SymPy was accepted as a GSoC org
Aaron Meurer
Interested students, please read our instructions on how to apply here
https://github.com/sympy/sympy/wiki/GSoC-2017-Student-Instructions.
The most important things are to pick an idea from our ideas list to
discuss with us, and to start working on a patch to fulfill your patch
requirement.
Mentors, I will be sending you invites on Google's website. If you
never signed up on the ideas page, please let me know so I can invite
you.
Aaron Meurer
Gaurav Dhingra
I've been thinking about applying again, though I am not sure
what the project should be. Are there any good algorithms that are
not implemented that could make a good project (I've read the
ideas page)? I have 3 projects in mind:
(a). Implementation of Karr's algorithm, I believe no one has done
much work except Matthew Rocklin who did work on Concrete module.
(b). Complex Analysis: I quote the statement by Kalevi:
> What I think should be added to SymPy is the Laurent series
expansion of meromorphic functions.
> The trouble with the current implementation of limit is that
it often goes too early to gruntz.
> That should only be used for functions that are not
meromorphic but have an essential singularity.
Also he opened a few issues on SymPy which might also use Complex
Analysis. Plus, since I have a course on Complex Analysis, which include
topics . But I am not sure even if it is possible to
implement these things in a Computer Algebra System.
(c). Last summer Sumith mentioned about A
dedicated bug fixing project in GSoC . Since I feel like I
can try to handle quite a few issues of multiple modules, I would
want to apply for this one if the above two doesn't have a mentor
alloted to them.
I might not want to mentor a project, since that seems like too
much pressure for me. Applying as a student would be my priority.
Gaurav Dhingra
Ondřej Čertík
https://github.com/sympy/sympy/issues/12233
Ondrej
>
> Gaurav Dhingra
>
>
> On Monday 27 February 2017 11:05 PM, Aaron Meurer wrote:
>
> SymPy was accepted as a GSoC org again this year.
>
> Interested students, please read our instructions on how to apply here
> https://github.com/sympy/sympy/wiki/GSoC-2017-Student-Instructions.
> The most important things are to pick an idea from our ideas list to
> discuss with us, and to start working on a patch to fulfill your patch
> requirement.
>
> Mentors, I will be sending you invites on Google's website. If you
> never signed up on the ideas page, please let me know so I can invite
> you.
>
> Aaron Meurer
>
>
> 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 sympy+un...@googlegroups.com.
> To post to this group, send email to sy...@googlegroups.com.
> Visit this group at https://groups.google.com/group/sympy.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sympy/fd4b4702-983b-4dbd-00f6-bff8b9142e0e%40gmail.com.
>
> For more options, visit https://groups.google.com/d/optout.
Aaron Meurer
To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/CADDwiVBUVroL6MCUFjtjoLBG1b6Pgqf_QA%2Bcu%2BdMv-zyi%2Bc_ZQ%40mail.gmail.com.
Robert Lee
Gaurav Dhingra
Hi Aaron,
I might want to look at Risch integration algorithm. Seeing the idea, it seems like Chetna worked on it in summer 2013. Did anyone else (other than Chetna) worked on the integration module (probably in GSoC) after your 2010 GSoC project? Since that would give me a list of things to look at for deciding as to what remains to be done by reading their blogs.
Gaurav Dhingra
To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/CAKgW%3D6JFLu%3DEx2KWJy%2Bi7x7TQ-OPrV4UZMw36Kh9xzx5cKEbZw%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.
-- Thanks Gaurav Dhingra (sent rom Thunderbird email client)
Gaurav Dhingra
Also I see a list of references for it here https://github.com/sympy/sympy/wiki/Technical-References#symbolic-integration
, which among these would be good to start first? (brief idea
would suffice)
Gaurav Dhingra
To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/CAKgW%3D6JFLu%3DEx2KWJy%2Bi7x7TQ-OPrV4UZMw36Kh9xzx5cKEbZw%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.
-- Thanks Gaurav Dhingra (sent from Thunderbird email client)
Aaron Meurer
Transcendental Functions". Most of what is in that book has already
been implemented, but the trigonometric case has not. Bronstein's
paper "Symbolic Integration Tutorial" (which can be found here
https://www-sop.inria.fr/cafe/Manuel.Bronstein/publications/mb_papers.html)
gives a high level overview of the whole algorithm, including the
algebraic and mixed cases (his book only deals with the transcendental
case). Unfortunately, the "easiest" bits from the book have already
been implemented, and what remains are algorithms that don't have
pseudocode. The algebraic case will require more references than
Bronstein's book.
As for the work since Chetna, other than the still open PRs
https://github.com/sympy/sympy/pulls/cheatiiit, the only work I am
aware of is this pull request by jksuom
https://github.com/sympy/sympy/pull/11761.
Aaron Meurer
Gaurav Dhingra
Considering my mathematical background and knowledge and after consulting Kalevi, I have doubt over my ability to work on Symbolic Integration (especially for algebraic part integration). But considering the amount work that remains to be done in the transcendental part integration + Completing the un-merged pull requests (by Kalevi and Chetna) do you think that would be enough for a GSoC project? Since I don't want to get into the Algebraic part integration.
Thanks
Aaron Meurer
be done for the trigonometric case.
Aaron Meurer
On Sun, Mar 12, 2017 at 10:55 AM, Gaurav Dhingra
Gaurav Dhingra
There does not seem to be any report of the project of
Chetna, https://github.com/sympy/sympy/wiki/GSoC-2013-Report,
or may be I couldn't find it anywhere. In any case I have started
writing the proposal for completing the Transcendental part
integration, since I very much like the impact of the
project.
-- Gaurav Dhingra (Sent from Thunderbird email client)
Archit Saxena
How should i search for a mentor?Should i fix a bug and then contact you again for the above mentioned projects?
Thanking you
ARCHIT SAXENA
BITS PILANI
Aaron Meurer
To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/7943b619-a2f6-3778-fb52-786318f0284b%40gmail.com.
Gaurav Dhingra
I am seeing what remains to be done. Chetna's pull request PR #2380 is an implementation of Coupled Differential System(+ other things), which is present as Chapter 8 in the Symbolic Integration (Manuel Bronstein). In total there are 9 chapters, and looking at chapter 9 doesn't contain much of the pseudo code or is it just a theoretical chapter for aide? I don't know if the all the chapters from book (Manuel Bronstein) were implemented in order without skipping any.
To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/CAKgW%3D6JJTi9ehkC2JPtjh3H6U3Ap-BVFd0wm215TNnbwPQZRpA%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.
Aaron Meurer
theoretical, but it is needed for the implementation. It tells you how
you can build the differential extension in a way that you know is
transcendental. For instance, you can prove (from the theorems in
chapter 9) that log(exp(x) + 1) is transcendental over QQ(x, exp(x)),
so it can be added as a differential extension. On the other hand,
log(exp(2*x)) is not, and the algorithm, which comes from the theorems
of chapter 9, tells you how to rewrite it in terms of x and exp(x) (in
this case, 2*x).
This is important because the algorithm everywhere assumes that the
extensions are transcendental. A nontranscendental (i.e., algebraic)
extension could produce a wrong answer (although, and this is an
interesting direction for research, I believe only a wrong answer in
the sense that it might tell you something is not elementary when it
isn't, see https://github.com/sympy/sympy/issues/11770).
For example, on that issue, I naively integrate f = exp(log(x)/2)
using risch, and it thinks it is nonelementary (it isn't, since f =
sqrt(x)). The problem is that the algorithm's decision procedures for
when an integral is nonelementary are based on the extensions being
transcendental, and QQ(x, log(x), exp(log(x)/2)) is not transcendental
(since y = exp(log(x)/2) is algebraic over the rest, satisfying y**2 -
x = 0). The structure theorems tell us how to reject exp(log(x)/2)
(or, if it were something like exp(log(x)*2), to just rewrite it in
terms of the rest). If you want to understand how this works, you
might try running through the code for risch_integrate(exp(log(x)/2,
evaluate=False), x) in a debugger.
For exponentials and logs, the theorems are pretty straightforward
(although the book doesn't have pseudocode, as I recall). They are
implemented in is_deriv_k and is_log_deriv_k_t_radical in prde.py. We
need a similar algorithm for the tangent case, to tell if a tangent is
transcendental over an extension. I forget what the exact structure
theorem looks like (I'll need to check my book later when I get home),
but I believe it is also relatively straightforward.
Yes, some of the book was skipped. Only half of chapter 7 was
implemented (Kalevi's PR does some work in this area). There are also
sections of the previous chapters that depend on things that aren't
yet implemented, so they aren't done yet either.
Aaron Meurer
On Tue, Mar 14, 2017 at 9:03 AM, Gaurav Dhingra