Inverting Large Matrices (Overflow error)

New to Julia
,
1

Hi all,

I’m trying to invert a large matrix (1001 by 1001) to perform Ridge Regression in Julia. I’m running into this “Stack overflow Error:”, I’ve recreated the problem with simpler code (not my Ridge Regression code, which is much messier) to illustrate what I’m dealing with. I’m using Julia in a Jupyter Notebook if that is of any concern. Could someone help me invert a large matrix or solve Ax=b for large A? Thank you!

(in this example I show that even a 100 by 100 matrix triggers this response, but 50 by 50 works)

Screen Shot 2021-12-19 at 7.17.38 PM
Screen Shot 2021-12-19 at 7.17.38 PM1052×558 75 KB

2

This is a known problem with Julia 1.7.0 and 1.6.4. 1.6.3 is unaffected, and 1.7.1 and 1.6.5 will fix this.

2 Likes
3

Oh ok, thank you for your fast response. I apologize for bringing up the old issue, I couldn’t find it on google. I’ll update my software then.

4

in version 1.6.4, you can also use:

using LinearAlgebra
x = 1001
pinv(rand(Uniform(-1, 1), x, x))

on my installation, it does not appear to suffer the same issue. Provided that pinv() is adequate in your regression context.

5

On my Intel processor running Julia 1.7.0,

using MKL

eliminates the problem, which is restricted to OpenBLAS.