A fuzzy-set theoretical framework for computing exit rates of rare events in potential-driven diffusion processes

This article is about molecular simulation. However, the theoretical results apply for general overdamped Langevin dynamics simulations. Molecular simulation is often used for determining the stability of a complex (e.g., ligand-receptor). The stability can be measured by computing the expected holding time of the complex before its dissociation. This dissociation can be seen as an exit event from a certain part S of the conformational state space. Determining exit rates (i.e, for SDE-based simulations exiting from a given starting set S) for a stochastic process in which the exit event occurs very rarely is obviously hard to solve by straight forward simulation methods. Finding a low variance procedure for computing rare event statistics is still an open problem. Imagine now, e.g., a simulation of a diffusion process. As long as the time-dependent state trajectory is inside the starting set S, no information is gained about the rare event statistics. Only at that point of time, when the process leaves the starting set, a piece of information about the exit rate is collected. If S, however, is a fuzzy set given by a membership function, then there might be additional information of the kind "the process is about to leave the set". However, how to define an exit rate from a fuzzy set?