Higher-order spring-coupled multilevel Monte Carlo method for invariant measures

Sankarasubramanian Ragunathan,Håkon Andreas Hoel

2024-03-11

Abstract:A higher-order change-of-measure multilevel Monte Carlo (MLMC) method is developed for computing weak approximations of the invariant measures of SDE with drift coefficients that do not satisfy the contractivity condition. This is achieved by introducing a spring term in the pairwise coupling of the MLMC trajectories employing the order 1.5 strong Itô--Taylor method. Through this, we can recover the contractivity property of the drift coefficient while still retaining the telescoping sum property needed for implementing the MLMC method.

Numerical Analysis,Probability

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: **How to calculate the weak approximation of the invariant measure of stochastic differential equations (SDEs) whose drift coefficients do not satisfy the contractivity condition**. Specifically, the author has developed a high - order variable - measure multi - level Monte Carlo (MLMC) method to calculate the weak approximation of the invariant measure of SDEs with drift coefficients that do not satisfy the contractivity condition.

### Problem Background

In many fields such as physics, biology, and statistical mechanics, calculating the invariant measure of SDEs is very important. However, the traditional MLMC method usually requires that the drift coefficient satisfies the contractivity condition (i.e., one - sided Lipschitz continuous and the constant is negative), which limits its application range. For SDEs that do not satisfy this condition, traditional methods may lead to unstable numerical solutions or high computational costs.

### Solution

To overcome this problem, the author introduced a "spring term" and restored the contractivity of the drift coefficient by using the 1.5 - order strong Itô - Taylor method while retaining the telescoping sum property required by the MLMC method. This method can effectively handle a wider class of SDEs while maintaining computational efficiency.

### Main Contributions

1. **Improved Computational Complexity**: For uniformly Lipschitz continuous payoff functions, this method can achieve weak approximation with a mean - square error accuracy of O(ϵ²) at a computational cost of O(ϵ⁻²|log ϵ|³/₂ (log |log ϵ|)¹/₂). For discontinuous payoff functions, the computational cost is O(ϵ⁻²|log ϵ|⁵/₃⁺ξ), where ξ > 0.
2. **Better Constant Factor**: Compared with the Milesian variable - measure method, this method has a significant improvement in the constant factor of the computational cost.
3. **Theoretical Verification**: Through multiple numerical tests, the theoretical results are verified and the robustness of the method is evaluated.

### Formula Representation

The key formulas involved in the paper are as follows:

- The form of SDE:
\[
dX(t) = a(X(t)) dt + dW(t), \quad X(0) = x_0 \in \mathbb{R}^d
\]
where \(a: \mathbb{R}^d \to \mathbb{R}^d\) is the drift coefficient and \(W(t)\) is the standard Wiener process.

- The expected value of the invariant measure:
\[
E[\Phi(X_\infty)] := \int_{\mathbb{R}^d} \Phi(x) \pi(dx) = \lim_{t \to \infty} E[\Phi(X_t)]
\]

- The fine - grid trajectory (Fine trajectory, P - measure) after introducing the spring term:
\[
\begin{aligned}
\tilde{Y}_f(t_{2n+1}) &= \tilde{Y}_f(t_{2n}) + S h (\tilde{Y}_c(t_{2n}) - \tilde{Y}_f(t_{2n})) + h a(\tilde{Y}_f(t_{2n})) + \Delta W_P^{2n} \\
&\quad + D a(\tilde{Y}_f(t_{2n})) \Delta Z_P^{2n} + \frac{h^2}{2} A a(\tilde{Y}_f(t_{2n}))
\end{aligned}
\]

Through these improvements, this method can effectively calculate the invariant measure of SDEs under more general conditions, thus expanding the application range of the MLMC method.