What I don't get is that "un-normalized" eigenvectors can be pretty much anything. If you care about the specific output of Matlab / Octave, it means you understand the particular "un-normalization" that these programs use. In that case you should be able to recover it from the normalized output from numpy.
-=- Olivier 2011/12/20 Fahreddın Basegmez <mangab...@gmail.com> > I don't think I can do that. I can go to the normalized results but not > the other way. > > > On Tue, Dec 20, 2011 at 9:45 PM, Olivier Delalleau <sh...@keba.be> wrote: > >> Hmm, sorry, I don't see any obvious logic that would explain how Octave >> obtains this result, although of course there is probably some logic... >> >> Anyway, since you seem to know what you want, can't you obtain the same >> result by doing whatever un-normalizing operation you are after? >> >> >> -=- Olivier >> >> 2011/12/20 Fahreddın Basegmez <mangab...@gmail.com> >> >>> I should include the scipy response too I guess. >>> >>> >>> scipy.linalg.eig(STIFM, MASSM) >>> (array([ 3937.15984097+0.j, 3937.15984097+0.j, 3937.15984097+0.j, >>> 3923.07692308+0.j, 3923.07692308+0.j, 7846.15384615+0.j]), >>> array([[ 1., 0., 0., 0., 0., 0.], >>> [ 0., 1., 0., 0., 0., 0.], >>> [ 0., 0., 1., 0., 0., 0.], >>> [ 0., 0., 0., 1., 0., 0.], >>> [ 0., 0., 0., 0., 1., 0.], >>> [ 0., 0., 0., 0., 0., 1.]])) >>> >>> On Tue, Dec 20, 2011 at 9:14 PM, Fahreddın Basegmez <mangab...@gmail.com >>> > wrote: >>> >>>> If I can get the same response as Matlab I would be all set. >>>> >>>> >>>> Octave results >>>> >>>> >> STIFM >>>> STIFM = >>>> >>>> Diagonal Matrix >>>> >>>> 1020 0 0 0 0 0 >>>> 0 1020 0 0 0 0 >>>> 0 0 1020 0 0 0 >>>> 0 0 0 102000 0 0 >>>> 0 0 0 0 102000 0 >>>> 0 0 0 0 0 204000 >>>> >>>> >> MASSM >>>> MASSM = >>>> >>>> Diagonal Matrix >>>> >>>> 0.25907 0 0 0 0 0 >>>> 0 0.25907 0 0 0 0 >>>> 0 0 0.25907 0 0 0 >>>> 0 0 0 26.00000 0 0 >>>> 0 0 0 0 26.00000 0 >>>> 0 0 0 0 0 26.00000 >>>> >>>> >> [a, b] = eig(STIFM, MASSM) >>>> a = >>>> >>>> 0.00000 0.00000 0.00000 1.96468 0.00000 0.00000 >>>> 0.00000 0.00000 0.00000 0.00000 1.96468 0.00000 >>>> 0.00000 0.00000 1.96468 0.00000 0.00000 0.00000 >>>> 0.19612 0.00000 0.00000 0.00000 0.00000 0.00000 >>>> 0.00000 0.19612 0.00000 0.00000 0.00000 0.00000 >>>> 0.00000 0.00000 0.00000 0.00000 0.00000 0.19612 >>>> >>>> b = >>>> >>>> Diagonal Matrix >>>> >>>> 3923.1 0 0 0 0 0 >>>> 0 3923.1 0 0 0 0 >>>> 0 0 3937.2 0 0 0 >>>> 0 0 0 3937.2 0 0 >>>> 0 0 0 0 3937.2 0 >>>> 0 0 0 0 0 7846.2 >>>> >>>> >>>> Numpy Results >>>> >>>> >>> STIFM >>>> array([[ 1020., 0., 0., 0., 0., 0.], >>>> [ 0., 1020., 0., 0., 0., 0.], >>>> [ 0., 0., 1020., 0., 0., 0.], >>>> [ 0., 0., 0., 102000., 0., 0.], >>>> [ 0., 0., 0., 0., 102000., 0.], >>>> [ 0., 0., 0., 0., 0., 204000.]]) >>>> >>>> >>> MASSM >>>> >>>> array([[ 0.25907, 0. , 0. , 0. , 0. , 0. >>>> ], >>>> [ 0. , 0.25907, 0. , 0. , 0. , 0. >>>> ], >>>> [ 0. , 0. , 0.25907, 0. , 0. , 0. >>>> ], >>>> [ 0. , 0. , 0. , 26. , 0. , 0. >>>> ], >>>> [ 0. , 0. , 0. , 0. , 26. , 0. >>>> ], >>>> [ 0. , 0. , 0. , 0. , 0. , 26. >>>> ]]) >>>> >>>> >>> a, b = linalg.eig(dot( linalg.pinv(MASSM), STIFM)) >>>> >>>> >>> a >>>> >>>> array([ 3937.15984097, 3937.15984097, 3937.15984097, 3923.07692308, >>>> 3923.07692308, 7846.15384615]) >>>> >>>> >>> b >>>> >>>> array([[ 1., 0., 0., 0., 0., 0.], >>>> [ 0., 1., 0., 0., 0., 0.], >>>> [ 0., 0., 1., 0., 0., 0.], >>>> [ 0., 0., 0., 1., 0., 0.], >>>> [ 0., 0., 0., 0., 1., 0.], >>>> [ 0., 0., 0., 0., 0., 1.]]) >>>> >>>> On Tue, Dec 20, 2011 at 8:40 PM, Olivier Delalleau <sh...@keba.be>wrote: >>>> >>>>> Hmm... ok ;) (sorry, I can't follow you there) >>>>> >>>>> Anyway, what kind of non-normalization are you after? I looked at the >>>>> doc for Matlab and it just says eigenvectors are not normalized, without >>>>> additional details... so it looks like it could be anything. >>>>> >>>>> >>>>> -=- Olivier >>>>> >>>>> 2011/12/20 Fahreddın Basegmez <mangab...@gmail.com> >>>>> >>>>>> I am computing normal-mode frequency response of a mass-spring >>>>>> system. The algorithm I am using requires it. >>>>>> >>>>>> On Tue, Dec 20, 2011 at 8:10 PM, Olivier Delalleau <sh...@keba.be>wrote: >>>>>> >>>>>>> I'm probably missing something, but... Why would you want >>>>>>> non-normalized eigenvectors? >>>>>>> >>>>>>> -=- Olivier >>>>>>> >>>>>>> >>>>>>> 2011/12/20 Fahreddın Basegmez <mangab...@gmail.com> >>>>>>> >>>>>>>> Howdy, >>>>>>>> >>>>>>>> Is it possible to get non-normalized eigenvectors from >>>>>>>> scipy.linalg.eig(a, b)? Preferably just by using numpy. >>>>>>>> >>>>>>>> BTW, Matlab/Octave provides this with its eig(a, b) function but I >>>>>>>> would like to use numpy for obvious reasons. >>>>>>>> >>>>>>>> Regards, >>>>>>>> >>>>>>>> Fahri >>>>>>>> >>>>>>> >>>>> _______________________________________________ >>>>> NumPy-Discussion mailing list >>>>> NumPy-Discussion@scipy.org >>>>> http://mail.scipy.org/mailman/listinfo/numpy-discussion >>>>> >>>>> >>>> >>> >>> _______________________________________________ >>> NumPy-Discussion mailing list >>> NumPy-Discussion@scipy.org >>> http://mail.scipy.org/mailman/listinfo/numpy-discussion >>> >>> >> >> _______________________________________________ >> NumPy-Discussion mailing list >> NumPy-Discussion@scipy.org >> http://mail.scipy.org/mailman/listinfo/numpy-discussion >> >> > > _______________________________________________ > NumPy-Discussion mailing list > NumPy-Discussion@scipy.org > http://mail.scipy.org/mailman/listinfo/numpy-discussion > >
_______________________________________________ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
