By the way, finishing this is one of the necessary conditions for the "systems of ODEs" solver. However in the coming weeks I will not have the time to help with all this.
On 15 July 2013 15:47, Stefan Krastanov <[email protected]> wrote: > In the mamueller's PR there are removed tests, but otherwise is seems > closer to completion. Given that you have already merged it, *if* it passes > the tests that are already in master feel free to use it instead of mine. > > Please be sure that there are not two parallel implementations created > this way, because there are already some placeholders and tests in master > (by Alexey (@goodok) I think). > > > On 15 July 2013 15:40, F. B. <[email protected]> wrote: > >> OK, you pointed me out two pending PR about the Jordan form. >> >> I worked on mamueller's one and I've merged it with the latest master >> branch. I did not inspect Krastanov's PR yet, but it looks like they are >> clashing against each other. >> >> In mamueller's one the Matrix.exp( ) already works on non-diagonalizable >> matrices, which seems good. >> >> Which PR do you suggest I should look for? >> >> >> On Monday, July 15, 2013 2:47:59 AM UTC+2, Rick Muller wrote: >>> >>> You're right. I guess I see the world through a numeric lens that I >>> don't even notice anymore. >>> >>> On Sunday, July 14, 2013 5:59:47 PM UTC-6, Aaron Meurer wrote: >>>> >>>> That paper mainly deals with numeric methods, and maintaining >>>> numerical stability, which are not issues for symbolic matrices, but >>>> the Jordan method is described (briefly) as method 16. It does >>>> actually give a closed form for the exponential of a Jordan block, >>>> which can be built much more efficiently by using the form of it than >>>> by taking the powers of the matrices directly. >>>> >>>> But maybe some other method there is also useful for symbolic >>>> computation. >>>> >>>> Aaron Meurer >>>> >>>> On Sun, Jul 14, 2013 at 6:21 PM, Rick Muller <[email protected]> >>>> wrote: >>>> > There's a great article from SIAM Review of matrix exponentiation >>>> called 19 >>>> > Dubious Ways to Exponentiate a Matrix that's fun reading if people >>>> aren't >>>> > already familiar with it. May have some useful tricks. >>>> > >>>> > >>>> > On Sunday, July 14, 2013 8:35:32 AM UTC-6, F. B. wrote: >>>> >> >>>> >> >>> m = Matrix([[0, 1], [0, 0]]) >>>> >> >>> exp(m) >>>> >> NotImplementedError: Exponentiation is implemented only for >>>> diagonalizable >>>> >> matrices >>>> >> >>>> >> >>>> >> What is the best way to implement the exponentiation for >>>> non-diagonalibale >>>> >> matrices? >>>> >> >>>> >> I thought a way to fix it could be by Taylor expansion (hoping >>>> >> non-diagonalizable matrices over the complexes are nilpotent). >>>> >> >>>> >> Any better ideas? Just suggest me something and I'll try to fix it. >>>> >> >>>> > -- >>>> > You received this message because you are subscribed to the Google >>>> Groups >>>> > "sympy" group. >>>> > To unsubscribe from this group and stop receiving emails from it, >>>> send an >>>> > email to [email protected]. >>>> > To post to this group, send email to [email protected]. >>>> > Visit this group at >>>> > http://groups.google.com/**group/sympy<http://groups.google.com/group/sympy>. >>>> >>>> > For more options, visit >>>> > https://groups.google.com/**groups/opt_out<https://groups.google.com/groups/opt_out>. >>>> >>>> > >>>> > >>>> >>> -- >> You received this message because you are subscribed to the Google Groups >> "sympy" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> To post to this group, send email to [email protected]. >> Visit this group at http://groups.google.com/group/sympy. >> For more options, visit https://groups.google.com/groups/opt_out. >> >> >> > > -- You received this message because you are subscribed to the Google Groups "sympy" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/sympy. For more options, visit https://groups.google.com/groups/opt_out.
