Abstract:In this paper, we present a novel explicit second order scheme with one step for solving the forward backward stochastic differential equations, with the Crank-Nicolson method as a specific instance within our proposed framework. We first present a rigorous stability result, followed by precise error estimates that confirm the proposed novel scheme achieves second-order convergence. The theoretical results for the proposed methods are supported by numerical experiments.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the numerical solution of forward-backward stochastic differential equations (FBSDEs). Specifically, the authors propose a new explicit single-step second-order numerical method to solve FBSDEs. This method maintains a high order of convergence while avoiding the issue of requiring multiple initial values in multi-step methods, thereby reducing computational complexity.
### Background and Motivation
Forward-backward stochastic differential equations (FBSDEs) have wide applications in fields such as mathematical finance, stochastic optimal control, and risk measurement. However, due to the complexity of these equations, it is difficult to find analytical solutions, thus necessitating the construction of numerical algorithms for approximate solutions. Among existing numerical methods, the Euler scheme is simple but only achieves first-order convergence, while multi-step methods require multiple initial values, increasing computational complexity.
### Main Contributions of the Paper
1. **Proposed a new explicit single-step second-order numerical method**: This method combines the idea of prediction-correction, solving \(Y\) and \(Z\) completely explicitly, avoiding the nonlinear equation solving and complex matrix operations in implicit methods, thereby improving computational efficiency.
2. **Established stability results**: Through rigorous theoretical analysis, the stability of the proposed method is proven, and precise error estimates are provided, confirming the second-order convergence of the method.
3. **Numerical experiments validation**: The theoretical results are validated through numerical experiments, demonstrating the effectiveness and accuracy of the method.
### Method Overview
- **Time discretization**: The time interval \([0, T]\) is divided into \(N\) equal parts, with a step size of \(h_t = \frac{T}{N}\).
- **Prediction-correction steps**: The explicit Euler scheme is used to predict intermediate values, and the final numerical solution is obtained through correction steps.
- **Calculation of conditional expectations**: Gaussian-Hermite quadrature and spatial interpolation are used to calculate the conditional expectation \(E_x\left[\cdot\right]\).
### Numerical Experiments
- **Experiment setup**: The time step \(h_t\) is tested with different values of \(N\), specifically \(N = \frac{1}{2^m}\) (\(m = 3, \ldots, 7\)).
- **Error metrics**: Define \(err_y := |Y_0 - Y^\pi_0|\) and \(err_z := |Z_0 - Z^\pi_0|\), representing the differences between the numerical solution and the exact solution at \(t = 0\).
- **Convergence rate**: The convergence rate \(CR\) of the time step \(h_t\) is calculated using the least squares regression method.
### Conclusion
Through rigorous theoretical analysis and numerical experiments, the paper successfully demonstrates that the proposed single-step second-order explicit method has high computational efficiency and accuracy, effectively solving forward-backward stochastic differential equations. This method not only has theoretical advantages but also performs excellently in practical applications.