[NMF-user] matrix size limit?

[Renaud Gaujoux]

Hi Thomas,

this is not really a large matrix and should not pose a problem, for
example on a toy dataset:

> x <- rmatrix(4000, 100)
> res <- nmf(x, 3)
> runtime(res)
   user  system elapsed
 19.888   0.004  19.899
> res <- nmf(x, 3, nrun = 3)
> runtime(res)
   user  system elapsed
 33.262   0.004  33.447
>

Do you see warnings as well? Sending a complete log would help.
The error you see is thrown if the following test is TRUE:

min(rowSums(x, na.rm = TRUE), na.rm = TRUE) == 0

Can you please directly run this test on your data? It should return
TRUE. Then look at:

s <- rowSums(x, na.rm = TRUE)
str(which( is.na(s) | s == 0 ))

which will give you a hint on which row(s) are detected as
problematic. By the way, is your dataset nonnegative?

Bests,
Renaud

On 19 May 2014 18:37, Thomas Tibbitts <Thomas.Tibbitts at infi.com> wrote:
> NMF-users,
>
>
>
> I can run NMF with a matrix of 1646 rows x 100 columns without error.
>
>
>
> However when I combine data sets into a larger matrix of 3495 rows x 100
> columns I keep getting the error:
>
> Error:  NMF::nmf  – Input matrix x contains at least one null or NA-filled
> row.
>
>
>
> There are no nulls or NA-filled rows in the larger matrix, I wrote it out to
> disk to make sure.
>
>
>
> So I ‘m thinking I hit an matrix size limitation.
>
> How can I fix this problem?
>
>
>
> Thank you.
>
> Tom
>
> This message and any attachments may contain confidential information and/or
> privileged material. If you received this message in error, please contact
> the sender and delete this message and any attachments immediately. Thank
> you.
>
> _______________________________________________
> nmf-user mailing list
> nmf-user at lists.r-forge.r-project.org
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/nmf-user
>
>



-- 
Renaud Gaujoux, PhD
Systems Immunology - Technion, Haifa, Israel