Statistical mechanics of histories: a cluster Monte Carlo algorithm.

Area of Science:

• Computational physics
• Statistical mechanics
• Stochastic processes

Background:

• Nonlinear stochastic processes are fundamental in many scientific fields.
• Efficiently sampling their histories, especially for rare events, remains a challenge.
• Existing methods struggle with complex systems and limited data.

Purpose of the Study:

• To present an efficient computational approach for sampling histories of nonlinear stochastic processes.
• To develop a cluster algorithm for improved sampling efficiency.
• To enable accurate state and parameter estimation using available measurements.

Main Methods:

• Casting a d-dimensional stochastic dynamical system into a (d+1)-dimensional equilibrium system via path-integral formulation.
• Introducing a novel cluster algorithm for efficient history sampling.
• Incorporating available measurements into the estimation of histories.

Main Results:

• The developed cluster algorithm significantly improves sampling efficiency, up to an order of magnitude.
• The approach is applicable to simulating rare events in stochastic systems.
• Demonstrated utility in Phi4 Langevin dynamics in two spatial dimensions.

Conclusions:

• The presented computational approach offers an efficient way to sample histories of nonlinear stochastic processes.
• This method enhances the simulation of rare events and optimal estimation tasks.
• The cluster algorithm provides a significant improvement over existing sampling techniques.