Strange behaviour of linalg.svd() and linalg.eigh()
Hi, Ive implemented the classical MultiDimensional Scaling for the scikit learn using both functions. Their behavior surprised me for "big" arrays (10000 by 10000, symmetric as it is a similarity matrix). linalg.svd() raises a memory error because it tries to allocate a (7000000,) array (in fact bigger than that !). This is strange because the test was made on a 64bits Linux, so memory should not have been a problem. linalg.eigh() fails to diagonalize the matrix, it gives me NaN as a result, and this is not very useful. A direct optimization of the underlying cost function can give me an adequate solution. I cannot attach the matrix file (more than 700MB when pickled), but if anyone has a clue, I'll be glad. Matthieu  French PhD student Website : http://matthieubrucher.developpez.com/ Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92 LinkedIn : http://www.linkedin.com/in/matthieubrucher
Hi, I tried using Matlab with the same matrix and its eig() function. It can diagonalize the matrix with a correct result, which is not the case for linalg.eigh(). Strange. Matthieu 2008/4/17 Matthieu Brucher <matthieu.brucher@gmail.com>:
Hi,
Ive implemented the classical MultiDimensional Scaling for the scikit learn using both functions. Their behavior surprised me for "big" arrays (10000 by 10000, symmetric as it is a similarity matrix). linalg.svd() raises a memory error because it tries to allocate a (7000000,) array (in fact bigger than that !). This is strange because the test was made on a 64bits Linux, so memory should not have been a problem. linalg.eigh() fails to diagonalize the matrix, it gives me NaN as a result, and this is not very useful. A direct optimization of the underlying cost function can give me an adequate solution.
I cannot attach the matrix file (more than 700MB when pickled), but if anyone has a clue, I'll be glad.
Matthieu  French PhD student Website : http://matthieubrucher.developpez.com/ Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92 LinkedIn : http://www.linkedin.com/in/matthieubrucher
 French PhD student Website : http://matthieubrucher.developpez.com/ Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92 LinkedIn : http://www.linkedin.com/in/matthieubrucher
participants (1)

Matthieu Brucher