1
0
mirror of https://github.com/kalmarek/PropertyT.jl.git synced 2024-12-11 23:06:32 +01:00
Commit Graph

803 Commits

Author SHA1 Message Date
3f11b2df9a
a bit on support for constraints 2024-02-06 13:30:06 +01:00
49b4c6f510
add symmetrized sos_problem_dual 2023-10-06 16:23:11 +02:00
9434f5999d
add functions to decompose vector in invariant basis 2023-10-06 16:18:36 +02:00
5f54ecb515
rework moment_matrix formulation 2023-10-06 14:16:49 +02:00
759d4c9ed7
update to PermutationGroups-0.4 2023-10-06 14:15:16 +02:00
7791fbdc42
ignore log 2023-10-06 14:11:59 +02:00
f4936dd50a
add/update scripts for SLNZ/SpNZ 2023-05-30 16:35:55 +02:00
cd7901b455
add G₂ to classify_root_system 2023-05-10 15:20:07 +02:00
902be093db
bump to 0.5.0 2023-05-09 12:39:34 +02:00
Marek Kaluba
0908c8ea1c
Merge pull request #11 from kalmarek/enh/G₂_here_we_come
G₂ - here we come!
2023-05-09 12:37:56 +02:00
d40a0fe117
fix error with printing of roots 2023-05-09 01:02:39 +02:00
34e19768d4
rm Manifest and update scripts 2023-05-08 17:49:32 +02:00
ed832d69fb
fix for solve_in_loop script 2023-04-13 01:00:22 +02:00
9afbccda15
drop zeros when reconstructing the solution 2023-04-11 10:32:24 +02:00
263af398eb
G₂_Adj: bring back solve_in_loop 2023-04-11 10:31:54 +02:00
56aed88416
use MKLDirect for G₂ 2023-04-06 16:16:27 +02:00
58f0ccb141
use IntervalMatrices
IntervalMatrices use Rump algorithm to matrix multiplication
This brings time to Qint'*Qint down to ~40s which is
5-8 × slower than Q'*Q (for size n=2^13).

The naive version is ~100 × slower than Q'*Q even for n = 2^10.
2023-04-06 13:17:52 +02:00
17274f895f
set eps_*_inf for COSMO solver 2023-04-06 11:40:49 +02:00
f0986982ce
reorganize Roots module 2023-04-06 11:39:54 +02:00
005ffc29cb
cosmo fails at high precision with zero condition 2023-04-05 17:50:12 +02:00
5b4a7f6804
reshuffle sos_sdps for clarity 2023-04-04 23:50:48 +02:00
150b5c2cba
skip the identity constraint if augmented 2023-04-04 23:48:59 +02:00
3f2be20152
make nzpairs(::ConstraintMatrix) type stable 2023-04-04 23:15:40 +02:00
132802feeb
use sparse matrices for invariant constraint 2023-04-04 23:14:56 +02:00
1a43a1b1be
in reconstruct: average the sum, not sum the averages! 2023-04-04 19:58:51 +02:00
f9f852439f
rewrite scripts for G2 2023-03-20 02:34:14 +01:00
15286c0c4a
add Manifest with a branch of SymbolicWedderburn 2023-03-20 01:42:41 +01:00
a14c6d2669
add G₂ script 2023-03-20 01:40:59 +01:00
92d4ba0c20
update script for SpN_Adj.jl 2023-03-20 01:40:40 +01:00
914b068070
update and reformat tests 2023-03-20 01:37:19 +01:00
fb3b51fd6e
fix sos_problem_dual 2023-03-20 01:37:19 +01:00
bacd170504
preserve trace when diagonalizing M_orb 2023-03-20 01:37:18 +01:00
a5f5a4ea35
lots of re-formatting 2023-03-20 01:37:18 +01:00
a1de0ecc85
improve __fast_recursive_dot a bit 2023-03-20 01:37:17 +01:00
b5fa1ac0ef
use Cartan matrix to classify root-subsystems 2023-03-20 01:37:16 +01:00
1fb324b49a
update constraints to StarAlgebras-0.2 2023-03-20 01:36:40 +01:00
4e43811ea3
fix a nasty bug with negatives in ConstraintMatrix 2023-03-20 01:36:39 +01:00
Marek Kaluba
0127d05594
Merge pull request #10 from kalmarek/mk/perf_tweaks
Various perf tweaks including moving to sparse matrices where possible
2022-12-02 14:21:24 +01:00
2cc9444667
small fixes 2022-11-17 11:22:17 +01:00
5349c21034
add script for SpN_Adj.jl 2022-11-17 02:50:48 +01:00
18f681b12b
add really quick tests to be run first
tests run really fast, but compilation time is killing it...
2022-11-16 00:16:01 +01:00
d1d46d13ef
use spzeros constructor compatible with julia-1.6 2022-11-16 00:14:06 +01:00
b92f12107f
update SymbolicWedderburn to 0.3.2 2022-11-15 19:36:39 +01:00
f053bffefe
rewrite reconstruct! with a better architecture 2022-11-15 19:36:39 +01:00
35c5110a37
move solve to a separate file 2022-11-15 18:51:43 +01:00
ca6a17acca
use ProgressMeter instead of poormans progressbar 2022-11-15 00:27:09 +01:00
fca7733280
remove old (unported) SOS_correctness tests 2022-11-14 19:51:38 +01:00
e6bd862e7a
small tweaks 2022-11-14 19:50:50 +01:00
703e69fc62
fix actions on SpNs 2022-11-14 19:50:27 +01:00
971e07b819
move to using sparse matrices in symmetrized sdp
dense are faster for small sizes only
2022-11-14 19:50:09 +01:00