CODEWITHSUNDEEP
mathnumpy
LeafGlowPath
LeafGlowPath

Reputation: 3799

How to solve homogeneous linear equations with NumPy?

If I have homogeneous linear equations like this

array([[-0.75,  0.25,  0.25,  0.25],
       [ 1.  , -1.  ,  0.  ,  0.  ],
       [ 1.  ,  0.  , -1.  ,  0.  ],
       [ 1.  ,  0.  ,  0.  , -1.  ]])

And I want to get a non-zero solution for it. How can it be done with NumPy?

EDIT

linalg.solve only works on A * x = b where b does not contains only 0.

Upvotes: 10

Views: 15749

Answers (3)

alon ein eli
alon ein eli

Reputation: 51

if you are familiar with the SVD decomposition you can use:

_, _, V = np.linalg.svd(A, full_matrices=True)

v=V[-1]

then v is the last row of the last matrix in the decomposition, which correspond to the minimal singular value. thus if you start with a matrix A nxm with n>m then the last singular value must be zero, it means that Av=0

Upvotes: 1

Sasha Nicolas
Sasha Nicolas

Reputation: 169

For that matter, the best solution of an over constrained homogeneous linear system is the eigenvector associated with the smallest eigenvalue. So given U as the coefficient matrix of the system, the solution is:

import numpy as np

def solution(U):
    # find the eigenvalues and eigenvector of U(transpose).U
    e_vals, e_vecs = np.linalg.eig(np.dot(U.T, U))  
    # extract the eigenvector (column) associated with the minimum eigenvalue
    return e_vecs[:, np.argmin(e_vals)]

Upvotes: 9

rcs
rcs

Reputation: 68829

You can use an SVD or a QR decomposition to compute the null space of the linear system, e.g., something like:

import numpy

def null(A, eps=1e-15):
    u, s, vh = numpy.linalg.svd(A)
    null_space = numpy.compress(s <= eps, vh, axis=0)
    return null_space.T

This yields for your example:

>>> A
matrix([[-0.75,  0.25,  0.25,  0.25],
        [ 1.  , -1.  ,  0.  ,  0.  ],
        [ 1.  ,  0.  , -1.  ,  0.  ],
        [ 1.  ,  0.  ,  0.  , -1.  ]])

>>> null(A).T
array([[-0.5, -0.5, -0.5, -0.5]])

>>> (A*null(A)).T
matrix([[ 1.66533454e-16, -1.66533454e-16, -2.22044605e-16, -2.22044605e-16]])

See also the section Numerical computation of null space on Wikipedia.

Upvotes: 9

Related Questions