in a step by comparing the results of two different integration methods of different orders. That is possible because they are integrating a second order equation. But the Brownian integrator is just integrating a first order equation, so that method doesn't apply.
We could get an error estimate by evaluating the hessian matrix, but that would be far too expensive to be useful.