2

After (quite a lengthy) discussion with Rody Oldenhuis here and another discussion here, I think it is best to make it a formal question:

  1. Why should one avoid using Matlab's inv()?
  2. What better alternatives exists for inv for different use cases?
Community
  • 1
  • 1
Shai
  • 111,146
  • 38
  • 238
  • 371
  • 5
    related question: [Why is Matlab's inv slow and inaccurate?](http://stackoverflow.com/questions/1419580/) – Eitan T Jun 24 '13 at 09:17
  • another related question is: [Singularity for inverse matrix](http://stackoverflow.com/questions/17263873/singularity-for-inverse-matrix) – Shai Jun 24 '13 at 09:20
  • 2
    A relevant post on [Loren's blog](http://blogs.mathworks.com/loren/2007/05/16/purpose-of-inv/). – Shai Jun 24 '13 at 09:57
  • 2
    In the case of linear systems you can find on the [FEX: Don't let that INV go past your eyes; to solve that system, FACTORIZE!](http://www.mathworks.co.uk/matlabcentral/fileexchange/24119-dont-let-that-inv-go-past-your-eyes-to-solve-that-system-factorize). This link has also a good publishable demo. – Oleg Jun 24 '13 at 12:16
  • 1
    I'd be interested in cases where `inv` is sufficiently accurate and actually faster than other common methods (I'm not talking about solving linear systems). I think that there's a strange sort of orthodoxy surrounding `inv` because it can be easily misused. Many of Matlab's methods are a tradeoff between precision and speed and require a good level of understanding in order to use properly, i.e., the ODE suite. It would be nice to bring back some subtlety. – horchler Jun 24 '13 at 15:04

0 Answers0