SciPy - solveh_banded()Function

A banded matrix is one in which most elements are zero except for some of the diagonals closest to the main diagonal.

SciPy's solveh_banded(), that solves linear equation systems Ax=b when A is Hermitian or symmetric positive definite banded matrix. This approach saves memory and processors time since it takes advantages of A's banded structure.

This technique is useful for sparse banded matrices. It proves useful in signal processing, physics simulations, and finite element analysis. It reduces computing complexity compared to dense solutions making it ideal for big problems.

This method works well for many math and science use because it handles both upper and lower triangular banded matrices. It's great for large systems ensuring top performance in real-world applications.

Compared to generic solvers, this approach is faster and uses less memory since it is specifically made to handle the structure of Hermitian or symmetric banded matrices.

Syntax

Following is the syntax of the SciPy solveh_banded() method −

.solveh_banded(ab, b, overwrite_ab=False, overwrite_b=False, lower=False, check_finite=True)

Parameters

Following are the parameters of solveh_banded() method −

• ab (array_like) − The upper or lower triangle of the banded Hermitian matrix A, stored in a 2D array.

• b array_like, shape (n,) or (n, k) − Input equation of Right-hand side.

• overwrite_ab (bool, optional) − Allows overwriting data in ab for memory efficiency.

• overwrite_b (bool, optional) − If True, allows overwriting data in b to save memory. Default is False.

• lower (bool, optional) − Specifies whether the data in ab corresponds to the lower triangular part of the matrix. Default is False (upper triangular).

• check_finite (bool, optional) − If True, checks if the input contains only finite numbers. Disabling this can improve performance but risks issues with invalid inputs.

Return Value

x (ndarray) − The solution to the system Ax=b. The shape of the solution matches with b.

Example 1

In this code, we created a Hermitian banded matrix represented by ab. The function solveh_banded() solves the system Ax = b.

import numpy as np
import scipy.linalg
from scipy.linalg import solveh_banded

# Upper triangle of the Hermitian banded matrix
ab = np.array([[0, 2, 3], [4, 5, 6]])

# Right-hand side vector
b = np.array([7, 8, 9])

# Solve the system
x = scipy.linalg.solveh_banded(ab, b)
print(x)

When we run the above program, it produces the following result

[1.75 0.   1.5 ]

Example 2

In-place matrix modification is done using the overwrite_ab=True parameter in solveh_banded(), which saves memory at runtime, especially for large systems.

In this code, we used the overwrite_ab=True parameter to allow in-place modifications of the matrix ab for memory efficiency.

import numpy as np
from scipy.linalg import solveh_banded

# Upper triangle of the Hermitian banded matrix
ab = np.array([[0, 2, 3], [4, 5, 6]], dtype=float)
b = np.array([10, 20, 30], dtype=float)

# Solve the system with overwrite_ab for memory efficiency
x = solveh_banded(ab, b, overwrite_ab=True)
print(x)

Following is an output of the above code

[2.5 0.  5. ]