Differences between revisions 1 and 38 (spanning 37 versions)
Revision 1 as of 2009-06-20 15:18:39
Size: 348
Editor: was
Comment:
Revision 38 as of 2009-06-27 09:55:46
Size: 8032
Comment:
Deletions are marked like this. Additions are marked like this.
Line 2: Line 2:
Line 6: Line 7:
PEOPLE: William Stein, wouter, dkohel
Line 7: Line 10:

== Create elliptic curve classes for elliptic curve models in the Explicit-Formulas Database ==

PEOPLE: David Kohel, Wouter Castryck

  In order to optimize and compare arithmetic, we should first implement alternative models and verify relative performance. The isomorphisms between different models should also be implemented, and classes for isogenies of these models developed, making use first of the new isogenies code, and eventually putting in place special optimized code for specific models.
(Wouter) As a comment to that, none of these alternative models cover the whole range of elliptic curves (as far as I know). E.g. an elliptic curve can be shaped into Edwards form only if it has a rational point of order 4. So one should in any case optimize arithmetic on pure Weierstrass curves (using weighted projective coordinates or so).

See the EFD: http://www.hyperelliptic.org/EFD/

== Rewrite abelian groups (hard) ==

PEOPLE: William Stein, rlmiller, dkohel, wstein, dloeffler, bjarke

  . It would be possible to use [[http://trac.sagemath.org/sage_trac/ticket/5882|trac 5882]] to rewrite abelian groups natively in Sage (not using GAP), in a way that is much more flexible than the current implementation. This could be useful for many number theory applications.

  . This project is "hard", since many have tried and always failed. The reason for this is the temptation to rewrite the current implementation all at once. See more details for the new plan, which avoids this pitfall...

  . [[/abgrp|More Details]]

== Multisets ==

PEOPLE: dkohel, rlmiller, brian

== Optimize/better document/generally improve graph theory library in Sage ==

PEOPLE: Robert Miller

  Ticket [[http://trac.sagemath.org/sage_trac/ticket/6085|#6085]] contains a lot of work so that a graph created by Graph(implementation='c_graph') is just as functional as a Sage graph. I will be sporadically working on improving documentation and optimizing graphs all week, and anyone interested is welcome to join.

== Cliquer SPKG for Sage ==

PEOPLE: Robert Miller, Nathann Cohen (remotely), brian, ncohen, rlmiller, bjarke

 * [[http://trac.sagemath.org/sage_trac/ticket/6355|#6355]]
  * [[http://trac.sagemath.org/sage_trac/ticket/5793|#5793]]
  * [[http://trac.sagemath.org/sage_trac/ticket/5669|#5669]]
 * [[http://groups.google.com/group/sage-devel/browse_thread/thread/8afc622316dc102d|sage-devel thread]]

== Take a look at the possibility of making GAP a dynamically loadable library ==

PEOPLE: Robert Miller, wstein

  * Robert Miller and William Stein succeeded in making a first very rough Cython module that links in GAP as a library and allows one to type commands to the interpreter. See http://trac.sagemath.org/sage_trac/ticket/6391/. This proof-of-concept that not only can GAP be linked as a library instead of used via pexpect, but that the result is likely to be *better* than libPARI and libSingular, in that one has full access to the GAP interpreter scripts and capabilities (just like embedding Python allows for that). Also, it turns out that so far there are no issues because of the GAP memory manager.

== Python implementation of Ford-Fulkerson algorithm ==

PEOPLE: Robert Miller, brian, michael, tom

I plan on at least copying the Python implementation on wikipedia, since now we have nothing at all for max flow problems. Hopefully then someone who really cares about it will try to use it, realize it is slow, start improving it, etc. etc. etc.

 * http://en.wikipedia.org/wiki/Ford-Fulkerson_algorithm

== Frobenius number and genus of numerical semigroups using toric Grobner bases ==

PEOPLE: Bjarke Hammersholt Roune, wouter

  I plan to code Frobenius number (largest gap) and genus (number of gaps) functions for numerical semigrups using two related algorithms based on toric ideals. These algorithms can handle random numerical semigroups generated by numbers with thousands of digits, as long as there are not too many minimal generators. I'm happy to explain either algorithm if you want to help or are just curious. The steps needed are these:

   * Find the best way to compute toric Grobner bases in Sage (4ti2?)
   * Improve the integration of the library Frobby for monomial ideal computations.
   * Code the algorithms (should be easy at this point)

  Help is welcome, especially for writing a Cython interface to Frobby.

== PolyBoRi and Singular ==

PEOPLE: Martin Albrecht, Burcin Erocal

Both PolyBoRi's (Gröbner bases in GF(2)[x_i]/<x_i^2+x_i>) and Singular's interface need some (engineering) work & documentation.

== Elliptic curves isogenies -- greatly improve usability ==

PEOPLE: William Stein, dkohel, wouter, rlmiller, hamish

Dan Shumow wrote new code for computing isogenies between elliptic curves, but it is still very rough around the edges. Fix it up. See e.g., http://trac.sagemath.org/sage_trac/ticket/6384

== Siegel Theta Functions ==

PEOPLE: guardia, holuk, drkohel, rlmiller, preston, schulze-pillot

See http://trac.sagemath.org/sage_trac/ticket/6371 for code that calculates classical theta functions, their derivatives, and some related functions I'm interested in, to arbitrary precision.

== Multimodular reconstruction implementation ==

PEOPLE: clement, wstein, burcin, rlmiller, dloeffler, sebastian, brian

   * CRT in rings/arith.py is unsanely slow
   * !MultimodularBasis class is broken

Fix / Unify code / optimize the hell out of it

Add the early termination technique for the reconstruction of multiple values
Apply it to Burcin's computation of nullspace over Z[x]

PEOPLE: Burcin Erocal, Clément Pernet

== Minpoly over small finite field ==

PEOPLE: clement, michael, brian, preston, emmanuel

[[http://trac.sagemath.org/sage_trac/ticket/6296|#6296]]
Implement a certified computation of minpoly over a small finite field.

PEOPLE: Clément Pernet

== Refactor symbolic functions ==

PEOPLE: burcin, wstein, tom

Merge the `PrimitiveFunction` and `SFunction` classes in `sage.symbolic.function` and make it easier to define new special functions with custom evaluation, printing, etc. methods. This also involves adding capability to represent symbolic sums and integrals in Sage as a first step to adding native Sage code to deal with these.


== M4RI: improvement of PLUQ factorization ==

PEOPLE: Martin Albrecht, Clément Pernet

Change/improve the slow column swap in PLUQ factorization

 * We improved the PLUQ/LQUP base case by using up to four tables now instead of only two.
 * We improved the implementation for various column swaps and permutation matrix multiplications.
 * We moved from PLUQ to LQUP because it requires less column swaps.

{{attachment:lqup-vs-pluq.png}}}

{{attachment:m4ri-pluq-sparse-ish.png}}
{{attachment:m4ri-dense-10.png}}
{{attachment:m4ri-dense-20.png}}
{{attachment:m4ri-dense-30.png}}

== LinBox wrappers ==

Rewrite LinBox wrappers using C++ wrappers of Cython directly

PEOPLE: Burcin Erocal, Clément Pernet, rlmiller, malb

== Ray class groups and Groessencharacters ==

Implement computations of the ray class group modulo an ideal of a number field, and the corresponding character group.

PEOPLE: David Loeffler, rlmiller, wstein, dkohel, haluk, michael


== Pari 2.4 Upgrade ==

  PEOPLE: wstein, dloeffler, hamish, michael

== Sage notebook security ==

  PEOPLE: Yoav, malb

 We had some nice design discussions.

== Texmacs --> Sage Worksheet converter ==
  PEOPLE: Offray

== SPD (simple python distro) ==
  PEOPLE Offray

Sage Days 16 Project Idea Page

Create a Cython class for points on elliptic curves and optimize basic arithmetic

PEOPLE: William Stein, wouter, dkohel

  • Right now basic arithmetic on elliptic curves is way too slow. It could be sped up by moving the point class to Cython, and possibly by using better formulas for arithmetic, e.g., using projective coordinates.

Create elliptic curve classes for elliptic curve models in the Explicit-Formulas Database

PEOPLE: David Kohel, Wouter Castryck

  • In order to optimize and compare arithmetic, we should first implement alternative models and verify relative performance. The isomorphisms between different models should also be implemented, and classes for isogenies of these models developed, making use first of the new isogenies code, and eventually putting in place special optimized code for specific models.

(Wouter) As a comment to that, none of these alternative models cover the whole range of elliptic curves (as far as I know). E.g. an elliptic curve can be shaped into Edwards form only if it has a rational point of order 4. So one should in any case optimize arithmetic on pure Weierstrass curves (using weighted projective coordinates or so).

See the EFD: http://www.hyperelliptic.org/EFD/

Rewrite abelian groups (hard)

PEOPLE: William Stein, rlmiller, dkohel, wstein, dloeffler, bjarke

  • It would be possible to use trac 5882 to rewrite abelian groups natively in Sage (not using GAP), in a way that is much more flexible than the current implementation. This could be useful for many number theory applications.

  • This project is "hard", since many have tried and always failed. The reason for this is the temptation to rewrite the current implementation all at once. See more details for the new plan, which avoids this pitfall...
  • More Details

Multisets

PEOPLE: dkohel, rlmiller, brian

Optimize/better document/generally improve graph theory library in Sage

PEOPLE: Robert Miller

  • Ticket #6085 contains a lot of work so that a graph created by Graph(implementation='c_graph') is just as functional as a Sage graph. I will be sporadically working on improving documentation and optimizing graphs all week, and anyone interested is welcome to join.

Cliquer SPKG for Sage

PEOPLE: Robert Miller, Nathann Cohen (remotely), brian, ncohen, rlmiller, bjarke

Take a look at the possibility of making GAP a dynamically loadable library

PEOPLE: Robert Miller, wstein

  • Robert Miller and William Stein succeeded in making a first very rough Cython module that links in GAP as a library and allows one to type commands to the interpreter. See http://trac.sagemath.org/sage_trac/ticket/6391/. This proof-of-concept that not only can GAP be linked as a library instead of used via pexpect, but that the result is likely to be *better* than libPARI and libSingular, in that one has full access to the GAP interpreter scripts and capabilities (just like embedding Python allows for that). Also, it turns out that so far there are no issues because of the GAP memory manager.

Python implementation of Ford-Fulkerson algorithm

PEOPLE: Robert Miller, brian, michael, tom

I plan on at least copying the Python implementation on wikipedia, since now we have nothing at all for max flow problems. Hopefully then someone who really cares about it will try to use it, realize it is slow, start improving it, etc. etc. etc.

Frobenius number and genus of numerical semigroups using toric Grobner bases

PEOPLE: Bjarke Hammersholt Roune, wouter

  • I plan to code Frobenius number (largest gap) and genus (number of gaps) functions for numerical semigrups using two related algorithms based on toric ideals. These algorithms can handle random numerical semigroups generated by numbers with thousands of digits, as long as there are not too many minimal generators. I'm happy to explain either algorithm if you want to help or are just curious. The steps needed are these:
    • Find the best way to compute toric Grobner bases in Sage (4ti2?)
    • Improve the integration of the library Frobby for monomial ideal computations.
    • Code the algorithms (should be easy at this point)
    Help is welcome, especially for writing a Cython interface to Frobby.

PolyBoRi and Singular

PEOPLE: Martin Albrecht, Burcin Erocal

Both PolyBoRi's (Gröbner bases in GF(2)[x_i]/<x_i^2+x_i>) and Singular's interface need some (engineering) work & documentation.

Elliptic curves isogenies -- greatly improve usability

PEOPLE: William Stein, dkohel, wouter, rlmiller, hamish

Dan Shumow wrote new code for computing isogenies between elliptic curves, but it is still very rough around the edges. Fix it up. See e.g., http://trac.sagemath.org/sage_trac/ticket/6384

Siegel Theta Functions

PEOPLE: guardia, holuk, drkohel, rlmiller, preston, schulze-pillot

See http://trac.sagemath.org/sage_trac/ticket/6371 for code that calculates classical theta functions, their derivatives, and some related functions I'm interested in, to arbitrary precision.

Multimodular reconstruction implementation

PEOPLE: clement, wstein, burcin, rlmiller, dloeffler, sebastian, brian

  • CRT in rings/arith.py is unsanely slow
  • MultimodularBasis class is broken

Fix / Unify code / optimize the hell out of it

Add the early termination technique for the reconstruction of multiple values Apply it to Burcin's computation of nullspace over Z[x]

PEOPLE: Burcin Erocal, Clément Pernet

Minpoly over small finite field

PEOPLE: clement, michael, brian, preston, emmanuel

#6296 Implement a certified computation of minpoly over a small finite field.

PEOPLE: Clément Pernet

Refactor symbolic functions

PEOPLE: burcin, wstein, tom

Merge the PrimitiveFunction and SFunction classes in sage.symbolic.function and make it easier to define new special functions with custom evaluation, printing, etc. methods. This also involves adding capability to represent symbolic sums and integrals in Sage as a first step to adding native Sage code to deal with these.

M4RI: improvement of PLUQ factorization

PEOPLE: Martin Albrecht, Clément Pernet

Change/improve the slow column swap in PLUQ factorization

  • We improved the PLUQ/LQUP base case by using up to four tables now instead of only two.
  • We improved the implementation for various column swaps and permutation matrix multiplications.
  • We moved from PLUQ to LQUP because it requires less column swaps.

lqup-vs-pluq.png}

m4ri-pluq-sparse-ish.png m4ri-dense-10.png m4ri-dense-20.png m4ri-dense-30.png

LinBox wrappers

Rewrite LinBox wrappers using C++ wrappers of Cython directly

PEOPLE: Burcin Erocal, Clément Pernet, rlmiller, malb

Ray class groups and Groessencharacters

Implement computations of the ray class group modulo an ideal of a number field, and the corresponding character group.

PEOPLE: David Loeffler, rlmiller, wstein, dkohel, haluk, michael

Pari 2.4 Upgrade

  • PEOPLE: wstein, dloeffler, hamish, michael

Sage notebook security

  • PEOPLE: Yoav, malb
  • We had some nice design discussions.

Texmacs --> Sage Worksheet converter

  • PEOPLE: Offray

SPD (simple python distro)

  • PEOPLE Offray

days16/projects (last edited 2009-06-27 14:41:17 by BurcinErocal)