Newton's Method for matrix-valued functions; finding square root of given matrix

snehahl

New member
Joined
Feb 15, 2018
Messages
1
Please let me know how to solve the question below:



1. Newton's method can be extended to matrix-functions as well. For example, given a square matrix A and a real number t, the matrix-exponential etA is defined via the Taylor series for the exponential function:

. . . . .\(\displaystyle e^{tA}\, =\, 1\, +\, tA\, +\, \dfrac{(tA)^2}{2!}\, +\, \dfrac{(tA)^3}{3!}\, +\, \dfrac{(tA)^4}{4!}\, +\, ...\)

Obviously, the matrix A must be square.

(a) Derive Newton's method for finding the root of an arbitrary matrix-valued function \(\displaystyle f =f(X)\), where by "root" we mean that X is a root of f if \(\displaystyle f(X)= \mathbf{0}\), where 0 is the matrix of all zeroes. Assume that the matrix arguments of f are square and invertible.

(b) The square root of a matrix A is a matrix X such that \(\displaystyle X^T\, X= A.\) For a symmetric, positive-definite matrix A, derive the Newton iteration for finding \(\displaystyle X= \sqrt{\strut A\,}.\)

(c) Write a program using the Newton iteration you derived above to find the square root of the matrix below:

. . . . .\(\displaystyle A\, =\, \left(\begin{array}{cccc}8&4&2&1\\4&8&4&2\\2&4&8&4\\1&2&4&8\end{array}\right)\)

The stopping criterion for your Newton iteration should be when the absolute difference between elements of successive iterations is at most \(\displaystyle 10^{-10}.\)
 

Attachments

  • problem.jpg
    problem.jpg
    133.4 KB · Views: 12
Last edited by a moderator:
Computer program for Newton's method

Please let me know how to solve the question below:



1. Newton's method can be extended to matrix-functions as well. For example, given a square matrix A and a real number t, the matrix-exponential etA is defined via the Taylor series for the exponential function:

. . . . .\(\displaystyle e^{tA}\, =\, 1\, +\, tA\, +\, \dfrac{(tA)^2}{2!}\, +\, \dfrac{(tA)^3}{3!}\, +\, \dfrac{(tA)^4}{4!}\, +\, ...\)

Obviously, the matrix A must be square.

(a) Derive Newton's method for finding the root of an arbitrary matrix-valued function \(\displaystyle f =f(X)\), where by "root" we mean that X is a root of f if \(\displaystyle f(X)= \mathbf{0}\), where 0 is the matrix of all zeroes. Assume that the matrix arguments of f are square and invertible.

(b) The square root of a matrix A is a matrix X such that \(\displaystyle X^T\, X= A.\) For a symmetric, positive-definite matrix A, derive the Newton iteration for finding \(\displaystyle X= \sqrt{\strut A\,}.\)

(c) Write a program using the Newton iteration you derived above to find the square root of the matrix below:

. . . . .\(\displaystyle A\, =\, \left(\begin{array}{cccc}8&4&2&1\\4&8&4&2\\2&4&8&4\\1&2&4&8\end{array}\right)\)

The stopping criterion for your Newton iteration should be when the absolute difference between elements of successive iterations is at most \(\displaystyle 10^{-10}.\)

What are your thoughts?

Please share your work with us ...even if you know it is wrong.

If you are stuck at the beginning tell us and we'll start with the definitions.

You need to read the rules of this forum. Please read the post titled "Read before Posting" at the following URL:

http://www.freemathhelp.com/forum/announcement.php?f=33
 
Last edited by a moderator:
Top