返回题库

有限差分法

Finite difference method

专题
Algorithmic Programming / 算法编程
难度
L4

题目详情

A. 简要解释有限差分法(finite difference methods)。

B. 对于抛物型 PDE 的显式格式,时间步太多更糟还是空间步太多更糟?为什么?


A. Explain finite difference methods briefly?

B. In an explicit scheme for a parabolic PDE, is it worse to have too many time steps or too many space steps?

解析

A. 把 PDE(如 Black–Scholes)在变量变换后写成扩散类方程,在 (τ,x)(\tau,x) 上建立网格(Δτ,Δx\Delta\tau,\Delta x),用差分近似偏导数(显式/隐式/Crank–Nicolson),结合边界条件逐点求解网格上的未知值。

B. 显式格式通常有稳定性条件(示例)

Δτ<(Δx)22.\Delta\tau < \frac{(\Delta x)^2}{2}.

因此空间步长 Δx\Delta x 取太小会迫使 Δτ\Delta\tau 更小(需要极多时间步),成本更高且更容易触发不稳定。通常“空间划分过细”更糟。


Original Explanation

AnswerA:
We transform PDE (like Black-Scholes) into a diffusion PDE in (τ,x)(\tau,x), set up a grid for Δτ,Δx\Delta\tau,\Delta x. Then approximate partial derivatives (explicit/implicit/Crank-Nicolson). Solve with boundary conditions for each grid node.


AnswerB:
Explicit demands Δτ<(Δx)22\Delta\tau < \frac{(\Delta x)^2}{2} for stability. If Δx\Delta x is too small, you need extremely many time steps => bigger cost or unstable. So too many space subdivisions is worse.