Overview
- The norm of a square matrix describes the magnitude of the matrix.
- Matrix norm is a function that combines all the values of a matrix to form a single positive real number that satisfies the following conditions.
- || A || >= 0 for any square matrix A
- || A || >= 0 if and only if the matrix A = 0
- || kA || = | k | || A || where k is a scalar
- || A + B || <= || A || + || B ||
- || AB || <= || A || || B ||
- There are several matrix norms and each of them are calculated differently.
- The parameter ord of the numpy.linalg.norm() function denotes which matrix norm needs to be calculated.
Types of matrix norms
- Frobenius norm or Euclidean norm: Returns the square root of the sum of the squares of the
all matrix elements. This is the default norm returned by the numpy.norm() function when ord = None.
- The 1-norm: The 1-norm is calculated by taking the maximum of the sums of the matrix columns. Column sums are calculated using the absolute value of the column elements.
- The -1-norm: The -1-norm is calculated by taking the minimum of the sums of the matrix columns. Column sums are calculated using the absolute value of the column elements.
- The 2-norm: Conventionally refers to the Euclidean norm. However when ord = 2, numpy.linalg.norm() returns the maximum of the singular values of the diagonal matrix obtained through singular value decomposition.
- The -2-norm: Returns the minimum of the singular values of the diagonal matrix obtained through singular value decomposition.
- Nuclear norm: The Nuclear norm is provided by summing up the singular values obtained through singular value decomposition.
Example
# Example Python program that finds the various matrix norms of a given
# matrix using the function norm() of numpy.linalg module
import numpy
import math
# Create a 2x2 square matrix
a1 = numpy.ndarray(shape = (2, 2),
dtype = numpy.int32)
a1[0][0] = -2
a1[0][1] = -4
a1[1][0] = 3
a1[1][1] = 5
print("Square matrix:")
print(a1)
# Frobenius norm / Euclidean norm
normVal = numpy.linalg.norm(a1)
print("Euclidean norm of the matrix:")
print(normVal)
# When ord = 1
normVal = numpy.linalg.norm(a1, ord = 1)
print("1-norm of the matrix:")
print(normVal)
# When ord = -1
normVal = numpy.linalg.norm(a1, ord = -1)
print("-1-norm of the matrix:")
print(normVal)
# When ord = 2
normVal = numpy.linalg.norm(a1, ord = -2)
print("2-norm of the matrix:")
print(normVal)
# When ord = -2
normVal = numpy.linalg.norm(a1, ord = -2)
print("-2-norm of the matrix:")
print(normVal)
# When ord=inf
normVal = numpy.linalg.norm(a1, ord=numpy.inf)
print("Infinity norm of the matrix:")
print(normVal)
# When ord=-inf
normVal = numpy.linalg.norm(a1, ord=-numpy.inf)
print("Negative infinity norm of the matrix:")
print(normVal)Output
Square matrix:
[[-2 -4]
[ 3 5]]
Euclidean norm of the matrix:
7.3484692283495345
1-norm of the matrix:
9.0
-1-norm of the matrix:
5.0
2-norm of the matrix:
0.272352646999216
-2-norm of the matrix:
0.272352646999216
Infinity norm of the matrix:
8.0
Negative infinity norm of the matrix:
6.0