Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Handling Singular matrices in Algorithms

  1. Mar 28, 2005 #1
    Hi All

    I'm new to this forum so please be kind :)

    I am doing a project on handling singular matrices in algorithms.

    Basically what i have to do is to find out how to solve the system Ax=B when A is not square or det(A)=0. Because it does not have an inverse, I dont know what to do or how to approach this problem. Any help would be appreciated

    P.S I tried the Guassian Elimination Method, but I have to find another way
  2. jcsd
  3. Mar 28, 2005 #2


    User Avatar
    Staff Emeritus
    Science Advisor
    Gold Member

    What was wrong with gaussian elimination? I had a similar problem a while back (specifically, I needed to do it numerically) and I finally settled on writing a gaussian elimination routine.
  4. Mar 28, 2005 #3


    User Avatar
    Staff Emeritus
    Science Advisor
    Gold Member

    You could apply some sort of regularization. e.g. instead solve:

    (A^T A + \lambda I) x = A^T b

    (for your favorite λ > 0)

    The system is no longer singular, but the answers you get will be biased away from the actual solutions. In particular, the answer will be unique, instead of getting a bunch of solutions.

    I don't think this would help with Gaussian elimination, though. I share SpaceTiger's question -- what's wrong with it?
  5. Mar 30, 2005 #4
    Thanx Guys,

    I really appreciated your help. Nothings wrong with the gaussian elimination method. I just have to find out "various" ways of solving the system and tha Gaussian Elimination is just another method.

    I have another question. Consider A , an n*n matrix which is invertible, and A^-1, its inverse. If A is transposed, how would it affect A^-1. (in other words how can i find the inverse of the Transpose of A without solving [A|I])

    Thanx in advence
  6. Mar 30, 2005 #5
    Look up Rank-Deficient Least Squares methods.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook