5 Best Ways to Use SciPy to Calculate the Inverse of a Matrix in Python

by

💡 Problem Formulation: How can one utilize the SciPy library in Python to compute the inverse of a square matrix? This article addresses this challenge by demonstrating methods to invert a matrix, with an example matrix as input [[4, 7], [2, 6]] and its inverse as the desired output.

Method 1: Using scipy.linalg.inv

The scipy.linalg.inv function is a straightforward approach to compute the inverse of a matrix. It’s a part of the scipy.linalg module, which is dedicated to linear algebra operations. The inv function requires a square matrix as input and returns its inverse, provided the matrix is non-singular (i.e., it has an inverse).

Here’s an example:

from scipy.linalg import inv
import numpy as np

# Our matrix
A = np.array([[4, 7], [2, 6]])

# Calculate the inverse
A_inv = inv(A)

print(A_inv)

Output:

[[ 0.6 -0.7]
 [-0.2  0.4]]

This code snippet imports the necessary functions from SciPy and NumPy. It then defines a 2×2 matrix A and uses the inv function from SciPy’s linear algebra module to compute its inverse, which is printed to the console.

Method 2: Using scipy.linalg.solve

Another method involves using the scipy.linalg.solve function, generally used to solve linear systems. By passing the identity matrix as the second argument, scipy.linalg.solve can work out the inverse of the first argument matrix.

Here’s an example:

from scipy.linalg import solve
import numpy as np

A = np.array([[4, 7], [2, 6]])
I = np.eye(2)

A_inv = solve(A, I)

print(A_inv)

Output:

[[ 0.6 -0.7]
 [-0.2  0.4]]

The above code snippet employs the solve function from SciPy’s linear algebra module to calculate the inverse of the matrix A by treating it as a system of equations with the identity matrix I as the outcome.

Method 3: Using numpy.linalg.inv

The NumPy library also offers a way to calculate the matrix inverse with numpy.linalg.inv. While this is in the NumPy library, SciPy builds on top of NumPy and generally recommends using SciPy’s functions for more advanced tasks.

Here’s an example:

import numpy as np

A = np.array([[4, 7], [2, 6]])

A_inv = np.linalg.inv(A)

print(A_inv)

Output:

[[ 0.6 -0.7]
 [-0.2  0.4]]

The code utilizes NumPy’s linalg.inv method to compute the inverse of the matrix A. Although this is not a SciPy method, NumPy’s functions are tightly intertwined with SciPy’s, making it relevant for comparison.

Method 4: Using scipy.linalg.pinv

For matrices that are singular or not square (and therefore do not have an inverse), the scipy.linalg.pinv function provides the pseudoinverse (the Moore-Penrose inverse) of a matrix using a singular value decomposition (SVD).

Here’s an example:

from scipy.linalg import pinv
import numpy as np

A = np.array([[4, 7], [2, 6]])

A_pinv = pinv(A)

print(A_pinv)

Output:

[[ 0.6 -0.7]
 [-0.2  0.4]]

Here we demonstrate the use of pinv, which computes the pseudoinverse of A. For the provided non-singular square matrix, the pseudoinverse is the same as the inverse.

Bonus One-Liner Method 5: Using a Lambda Function

If you’re looking for a quick one-liner to compute the inverse of a matrix, combining a lambda function with the inv function can provide a compact solution.

Here’s an example:

import numpy as np
from scipy.linalg import inv

A = np.array([[4, 7], [2, 6]])

# Define the one-liner lambda function
inverse = lambda x: inv(x)

# Use it
print(inverse(A))

Output:

[[ 0.6 -0.7]
 [-0.2  0.4]]

This code creates an anonymous function that computes the inverse of a matrix using the inv function and applies it immediately to the matrix A.

Summary/Discussion

  • Method 1: Using scipy.linalg.inv. Simple and direct. Assumes the matrix is non-singular.
  • Method 2: Using scipy.linalg.solve. Versatile, can solve various linear algebra problems including matrix inversion. More computationally intensive.
  • Method 3: Using numpy.linalg.inv. Not a SciPy method but relevant due to SciPy’s reliance on NumPy. Limited to NumPy’s capabilities.
  • Method 4: Using scipy.linalg.pinv. Provides a more general solution through the pseudoinverse for singular or non-square matrices. More computationally expensive.
  • Bonus Method 5: One-liner lambda function with inv. Compact, but syntactic sugar that doesn’t improve performance or capabilities.