Em Sex, 2009-02-06 às 22:18 +0200, ahmet alper parker escreveu: > One more comment on the subject. As we were trying to solve a > Sturm-Lioville boundary value problem by finite difference method, we > calculated eigenvalues and eigenvectors of some matrices on matlab and > mathematica. In both of the programs, their 32 bit and 64 bit results > were different (we tried these calculations on the same machine with > different windows builds). So I want to know (in my future studies) > what kind/degree/level (I have to look at my of numerical error > books) I am making in my calculations. So a good multiple precision > arithmetic and error estimation support on a Cas is extremely > important from my point of view. If you are welcome for future > development ideas on Sage, this is my personal recommendation. > Thanks for all your valuable recommendations and answers. > Best wishes... > AAP
I haven't used Mathematica, and I just touched the surface on Matlab, but it's worth pointing out that Matlab's matrices are numerical, and the displayed values may be a bit off. That's not to mention that even when displaying the result, Matlab uses approximation. For example, when one calculates an eigenvalues matrix, in Matlab it's shown as it should be, just the main diagonal with values and the others as zero, because it's approximated; in Numpy you see what the computer really calculates, which result in values in the order of 10^-9~10^-16. There's an option in Matlab which can change the order of approximation, which makes it show what's really under it all. This has generated confusion for me when I didn't get zeroes where they should be. Ronan --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
