0
$\begingroup$

I have a big matrix of size 200000 x 200000. I need to compute its inverse. But it gives out of memory error. Is there any algorithm to approximate and compute the inverse of a large sized matrix.

$\endgroup$
8
  • $\begingroup$ For what do yo need to invert the matrix?, is for solve a linear system? $\endgroup$ Mar 23, 2019 at 16:55
  • $\begingroup$ @TheStudent No, it is not a linear system. It is actually a hessian matrix in neural network. I need to invert it to compute the weight salience. $\endgroup$ Mar 23, 2019 at 17:01
  • 1
    $\begingroup$ Without a lot more context this question is impossible to answer except for the trivial "Get more memory!". $\endgroup$
    – Somos
    Mar 23, 2019 at 17:11
  • $\begingroup$ If the matrix is sparse there are ways to do it. But otherwise it should be impossible to even store the original matrix in the first place. You have $4\times 10^{10}$ entries which is a ton. $\endgroup$
    – Asinomás
    Mar 23, 2019 at 20:26
  • $\begingroup$ @JorgeFernándezHidalgo Here the matrix is not sparse. Is there an algorithm to approximate the inverse when matrix is large? $\endgroup$ Mar 23, 2019 at 20:50

0

You must log in to answer this question.

Browse other questions tagged .