"Index out of bounds" error in SingularValueDecomposition.Solve

[ChristerModin]

Hi

First thanks a lot for all your work with this, at it seems, amazing library :)

I've run into trouble though when fitting an overdetermined system with a two-dimensional B-matrix.
Seems as the VLU isn't dimensioned properly with respect to the number of rows in U
Also, the inner loop should have the number of columns in U as upper bound instead of urows.

I suggest from line 934 in SingularValueDecomposition.cs :

        int ucols = u.Columns();
        var VLU = new Double[vrows, urows];
        for (int i = 0; i < vrows; i++)
        {
            for (int j = 0; j < urows; j++)
            {
                Double sum = 0;
                for (int k = 0; k < ucols; k++)
                    sum += VL[i, k] * u[j, k];
                VLU[i, j] = sum;
            }
        }         

BR

Christer Modin

[cesarsouza]

Hi Christer,

Many thanks for opening the issue! Can you please provide a sample matrix that triggers the problem so I can write an unit test for it?

Regards,
Cesar

[ChristerModin]

Hi Cesar Of course, how about: double[,] A = { { 1 }, { 0 } }; double[,] B = { { 1 }, { 0 } }; double[,] X = Accord.Math.Matrix.Solve(A, B, true); BR Christer From: César Souza [mailto:[email protected]] Sent: den 5 juni 2017 17:58 To: accord-net/framework <[email protected]> Cc: Christer Modin <[email protected]>; Author <[email protected]> Subject: Re: [accord-net/framework] "Index out of bounds" error in SingularValueDecomposition.Solve (#614) Hi Christer, Many thanks for opening the issue! Can you please provide a sample matrix that triggers the problem so I can write an unit test for it? Regards, Cesar — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#614 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/Aboh-mYgkCcUHBtZepwxMjI96nbZX6tZks5sBCV9gaJpZM4NnSnf>.

[ndcomplete]

This is very similar to the question I asked previously, which also includes a sample. I'm not sure if the sample provided will highlight the problem, as it is for JaggedSingularValueDecomposition, though I think it's the same underlying issue:

#602

[ChristerModin]

Yep, seems to be similar code with the same issue for all versions of svd-decomposition-solve where B isn't one-dimensional.

[cesarsouza]

Fixed in release 3.6.0.