AN AUTOMATIC ADAPTIVE IMPORTANCE SAMPLING ALGORITHM FOR MOLECULAR DYNAMICS IN REACTION COORDINATES

In this article, we propose an adaptive importance sampling scheme for dynamical quantities of complex systems, which are metastable. The main idea of this article is to combine metadynamics, an algorithm from molecular dynamics simulation, with Girsanov's theorem, a result from stochastic analysis. With an assimilated version of the metadynamics algorithm we build a bias to reduce the metastability in the dynamical system. To correct the sampling of the modified system we apply a reweighting strategy based on Girsanov's theorem. The proposed algorithm has two advantages compared to a standard estimator of dynamic quantities: first, it is possible to produce estimators with a lower variance, and second, we speed up the generation of a typical sample.