Preserving correlations between trajectories for efficient path sampling

摘要: Importance sampling of trajectories has proved a uniquely successful strategy for exploring rare dynamical behaviors complex systems in an unbiased way. Carrying out this sampling, however, requires ability to propose changes pathways that are substantial, yet sufficiently modest obtain reasonable acceptance rates. Satisfying requirement becomes very challenging the case long trajectories, due characteristic divergences chaotic dynamics. Here, we examine schemes addressing problem, which engineer correlation between trial trajectory and its reference path, instance using artificial forces. Our analysis is facilitated by modern perspective on Markov chain Monte Carlo inspired non-equilibrium statistical mechanics, clarifies types strategies can scale trajectories. Viewed light, most promising such guides manipulating sequence random numbers advan...