Hi Philip,
This does not look it's being related to memory - instead I suspect that there is something fishy in the data. What happens when you run individual ICAs on a few of the input files - does that work OK? Also, have you checked (eg. with fslstats or by visual inspection in fslview) that the data is fine and has a proper intensity range? If it does not then you could try and switch off variance normalisation to get it going.
hth
Christian
On 24 Aug 2011, at 20:52, Philipp G. Saemann wrote:
> Dear FSLers,
>
> after calling melodic (group ICA, N=160, each 177 time points) with a shell command I get an error right after the PCA starts that reads:
>
> Data size : 3040 x 50801
> Starting PCA ...
> An exception has been thrown
> Logic error:- detected by Newmat: element error: requested index (wrt 0) = 3040
> MatrixType = Diag # Rows = 3040; # Cols = 3040; lower BW = 0; upper BW = 0
> Trace: Evalue(tql2).
>
> The program then stops.
>
> A totally unrelated analysis could be successfully finished though the Data size of the matrix for PCA was even larger:
>
> Data size : 5670 x 56312
> Starting PCA ... done
> [successfully finished after that]
>
>
> Does anyone have an idea what might go wrong here? The 4D-data files look okay, have the same size; the mask is correct, the call of melodic itself seems correct. The swap space of the machine is quite large.
>
> Thanks a lot for any hints to solve this problem!
> Philipp
|