Alix Leroy,Benedict Leimkuhler,Jonas Latz,Desmond J. Higham
Abstract:We discuss the design of an invariant measure-preserving transformed dynamics for the numerical treatment of Langevin dynamics based on rescaling of time, with the goal of sampling from an invariant measure. Given an appropriate monitor function which characterizes the numerical difficulty of the problem as a function of the state of the system, this method allows the stepsizes to be reduced only when necessary, facilitating efficient recovery of long-time behavior. We study both the overdamped and underdamped Langevin dynamics. We investigate how an appropriate correction term that ensures preservation of the invariant measure should be incorporated into a numerical splitting scheme. Finally, we demonstrate the use of the technique in several model systems, including a Bayesian sampling problem with a steep prior.
What problem does this paper attempt to address?
The core problem that this paper attempts to solve is: in Langevin dynamics simulations, how to effectively sample the invariant measure through an adaptive step - size algorithm, in order to improve the efficiency of recovering long - time behavior and reduce computational costs. Specifically, the article focuses on how to intelligently adjust the step - size of numerical integration in cases where the system state changes drastically or exhibits highly oscillatory patterns, so that the step - size is decreased only when necessary, thereby achieving more efficient sampling.
It is mentioned in the paper that in the traditional fixed - step - size method, in order to ensure the stability of the numerical method, the step - size must be small enough to adapt to the most extreme cases. However, this method will lead to a significant increase in computational overhead, especially in cases where the frequency changes when the system visits different regions. Therefore, the author proposes a method based on time rescaling, combined with an appropriate monitoring function, which can reduce computational costs while maintaining stability and accuracy.
The key lies in introducing a time variable τ that can be adjusted according to the system state and defining a new time relation:
\[ dt = g(x(t(τ))) dτ \]
and the corresponding Brownian motion transformation:
\[ dW(t) = \sqrt{g(\tilde{x})} d\tilde{W}(τ) \]
In addition, in order to ensure the fidelity of the invariant measure, the author also introduces a correction term to ensure that the transformed SDE still has the required Gibbs - Boltzmann distribution as its invariant measure. The ultimate goal is to design an adaptive step - size algorithm that can both sample efficiently and maintain the accuracy of the target distribution.
### Formula Summary
1. **Langevin Dynamics Equation**:
\[
\begin{cases}
dx = M^{-1}p \, dt \\
dp = -\nabla_x V(x) \, dt - \gamma p \, dt + \sqrt{\frac{2\gamma}{\beta}} M^{1/2} dW(t)
\end{cases}
\]
where \( V(x) \) is the potential energy function, \( W(t) \) is the standard Brownian motion, \( \beta=(k_B T)^{-1} \) is the inverse temperature, and \( k_B \) is the Boltzmann constant.
2. **Time Rescaling Relation**:
\[
dt = g(x(t(τ))) dτ
\]
\[
dW(t) = \sqrt{g(\tilde{x})} d\tilde{W}(τ)
\]
3. **Direct Time Rescaling SDE**:
\[
d\tilde{x} = -\nabla V(\tilde{x}) g(\tilde{x}) dτ + \sqrt{\frac{2}{\beta} g(\tilde{x})} d\tilde{W}(τ)
\]
4. **Invariant Measure Fidelity - Transformed SDE**:
\[
dx = -g(x) \nabla V(x) dt + \frac{1}{\beta} \nabla g(x) dt + \sqrt{\frac{2}{\beta} g(x)} dW
\]
5. **Euler - Maruyama Discretization Scheme**:
\[
X_{n + 1}=X_n - \nabla V(X_n) g(X_n) h + \frac{1}{\beta} \nabla g(X_n) h + \sqrt{\frac{2}{\beta} g(X_n) h} Z_n
\]
Through these formulas and methods, the author aims to provide an adaptive step - size algorithm that can perform numerical simulations efficiently and accurately in complex systems.