[SciPy-user] Rank-Revealing QR factorization of dense matrices in scipy

[Nils Wagner]

Robert Kern wrote:

> Nils Wagner wrote:
>
>> Hi all,
>>
>> Has someone written a function for Rank-Revealing QR factorization of 
>> dense matrices in scipy ?
>
>
> No. The relevant LAPACK function set is xGEQP3, I believe. 
> Contributions are welcome.
>
BTW, a good source is

C. H. Bischof, G. Quintana-Orti
Algorithm 782: Codes for Rank-Revealing QR Factorizations of Dense
Matrices

ACM Transactions on Mathematical software, Vol.24 No. 2 pp. 254-257


>> Is there a function to determine the numerical rank of a matrix in 
>> scipy ?
>
>
> Tip: grep the source before asking here.
>
> Offhand, I don't see one standing alone.
>
> If your arrays aren't too large, you can use linalg.svdvals to compute 
> the singular values. The number of values greater than the largest 
> value times machine epsilon is a good estimate of the rank.
>
> Alternatively, the rank* of A can be computed by linalg.lstsq(A, 
> ones(A.shape[0]))[2].
>
> * It's documented as the "effective rank." I think we're talking about 
> the same thing, though.
>
It would be nice to have a built-in function (e.g. rank(A)) for this 
purpose.
Maybe linalg.orth can be used in this context.

Nils