How to reduce the time to calculate the inverse of a matrix

Question

Raviteja Panamgipalli 2013 年 4 月 23 日

0
リンク

この質問への直接リンク

https://jp.mathworks.com/matlabcentral/answers/73226-how-to-reduce-the-time-to-calculate-the-inverse-of-a-matrix

I have a 64 x 64 matrix,and it is taking a lot of time like 15-20 minutes to compute the inverse,Let my matrix be A,I'm using the command inv(A) to compute the inverse,Is there any alternative method to make my function work faster,Thank you in advance :)

2 件のコメント
なしを表示なしを非表示

Friedrich 2013 年 4 月 23 日

編集済み: Friedrich 2013 年 4 月 23 日

Are you sure your dimension is correct? A Matrix like this shouldnt take longer than a few seconds. Can you give us an example MATRIX and a tic,toc output of tic,inv(A);toc? What machine are you running MATLAB on? I cant believe it takes 15-20min.

Edric Ellis 2013 年 4 月 23 日

Why do you need inv? See Loren's blog on the purpose of inv here: http://blogs.mathworks.com/loren/2007/05/16/purpose-of-inv/

サインインしてコメントする。

サインインしてこの質問に回答する。

Answer 1

Matthias 2013 年 4 月 23 日

0
リンク

この回答への直接リンク

https://jp.mathworks.com/matlabcentral/answers/73226-how-to-reduce-the-time-to-calculate-the-inverse-of-a-matrix#answer_83186

I think the inv function implemented by Mathworks is as fast as it gets for the general purpose. There are two inv-functions in the Ligthspeed library, but it seems like these are designed for special cases. Your should see if these cases apply with your Matrix.

The other possibility will be using better Hardware. Since inv() is implemented in cuda by Mathworks ill assume it is parallelizable. Make sure to load all your cores using 'matlabpool' and see if your GPU hat CUDA 1.3.