]>
gitweb.michael.orlitzky.com - sage.d.git/log
Michael Orlitzky [Wed, 4 Nov 2020 22:28:56 +0000 (17:28 -0500)]
eja: speed up _charpoly_coefficients when rank is cached.
Michael Orlitzky [Wed, 4 Nov 2020 15:14:52 +0000 (10:14 -0500)]
eja: fix the determinant in trivial algebras.
Michael Orlitzky [Wed, 4 Nov 2020 15:14:34 +0000 (10:14 -0500)]
eja: remove more vestigial charpoly stuff.
Michael Orlitzky [Wed, 4 Nov 2020 14:52:52 +0000 (09:52 -0500)]
makefile: don't allow tests to time out.
Michael Orlitzky [Wed, 4 Nov 2020 13:31:51 +0000 (08:31 -0500)]
eja: simplify and unify the charpoly/rank stuff.
Michael Orlitzky [Wed, 4 Nov 2020 13:01:11 +0000 (08:01 -0500)]
eja: replace rank computation with length of charpoly coefficients.
If we cache the solution to the charpoly system instead, then the
result is useful in more than one place (for the charpoly itself,
in particular).
Michael Orlitzky [Tue, 3 Nov 2020 19:43:30 +0000 (14:43 -0500)]
eja: get rid of the KnownRankEJA class.
Not needed now that I know how to compute rank.
Michael Orlitzky [Tue, 3 Nov 2020 13:34:36 +0000 (08:34 -0500)]
eja: reimplement random_eja() without KnownRankEJA.
Michael Orlitzky [Tue, 3 Nov 2020 13:15:48 +0000 (08:15 -0500)]
eja: fix tests and pre-cache ranks.
Michael Orlitzky [Tue, 3 Nov 2020 12:12:14 +0000 (07:12 -0500)]
eja: replace the rank accessor with the full computation.
Michael Orlitzky [Sun, 1 Nov 2020 00:19:42 +0000 (20:19 -0400)]
eja: use now somewhat-proven and fast rank computation.
Michael Orlitzky [Fri, 30 Oct 2020 19:32:38 +0000 (15:32 -0400)]
eja: add TODON'T use Cramer's rule.
Michael Orlitzky [Wed, 28 Oct 2020 15:04:52 +0000 (11:04 -0400)]
eja: add two more experimental rank computations.
Michael Orlitzky [Fri, 23 Oct 2020 20:16:58 +0000 (16:16 -0400)]
eja: add voodoo _rank_computation() method for EJAs.
Michael Orlitzky [Mon, 19 Oct 2020 14:30:10 +0000 (10:30 -0400)]
eja: drop the a_jordan_frame() method on EJAs.
I understand this better now, and don't expect this method to work
in general. It works for associative algebras, but we already know
how to find a Jordan frame in associative algebras: the spectral
decomposition, which works on the associative algebra alg({x}).
Michael Orlitzky [Mon, 19 Oct 2020 14:25:20 +0000 (10:25 -0400)]
Revert "eja: start experiment full_spectral_decomposition() method."
This reverts commit
6d32150d3f7c175473a1d590facce1ea7c5ca77a .
I don't think this will ever work reliably.
Michael Orlitzky [Sun, 18 Oct 2020 14:56:38 +0000 (10:56 -0400)]
eja: start experiment full_spectral_decomposition() method.
Michael Orlitzky [Sun, 18 Oct 2020 05:24:18 +0000 (01:24 -0400)]
eja: fix sub-subalgebra element construction.
Michael Orlitzky [Sun, 18 Oct 2020 05:23:58 +0000 (01:23 -0400)]
mjo/eja/eja_element_subalgebra.py: switch two tests back to QQ.
Michael Orlitzky [Sun, 18 Oct 2020 05:23:09 +0000 (01:23 -0400)]
mjo/eja/eja_element.py: test spectral decomposition in subalgebras.
Michael Orlitzky [Sun, 18 Oct 2020 05:22:51 +0000 (01:22 -0400)]
mjo/eja/eja_element.py: remove a dead line.
Michael Orlitzky [Sun, 11 Oct 2020 21:24:17 +0000 (17:24 -0400)]
eja: make AA the default field because everything cool requires it.
Michael Orlitzky [Sun, 11 Oct 2020 19:22:49 +0000 (15:22 -0400)]
mjo/eja/eja_algebra.py: fix unimportant left/right eigenspace terminology.
Michael Orlitzky [Sun, 11 Oct 2020 13:10:38 +0000 (09:10 -0400)]
eja: ignore errors more specifically in a_jordan_frame().
Michael Orlitzky [Sun, 11 Oct 2020 12:35:18 +0000 (08:35 -0400)]
eja: rename orthogonal_idempotents() to a_jordan_frame().
Michael Orlitzky [Sun, 11 Oct 2020 01:47:50 +0000 (21:47 -0400)]
mjo/eja/eja_algebra.py: add WIP orthogonal_idempotents() method.
Michael Orlitzky [Wed, 7 Oct 2020 21:21:59 +0000 (17:21 -0400)]
mjo/ldlt.py: goodbye, use https://trac.sagemath.org/ticket/10332 instead.
Michael Orlitzky [Mon, 5 Oct 2020 19:23:00 +0000 (15:23 -0400)]
mjo/ldlt.py: add fast block-LDLT-based is_positive_semidefinite().
Michael Orlitzky [Mon, 5 Oct 2020 17:55:19 +0000 (13:55 -0400)]
mjo/ldlt.py: move implementation to SageMath proper.
The actual block-LDLT code is now on the u/mjo/ticket/10332 branch.
Michael Orlitzky [Sun, 4 Oct 2020 16:37:08 +0000 (12:37 -0400)]
mjo/ldlt.py: refactor into user-(un)friendly portions.
Michael Orlitzky [Sun, 4 Oct 2020 16:11:25 +0000 (12:11 -0400)]
mjo/ldlt.py: delete naive implementations; fix tests.
Michael Orlitzky [Sun, 4 Oct 2020 16:07:00 +0000 (12:07 -0400)]
mjo/ldlt.py: add examples and tests for block_ldlt().
Michael Orlitzky [Sun, 4 Oct 2020 13:40:50 +0000 (09:40 -0400)]
mjo/ldlt.py: make block_ldlt() work on Hermitian matrices, too.
Michael Orlitzky [Sun, 4 Oct 2020 12:53:48 +0000 (08:53 -0400)]
mjo/ldlt.py: switch some row/column indexing around.
Michael Orlitzky [Sun, 4 Oct 2020 12:27:37 +0000 (08:27 -0400)]
mjo/ldlt.py: rule #1 is never compute the inverse of a matrix.
Michael Orlitzky [Sun, 4 Oct 2020 02:48:04 +0000 (22:48 -0400)]
mjo/ldlt.py: get fast block_ldlt() working.
Michael Orlitzky [Sun, 4 Oct 2020 02:11:19 +0000 (22:11 -0400)]
mjo/ldlt.py: fix a dimension in block_ldlt_naive().
Michael Orlitzky [Fri, 2 Oct 2020 22:41:30 +0000 (18:41 -0400)]
mjo/ldlt.py: fix two bugs in the imperative block_ldlt() function.
Michael Orlitzky [Fri, 2 Oct 2020 22:16:06 +0000 (18:16 -0400)]
mjo/ldlt.py: begin fast block-LDLT implementation.
Michael Orlitzky [Fri, 2 Oct 2020 12:25:13 +0000 (08:25 -0400)]
mjo/ldlt.py: refactor the one-by-one block-LDLT pivot.
Michael Orlitzky [Fri, 2 Oct 2020 11:37:35 +0000 (07:37 -0400)]
mjo/ldlt.py: start a naive Bunch-Kaufman block LDLT.
Michael Orlitzky [Mon, 28 Sep 2020 19:27:37 +0000 (15:27 -0400)]
mjo/ldlt.py: various efficiency improvements to ldlt_fast().
Michael Orlitzky [Mon, 28 Sep 2020 18:41:12 +0000 (14:41 -0400)]
mjo/ldlt.py: avoid some intermediate variables in ldlt_fast().
Michael Orlitzky [Mon, 28 Sep 2020 18:28:22 +0000 (14:28 -0400)]
mjo/ldlt.py: add a vaguely correct non-recursive ldlt_fast().
Michael Orlitzky [Mon, 28 Sep 2020 15:08:37 +0000 (11:08 -0400)]
mjo/ldlt.py: actually permute with the permutation matrix.
Michael Orlitzky [Mon, 28 Sep 2020 01:40:13 +0000 (21:40 -0400)]
mjo/ldlt.py: add naive, pivoted LDLT matrix factorization.
Michael Orlitzky [Sun, 13 Sep 2020 21:05:21 +0000 (17:05 -0400)]
makefile: delete mjo/__pycache__ with "make clean".
Michael Orlitzky [Sun, 13 Sep 2020 21:03:59 +0000 (17:03 -0400)]
mjo: rename matrix_vector.py to basis_repr.py.
Michael Orlitzky [Sun, 13 Sep 2020 20:40:54 +0000 (16:40 -0400)]
mjo/matrix_vector: rename basis_representation() to basis_repr().
Michael Orlitzky [Sun, 13 Sep 2020 20:33:06 +0000 (16:33 -0400)]
mjo/matrix_vector: replace matrix_of_transformation with basis_repr_of_operator.
Michael Orlitzky [Sun, 13 Sep 2020 20:32:43 +0000 (16:32 -0400)]
mjo/matrix_vector: fix phi_inv implementation in basis_representation().
Michael Orlitzky [Sun, 13 Sep 2020 16:51:45 +0000 (12:51 -0400)]
mjo/matrix_vector: replace isomorphism with basis_representation().
Michael Orlitzky [Fri, 28 Aug 2020 17:59:42 +0000 (13:59 -0400)]
init.sage: use readline-like tab completion.
Michael Orlitzky [Fri, 28 Aug 2020 17:53:25 +0000 (13:53 -0400)]
init.sage: let the default LightBg colors apply.
I use a light terminal background now.
Michael Orlitzky [Tue, 9 Jun 2020 11:56:07 +0000 (07:56 -0400)]
eja: implement JordanSpinEJA with BilinearFormEJA.
Michael Orlitzky [Tue, 9 Jun 2020 11:41:47 +0000 (07:41 -0400)]
eja: make a BilinearFormEJA test do what it says.
Michael Orlitzky [Tue, 9 Jun 2020 10:25:01 +0000 (06:25 -0400)]
eja: re-disable base ring to algebra coercion.
Michael Orlitzky [Mon, 8 Jun 2020 23:16:52 +0000 (19:16 -0400)]
eja: new half-baked BilinearFormEJA class.
Michael Orlitzky [Mon, 8 Jun 2020 23:16:29 +0000 (19:16 -0400)]
eja: fix the docs for the spin algebra multiplication.
Michael Orlitzky [Thu, 2 Apr 2020 12:47:14 +0000 (08:47 -0400)]
eja: add another TODO.
Michael Orlitzky [Sat, 1 Feb 2020 01:26:21 +0000 (20:26 -0500)]
mjo/cone: rawify some docstrings.
Michael Orlitzky [Sat, 1 Feb 2020 01:25:39 +0000 (20:25 -0500)]
init.sage: add an alternate sitedir name.
Michael Orlitzky [Mon, 6 Jan 2020 13:31:28 +0000 (08:31 -0500)]
mjo/eja/all.py: fix the name of an import.
Michael Orlitzky [Fri, 27 Dec 2019 15:09:55 +0000 (10:09 -0500)]
eja: rename RealCartesianProductEJA -> HadamardEJA.
Michael Orlitzky [Mon, 9 Dec 2019 02:23:33 +0000 (21:23 -0500)]
mjo/polynomial.py: add hacks for SageMath Trac #28855.
Michael Orlitzky [Mon, 9 Dec 2019 02:01:24 +0000 (21:01 -0500)]
mjo/symbol_sequence.py: fix another backslash escaping issue.
Michael Orlitzky [Sun, 8 Dec 2019 00:01:13 +0000 (19:01 -0500)]
eja: listify map/zip results for python-3.x compatibility.
Michael Orlitzky [Sat, 7 Dec 2019 23:31:53 +0000 (18:31 -0500)]
mjo/cone/doubly_nonnegative.py: don't check rank validity of None.
Michael Orlitzky [Sat, 7 Dec 2019 23:22:48 +0000 (18:22 -0500)]
mjo/interpolation.py: wrap map() calls in list() for python-3.x.
There's probably a better way to implement these functions for
python-3.x, but this fixes the tests for now.
Michael Orlitzky [Sat, 7 Dec 2019 23:16:35 +0000 (18:16 -0500)]
cone/doubly_nonnegative.py: make two docstrings raw.
Michael Orlitzky [Sat, 7 Dec 2019 23:15:29 +0000 (18:15 -0500)]
mjo: replace xrange() with range() for python-3.x compatibility.
Michael Orlitzky [Sat, 7 Dec 2019 23:09:48 +0000 (18:09 -0500)]
eja: use zip() instead of izip(), which doesn't exist in python-3.x.
Michael Orlitzky [Sat, 7 Dec 2019 23:09:26 +0000 (18:09 -0500)]
symbol_sequence.py: make a docstring raw.
Michael Orlitzky [Fri, 15 Nov 2019 17:44:03 +0000 (12:44 -0500)]
eja: rename is_minimal_idempotent() to is_primitive_idempotent().
The term "primitive" is more common in optimization, anyway. Also the
word "minimal" is a bit overloaded. Oh, and I added two more tests to
preclude zero from being primitive.
Michael Orlitzky [Fri, 15 Nov 2019 13:15:53 +0000 (08:15 -0500)]
eja: use peirce_decomposition() method to get the Peirce decomposition.
Michael Orlitzky [Mon, 11 Nov 2019 16:26:21 +0000 (11:26 -0500)]
eja: add another TODO before I forget.
Michael Orlitzky [Sun, 10 Nov 2019 18:39:19 +0000 (13:39 -0500)]
eja: fix the Peirce decomposition and add docs/tests for it.
Michael Orlitzky [Sun, 10 Nov 2019 15:47:42 +0000 (10:47 -0500)]
eja: allow subalgebra_idempotent() to work in trivial EJAs.
Every element is nilpotent in a trivial EJA, but we also know the only
idempotent element in such an algebra: its sole occupant. So return it.
Michael Orlitzky [Sun, 10 Nov 2019 14:57:58 +0000 (09:57 -0500)]
eja: add another non-element subalgebra example.
Michael Orlitzky [Sun, 10 Nov 2019 14:40:25 +0000 (09:40 -0500)]
eja: make two subalgebra tests more general.
All of the subalgebra tests are element-subalgebra tests, because I
copied them from the element-subalgebra class. I'll slowly make them
more robust.
Michael Orlitzky [Sun, 10 Nov 2019 14:20:59 +0000 (09:20 -0500)]
eja: remove redundant subalgebra stuff out of element subalgebras.
These were simply remnants of a copy/paste refactoring.
Michael Orlitzky [Sun, 10 Nov 2019 14:10:43 +0000 (09:10 -0500)]
eja: don't wait for a non-nilpotent element in trivial algebras.
Michael Orlitzky [Sun, 10 Nov 2019 13:52:44 +0000 (08:52 -0500)]
eja: remove broken one() implementation in subalgebras.
The one() implementetion for the algebra generated by an element is
not, in general, correct for a subalgebra. Subalgebras can have
different identities than superalgebras -- take the Peirce
decomposition, for example.
Michael Orlitzky [Sun, 10 Nov 2019 01:56:51 +0000 (20:56 -0500)]
eja: remove a TODO that is done in theory.
Michael Orlitzky [Sun, 10 Nov 2019 01:56:18 +0000 (20:56 -0500)]
eja: start experimenting with single-element Peirce decomposition.
Michael Orlitzky [Sun, 10 Nov 2019 00:42:25 +0000 (19:42 -0500)]
eja: refactor the element subalgebra stuff into generic subalgebra.
A messy job, but I got the tests passing. They certainly need some
cleanup, and we should test non-element subalgebras too.
Michael Orlitzky [Sat, 9 Nov 2019 22:57:42 +0000 (17:57 -0500)]
eja: move eja_subalgebra to eja_element_subalgebra.
The plan is to cut/paste out all the stuff that works for a generic subalgebra.
Michael Orlitzky [Sat, 9 Nov 2019 19:51:23 +0000 (14:51 -0500)]
eja: add a priority TODO.
Michael Orlitzky [Tue, 29 Oct 2019 01:57:11 +0000 (21:57 -0400)]
eja: drop custom gram_schmidt() routine that isn't noticeably faster.
Using Gram-Schmidt in SageMath is a bit of a pain, since you have to
put your vector space elements into a matrix, which means you have to
convert them to vectors first. And then you wind up transposing the
matrix twice, and then you have to convert everything back to vectors
and then back to algebra elements, dropping the zero vectors/elements
in the process. Sounds slow, right?
Well, testing shows that jumping through all those hoops and using the
default implementation isn't much slower than the custom routine I
wrote. So, for now, let's delete it.
Beware that this breaks the construction of subalgebras when the field
is not exact because SageMath won't orthonormalize over an inexact
field.
Michael Orlitzky [Fri, 18 Oct 2019 23:13:16 +0000 (19:13 -0400)]
eja: add is_minimal_idempotent() for elements.
Michael Orlitzky [Tue, 15 Oct 2019 12:29:25 +0000 (08:29 -0400)]
eja: update the TODO.
Michael Orlitzky [Tue, 15 Oct 2019 12:28:08 +0000 (08:28 -0400)]
eja: add a new TrivialEJA class and some tests for it.
Michael Orlitzky [Tue, 15 Oct 2019 12:27:44 +0000 (08:27 -0400)]
eja: define the trace of an element in a trivial algebra.
Michael Orlitzky [Tue, 15 Oct 2019 12:23:10 +0000 (08:23 -0400)]
eja: simplify the algebra charpoly method a bit.
Michael Orlitzky [Tue, 15 Oct 2019 12:15:43 +0000 (08:15 -0400)]
eja: fix a rank test to work on trivial algebras.
Michael Orlitzky [Tue, 15 Oct 2019 12:13:41 +0000 (08:13 -0400)]
eja: make some element tests work in trivial algebras.
Michael Orlitzky [Tue, 15 Oct 2019 12:10:46 +0000 (08:10 -0400)]
eja: make two operator tests work in trivial algebras.
Michael Orlitzky [Mon, 14 Oct 2019 13:43:38 +0000 (09:43 -0400)]
eja: add the unique spectral_decomposition() for elements.
Michael Orlitzky [Mon, 14 Oct 2019 13:43:17 +0000 (09:43 -0400)]
eja: define subalgebra_generated_by() to contain the identity.
Michael Orlitzky [Mon, 14 Oct 2019 13:26:26 +0000 (09:26 -0400)]
eja: remove obsolete workaround for SageMath trac ticket #28327.