I have noticed, then whenever nEigen=nDim, that there is a failure in calculating eigenvectors.
Looking at the code, the error code returned from Arpack seems to indicate that
“NCV must be greater than NEV and less than or equal to N”
Which is not possible if nEigen=nDim. Is there some workaround to get all eigenvectors?
Attached is a simple example showing the issue for the matrix [(3,0),(0,0.5)].
address@hidden eigen]$ g++ main2.cpp -L /users/magala/sasopt/ppmatt/igraph-0.5.4/build-g/lib -ligraph -I /users/magala/sasopt/ppmatt/igraph-0.5.4/build-g/include/igraph -static
address@hidden eigen]$ ./a.out
Error at ../../src/arpack.c:380 :ARPACK error, NCV must be bigger
Aborted
This should return eigenvalues=(0.5,3) and eigenvectors=(0,1) and (1,0).
Note: I am using igraph-0.5.4.