Stationary Iterative Methods for Linear Systems

Copyright (C) 2020 Andreas Kloeckner

MIT License Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
In [2]:
import numpy as np
import scipy.linalg as la

import matplotlib.pyplot as pt

Let's solve $u''=-30x^2$ with $u(0)=1$ and $u(1)=-1$.

In [3]:
n = 50

mesh = np.linspace(0, 1, n)
h = mesh[1] - mesh[0]

Set up the system matrix A to carry out centered finite differences

$$ u''(x)\approx \frac{u(x+h) - 2u(x) + u(x-h)}{h^2}. $$

Use np.eye(n, k=...). What needs to be in the first and last row?

In [5]:
A = (np.eye(n, k=1) + -2*np.eye(n) + np.eye(n, k=-1))/h**2
A[0] = 0
A[-1] = 0
A[0,0] = 1
A[-1,-1] = 1

Next, fix the right hand side:

In [9]:
b = -30*mesh**2
b[0] = 1
b[-1] = -1

Compute a reference solution x_true to the linear system:

In [10]:
x_true = la.solve(A, b)
pt.plot(mesh, x_true)
Out[10]:
[<matplotlib.lines.Line2D at 0x7f1cd81932e8>]

Next, we'll try all the stationary iterative methods we have seen.

Jacobi

In [44]:
x = np.zeros(n)

Next, apply a Jacobi step:

In [42]:
x_new = np.empty(n)

for i in range(n):
    x_new[i] = b[i]
    for j in range(n):
        if i != j:
            x_new[i] -= A[i,j]*x[j]
        
    x_new[i] = x_new[i] / A[i,i]

x = x_new
In [43]:
pt.plot(mesh, x)
pt.plot(mesh, x_true, label="true")
pt.legend()
Out[43]:
<matplotlib.legend.Legend at 0x7f1cd2f38ef0>
  • Ideas to accelerate this?
  • Multigrid

Gauss-Seidel

In [45]:
x = np.zeros(n)
In [64]:
x_new = np.empty(n)

for i in range(n):
    x_new[i] = b[i]
    for j in range(i):
        x_new[i] -= A[i,j]*x_new[j]
    for j in range(i+1, n):
        x_new[i] -= A[i,j]*x[j]
        
    x_new[i] = x_new[i] / A[i,i]

x = x_new
pt.plot(mesh, x)
pt.plot(mesh, x_true, label="true")
pt.legend()
Out[64]:
<matplotlib.legend.Legend at 0x7f1cd28169e8>

And now Successive Over-Relaxation ("SOR")

In [92]:
x = np.zeros(n)
In [106]:
x_new = np.empty(n)

for i in range(n):
    x_new[i] = b[i]
    for j in range(i):
        x_new[i] -= A[i,j]*x_new[j]
    for j in range(i+1, n):
        x_new[i] -= A[i,j]*x[j]
        
    x_new[i] = x_new[i] / A[i,i]

direction = x_new - x
omega = 1.5
x = x + omega*direction

pt.plot(mesh, x)
pt.plot(mesh, x_true, label="true")
pt.legend()
pt.ylim([-1.3, 1.3])
Out[106]:
(-1.3, 1.3)
In [ ]: