Linear Algebra in Python

Modules - Basics

Last edited: February 7th 2018

Even though matrix operations might not be as naturally built in to Python as in for instance MATLAB, Python can be used to solve linear algebra problems. In addition to functions in the numpy base library, numpy also has its own linear algebra package called linalg. In this module we will take a look at how numpy and linalg can be used to handle matrices. First of all we need to import numpy and linalg:

In [1]:
import numpy as np
import numpy.linalg as la

Basic operations

Let's first define two matrices $A$ and $B$ by using numpy arrays:

In [2]:
A = np.array([[1, 2],
              [1, 2]])
B = np.array([[2, 1],
             [1, 2]])

In order to do a regular matrix multiplication between $A$ and $B$ we use the numpy function inner(A,B):

In [5]:
print(np.inner(A, B))
[[4 5]
 [4 5]]

Powers of the same matrix can be calculated using linalg.matrix_powers(A,n), where $n$ is the integer power.

In [7]:
print(la.matrix_power(A, 2))
[[3 6]
 [3 6]]

Solving Linear Systems of Equations

Linear systems of equations $A\vec{x}=b$ can be solved using linalg.solve(A,b), which returns the vector $\vec{x}$. As an example, let's solve the system of equations $$\begin{align} 2x + y - z &= 8\\ -3x - y + 2z &= -11\\ -2x + y + 2z &= -3, \end{align}$$

In [8]:
A = np.array([[2, 1, -1],
             [-3, -1, 2],
             [-2, 1, 2]])
b = np.array([8, -11, -3])

x = la.solve(A, b)
[ 2.  3. -1.]

We see that the solution is $x=2$, $y=3$ and $z=-1$. We could also have solved the above by calculating the inverse of $A$ and multiplying it with $b$. The inverse of $A$ can be calculated using linalg.inv(A).

In [9]:
A_inv = la.inv(A)
x = np.inner(A_inv, b)
[ 2.  3. -1.]

We see that this gives the same solution, as it should!

Eigenvalue Problems

The linalg package can also be used to find the eigenvalus and eigenvectors of a matrix, that is, the vector $\vec{x}$ of a matrix $A$ which fulfills $$ A\vec{x} = a \vec{x},$$ where $a$ is a constant. Let's determine the eigenvalues and eigenvectors of the matrix $B$ defined above.

In [12]:
(a, x) = la.eig(B)
print("Eigenvalues:\n", a)
for i in range(len(x)):
    print(x[:, i])
 [3. 1.]
[0.70710678 0.70710678]
[-0.70710678  0.70710678]

From the above we see that $B$ has the eigenvalues $3$ and $1$ with eigenvectors $[1, 1]/\sqrt{2}$ and $[-1,1]/\sqrt{2}$ respectively. The function eig returns an array $a$ of eigenvalues and array $x$ with normalized eigenvectors, such that the vector x[:,i] corresponds to the eigenvalue a[i].

The above shows just some of the various functions available in the numpy and linalg libraries. For more information, see the link to the documentation below.


[1] Documentation of the Numpy Linear Algebra package.