Commit Graph

583 Commits

Author SHA1 Message Date
kalmarek 523003cc2a move fillfrominternal! to SDPs.jl 2018-01-01 23:59:31 +01:00
kalmarek d5a4f4957a rework filename logic 2018-01-01 23:59:00 +01:00
kalmarek f48e8ba8d7 return the global logger for now 2018-01-01 23:55:07 +01:00
kalmarek 7ac194f98a fix logtime macro 2018-01-01 23:54:36 +01:00
kalmarek 3066046fdb move checking logic to check_distance_to_cone function 2018-01-01 23:50:36 +01:00
kalmarek b2f4360ef4 distence_to_cone(...) where {T<:...} does all the logging 2018-01-01 23:49:27 +01:00
kalmarek 5918e20569 replace distance_to_cone by distances_to_cone without logging 2018-01-01 23:48:17 +01:00
kalmarek 923d892da0 fix indentation, again 2018-01-01 23:45:49 +01:00
kalmarek 47f6d3637e fix indentation to 4 spaces 2018-01-01 14:06:33 +01:00
kalmarek fc54803b58 fix Memento deprecations; make clear when logger is global 2018-01-01 13:13:44 +01:00
kalmarek 7054fcffe2 remove old loging 2017-12-02 23:21:00 +01:00
kalmarek 29f347a228 only check the solution when not warmstarting 2017-12-01 18:28:48 +01:00
kalmarek 5e4b2203c3 Nemo.perm is a function in julia v0.6.0 2017-12-01 18:26:55 +01:00
kalmarek 53d413633b add warmstart to Settings 2017-12-01 18:26:08 +01:00
kalmarek c849ead0ac look for warmstart in the solutions dir 2017-12-01 17:27:18 +01:00
kalmarek e94662e615 fill the JuMP model variables from its internal model 2017-12-01 17:06:20 +01:00
kalmarek 659e5e03a1 grab warmstart, P, lambda from the solved model 2017-12-01 17:05:35 +01:00
kalmarek bbcaec5bae build JuMP.Model first 2017-12-01 17:04:53 +01:00
kalmarek 56adccdfd1 using MathProgBase 2017-12-01 17:03:24 +01:00
kalmarek e53486f394 since we warmstart internal model, optimize! it
JuMP.solve discards the internal one
2017-12-01 17:02:45 +01:00
kalmarek 10808583e3 Merge remote-tracking branch 'origin/enh/warmstart' into enh/warmstart-0.6
# Conflicts:
#	src/OrbitDecomposition.jl
2017-12-01 14:21:10 +01:00
kalmarek 02cdee0708 remove old check for the existence of solver.log 2017-12-01 14:13:19 +01:00
kalmarek da6a2df814 save primal, dual solutions and slack for future warmstart 2017-12-01 14:12:52 +01:00
kalmarek 7ee7a30829 fix rankOne_projections for the new type of GroupRingElem 2017-11-17 15:29:09 +01:00
kalmarek bdadac33f4 import Generic.perm 2017-11-17 15:28:11 +01:00
kalmarek 0d20df8207 add dim function for Characters 2017-11-16 20:01:12 +01:00
kalmarek 498a6700ec Merge branch 'master' into enh/julia-v0.6 2017-11-15 20:28:57 +01:00
kalmarek 39bef16aa7 filter Upis and dims at LOADTIME for only those which are non-zeros 2017-11-13 10:58:22 +01:00
kalmarek 04a13616c6 Allow different AbstractMathProgSolvers in Settings 2017-11-13 10:57:02 +01:00
kalmarek a35122f8bb cosmetic 2017-11-08 15:40:38 +01:00
kalmarek 38ae1e0656 remove minimalprojections in favour of rankOne_projections 2017-11-08 15:40:22 +01:00
kalmarek e746aaeeff rankOne_projections MUST be computed over exact field (e.g. Rational{Int}) 2017-11-08 15:39:51 +01:00
kalmarek d7652c4e02 make AbstractCharacter just an abstract type 2017-11-08 15:38:51 +01:00
kalmarek 6f0087b8c7 indentation 2017-11-08 11:10:30 +01:00
kalmarek ea6a6722be init model on number and size of SD blocks, not on Uπs 2017-11-08 10:02:38 +01:00
kalmarek bcbf8a03c8 store the dense Uπs, sparsify at read-time 2017-11-08 10:01:45 +01:00
kalmarek fe1e9ee37c let sparsify actually return sparse matrix 2017-11-08 10:00:34 +01:00
kalmarek 2aa82db1ac specify exact types for OrbitData 2017-11-08 09:58:37 +01:00
kalmarek c1dc3850dd rename AutS -> autS 2017-11-08 09:57:51 +01:00
kalmarek 4ba2f54bb4 improve slightly orthSVD 2017-11-08 09:50:58 +01:00
kalmarek 3c51a463dc rename variables 2017-11-08 09:38:56 +01:00
kalmarek ff14b33199 AutS_mreps -> reps 2017-11-08 09:37:39 +01:00
kalmarek 95e19e6f9d replace stray constraints_from_pm() -> constraints() 2017-11-08 09:36:34 +01:00
kalmarek 80deaebb9f remove matrix_reprs - we create them through perm_reprs 2017-11-08 09:35:26 +01:00
kalmarek 72261afdeb use preps_reprs instead of matrix_reprs
and save them in preps.jld
2017-11-08 09:33:00 +01:00
kalmarek 2d02c98947 faster & simpler Cstar_repr 2017-11-08 09:29:12 +01:00
kalmarek e636c97738 unified marix_reps and perm_reps 2017-11-08 09:28:04 +01:00
kalmarek 0c9fb40e65 constraints_from_pm -> constraints
returns a vector of constraints, i.e. vectors of Tuple{Int, Int}
2017-11-08 09:25:40 +01:00
kalmarek 31715701cc de-thread Projections
to avoid UndefRefError: access to undefined reference in
rankOne_projections(::WreathProduct)
2017-11-07 09:44:10 +01:00
kalmarek 12c5c47ca8 add Nemo.isone(::GroupElem) 2017-11-06 11:31:15 +01:00
kalmarek 84813fedc4 indentation 2017-11-05 20:55:53 +01:00
kalmarek 093cced764 Merge branch 'enh/logtime' 2017-11-05 20:55:05 +01:00
kalmarek bc8c28170e missing @logtime in orbit-check_property_T 2017-11-05 20:54:33 +01:00
kalmarek cda741cc0c Merge branch 'enh/check_solution' 2017-11-05 19:30:58 +01:00
kalmarek 36e5979774 @logtime Q = sqrtm(P) 2017-11-05 19:30:11 +01:00
kalmarek f7f9ceef07 pass only the i-th column in @parallel mode
this cuts on memory per worker drastically
2017-11-05 19:28:51 +01:00
kalmarek c3ea80e6a9 replace remaining @time with '@logtime logger' 2017-11-05 16:38:51 +01:00
kalmarek c0dc6937a5 cut on allocations in compute_SOS 2017-11-03 16:54:41 +01:00
kalmarek fba2464a80 Specify concrete Interval types in augIdproj 2017-11-03 16:42:53 +01:00
kalmarek a000290074 remove rationalization functions 2017-11-03 16:41:12 +01:00
kalmarek 79d5c7f933 rename parameter S -> T 2017-11-02 13:48:20 +01:00
kalmarek 1c67cdbcea remove timed_msg 2017-11-02 13:45:27 +01:00
kalmarek 3eab5b721a replace @timed by @logtime 2017-11-02 13:45:12 +01:00
kalmarek 7fa49756ba @logtime macro and time_string function 2017-11-02 13:44:08 +01:00
kalmarek 5bb311141d splaplacian doesn't have Id argument 2017-10-27 18:36:59 +02:00
kalmarek ca18f75efd clean the main check_property_T in both cases 2017-10-27 18:36:32 +02:00
kalmarek f206a13980 update function names 2017-10-27 18:35:50 +02:00
kalmarek 098a7ce948 check_distance_to... -> distance_to_...
and a lot of old code removal
2017-10-27 18:34:07 +02:00
kalmarek 4ae2c617cb Higher level compute_SOS 2017-10-27 18:33:18 +02:00
kalmarek 96ce02852f replace rationalise&project with interval&project 2017-10-27 18:32:58 +02:00
kalmarek 0225f38066 just two simple distance_to_cone functions
for Numbers and for Intervals
2017-10-27 18:30:59 +02:00
kalmarek 8defd71b59 new low-level compute_SOS function 2017-10-27 18:29:19 +02:00
kalmarek d8b2f0ab30 increase sigfigs to 12; remove explicit precision setting
(it's 53 anyway);
2017-10-27 18:25:13 +02:00
kalmarek b96b39e910 Revert "make initial P nullable"
This reverts commit 32c020fd88.
2017-10-27 18:20:28 +02:00
kalmarek 01e6625571 Revert "Allow for AbstractMathProfSolver in Settings"
This reverts commit 6cbf5e6944.
2017-10-27 18:19:53 +02:00
kalmarek 94c9a75a21 we don't need to pass id element -- it could be derived from S 2017-10-27 18:18:34 +02:00
kalmarek 32c020fd88 make initial P nullable 2017-10-27 18:18:34 +02:00
kalmarek 46a9150266 solver is not defined here 2017-10-27 18:18:34 +02:00
kalmarek 6cbf5e6944 Allow for AbstractMathProfSolver in Settings 2017-10-27 18:18:34 +02:00
kalmarek bc2a827342 use prepath(sett)/fullpath(sett) instead of sett.name 2017-10-09 19:49:49 +02:00
kalmarek b3f1961fbf use prepath/fullpath(::Settings) in OrbitData
and in create_SDP_problem generation
2017-10-09 19:49:01 +02:00
kalmarek 36a1151a73 add functions to produce prepath, fullpath out of Settings 2017-10-09 19:45:28 +02:00
kalmarek 6070bbf709 rename name -> prefix in filenames functions 2017-10-09 19:44:54 +02:00
kalmarek 4c777073a7 Merge branch 'master' of https://git.wmi.amu.edu.pl/kalmar/PropertyT.jl 2017-10-09 18:08:57 +02:00
kalmarek d7bc94b64c @parallel compute_SOS 2017-10-09 18:08:22 +02:00
kalmarek 344e11a974 overwrite Q to allow gc 2017-10-09 18:06:29 +02:00
kalmarek 0fabd21e53 much faster Cstar_repr 2017-10-03 14:46:05 +02:00
kalmarek dbca3945a8 call flush_cstdio() AFTER redirect_stdout
to fix SCS log buffering issue
2017-09-19 17:38:16 +02:00
kalmarek 154a66939f call flush_cstdio() AFTER redirect_stdout
to fix SCS log buffering issue
2017-09-19 17:37:35 +02:00
kalmarek 4e3bcbba6e update to the new Nemo.Generic module 2017-09-15 18:19:20 +02:00
kalmarek 5ac2876e31 update to julia v0.6.0 2017-09-15 18:16:31 +02:00
kalmarek 0693455656 Merge branch 'master' of https://git.wmi.amu.edu.pl/kalmar/PropertyT.jl 2017-09-10 13:39:52 +02:00
kalmarek b46ed5ae96 Revert "ccall((:printf, "libc"), ...) produces segfault for large examples??"
This reverts commit 69a3ab8d87.
This should be commented on machine-basis
2017-09-09 20:47:12 +02:00
kalmarek d8f04957b6 replace isfile by exists
which returns true also for symlinks
2017-09-06 16:02:46 +02:00
kalmarek a61cff986e verbose&check for possible rank reduction in Uπs 2017-08-27 21:25:12 +02:00
kalmarek 3b723e7f0e Add logging when verbose=true 2017-08-27 21:23:42 +02:00
kalmarek 6df4c790c0 Sparsification of a dense matrix is not required to be rank preserving
This can bite!
2017-08-27 21:23:14 +02:00
kalmarek 972697e2ba cosmetic 2017-08-27 20:24:18 +02:00
kalmarek be1251f274 Merge branch 'enh/logging' 2017-08-27 19:57:16 +02:00
kalmarek b781dae356 @logtime(logger, ex) macro 2017-08-27 19:56:35 +02:00
kalmarek 69a3ab8d87 ccall((:printf, "libc"), ...) produces segfault for large examples?? 2017-08-27 19:13:20 +02:00
kalmarek f543d847ea Save memory: ΔandSDPconstraints takes a lot of it 2017-08-27 19:12:21 +02:00
kalmarek 830ef95181 Merge branch 'enh/aug_correction' 2017-08-27 19:11:11 +02:00
kalmarek 757f89e54b no need to limit precision in rationalise_and_project
since we project directly to intervals the computational load is negligible
2017-08-27 19:09:56 +02:00
kalmarek 2724612ea3 better logging in Orbit-wise.jl 2017-08-27 19:06:17 +02:00
kalmarek 0c40e16e9e fix double assignment of constraints 2017-08-27 19:01:32 +02:00
kalmarek 8a6707eddd rework addconstraints! with progress printing 2017-08-27 19:00:49 +02:00
kalmarek db55927729 sparsify! in transform and in constrLHS 2017-08-27 18:59:16 +02:00
kalmarek 7dbd11bebf eps-sparsify for both sparse and dense matrices 2017-08-27 18:58:13 +02:00
kalmarek 151a6ff6da better density for dense matrices 2017-08-27 18:57:27 +02:00
kalmarek 7fd18e9e57 logging 2017-08-27 18:50:56 +02:00
kalmarek 56d944e8af correct columns of Q directly to intervals
hopefully containing 0
2017-08-27 18:46:31 +02:00
kalmarek 9840268250 use threaded version of matrix reps 2017-08-27 18:37:44 +02:00
kalmarek f3d813cb67 reworked reconstruct_sol with axpy! 2017-08-27 18:37:09 +02:00
kalmarek 33dc278762 threaded matrix_reps and perm_reps 2017-08-27 18:35:35 +02:00
kalmarek 6061f731b3 threaded generation of matrix reps 2017-08-27 18:32:50 +02:00
kalmarek 0cc0ad94a6 threaded orbit_decomposition 2017-08-27 18:32:19 +02:00
kalmarek 9328fa9187 single-threaded compute_SOS with @parallel option 2017-08-27 18:30:49 +02:00
kalmarek 9a62cc2444 threaded minimalprojections 2017-08-27 18:29:38 +02:00
kalmar 94074750cf limit the accuracy of rationalization to sensible value
to avoid running out of memory!
2017-08-04 21:27:10 +02:00
kalmar 8f88035bb7 in-place projection to augmentation ideal 2017-08-04 20:50:06 +02:00
kalmar c4dcaf25fe @parallelising computation of SOS 2017-08-04 20:47:38 +02:00
kalmar 7d42f28b29 put rationalization and projection to augmentation in a separate function
to have gc collect unused matrices of Rationals after this is finished
2017-08-04 20:36:25 +02:00
kalmar 845d7733a4 indentation 2017-08-04 20:33:21 +02:00
kalmar 9ed8679c95 simplify groupring_square 2017-08-04 20:33:02 +02:00
kalmar b97a0e30dd threaded matrix_reps 2017-08-04 18:25:10 +02:00
kalmar 666fac1775 Threaded version of projection to the augmentation ideal 2017-08-04 15:30:12 +02:00
kalmar 47a41ac9fe Threaded version of reconstruct_sol 2017-08-03 11:35:34 +02:00
kalmar a460b71db9 redirect stderr to full_log 2017-08-03 11:34:38 +02:00
kalmar a6eed4cd11 add @time to reconstruct_sol 2017-08-01 10:30:33 +02:00
kalmar 1f508ea85c replace fixed rankOne_projections with automatically generated ones 2017-08-01 10:29:38 +02:00
kalmar c86f46666a Cstar_reprs are usually sparse 2017-07-31 12:13:45 +02:00
kalmar cda9538653 fix Cstar_repr: explicit Float64 conversion 2017-07-31 12:13:14 +02:00
kalmar 15140d4e39 use sparsified U\pi's by default 2017-07-28 12:45:04 +02:00
kalmar f6674337ed sparsify only once during the reconstruction 2017-07-28 12:44:25 +02:00
kalmar 706097931a fix: central_projection asks for AbstractCharacter 2017-07-28 12:25:14 +02:00
kalmar 4bb52bd34b my Characters return Julia's Ints 2017-07-27 22:06:32 +02:00
kalmar 93ba763402 Allow evaluation of Characters on GroupRingElems 2017-07-27 22:06:02 +02:00
kalmar 536542c3a6 Introduce {Perm,DirectProd}Character <: AbstractCharacter <: Function 2017-07-27 22:05:15 +02:00
kalmar b89598c67c slight(?!) optimisation to Cstar_repr 2017-07-27 22:02:07 +02:00
kalmar 0a3b5b227d accommodate latest changes in Nemo characters 2017-07-27 20:36:22 +02:00
kalmar 9d3eec6812 fix: mixed-up projections 2017-07-27 20:35:39 +02:00
kalmar 582b7bc3c7 avoid transposing Us for every constraint 2017-07-26 18:02:21 +02:00
kalmar 5332d806cc Merge branch 'enh/sparsification' 2017-07-26 12:59:09 +02:00
kalmar 335a2aa216 change rounding mode to :tight 2017-07-26 12:58:39 +02:00
kalmar 66f860bac1 sparsify both on the forward and backward transform 2017-07-26 12:15:31 +02:00
kalmar 3ff649683a FIXXX: clamp to 0 only those eps-close to 0, not < eps 2017-07-26 12:12:58 +02:00
kalmar 33b45dd518 merge two sparsify functions 2017-07-26 10:29:11 +02:00
kalmar da32893d64 move setup-logging to the top of the file 2017-07-21 17:12:27 +02:00
kalmar a82eb13ef5 use views and broadcasting 2017-07-21 17:11:58 +02:00
kalmar 95bc1eaff5 make EOI a one-liner 2017-07-21 17:11:25 +02:00
kalmar e74b4c49ef use the new definition of base-less mul! in GroupRings 2017-07-21 17:10:52 +02:00
kalmar b8a40c42c8 indentation 2017-07-17 15:40:54 +02:00
kalmar 5d6074c368 rework ranOne_projections(::WreathProduct...) using emb functions
Still messy
2017-07-17 12:28:17 +02:00
kalmar 6ecd63197b account for changes in order of characters 2017-07-17 12:27:28 +02:00
kalmar 4e48917a38 characters of symmetric group are now implemented in Nemo 2017-07-17 12:26:10 +02:00
kalmar d7b7759e5d fix: delete superfluous parenthesis 2017-07-17 10:28:18 +02:00
kalmar d603bbd329 cut on allocating another GrupRingElement 2017-07-17 10:18:26 +02:00
kalmar 143695ed24 Revert "parametrise central_projection by the type of character"
This reverts commit 820517a04e.
2017-07-17 10:12:09 +02:00
kalmar 820517a04e parametrise central_projection by the type of character 2017-07-17 09:43:19 +02:00
kalmar 05691b5202 Merge branch 'enh/S_5projections' of https://git.wmi.amu.edu.pl/kalmar/PropertyT.jl into enh/S_5projections 2017-07-17 09:41:05 +02:00
kalmar aec153ca21 cosmetic 2017-07-17 09:39:46 +02:00
kalmar cb10afe6f6 save one value computation 2017-07-16 21:42:53 +02:00
kalmar 680d888b68 keep permtype consistent with partitions (i.e. rev-sorted) 2017-07-13 17:47:40 +02:00
kalmar af74ae6fb0 add hand-crafted projections for PermutationGroup(5) 2017-07-12 20:23:27 +02:00
kalmar 5906715902 use T to specify type of projection elements 2017-07-12 20:23:00 +02:00
kalmar 8819269689 central projections tend to be dense 2017-07-12 20:18:42 +02:00
kalmar 451245419c add characters for PermutationGroup(5) 2017-07-12 20:15:31 +02:00
kalmar c056c99c1f use dictionary for characters values 2017-07-12 20:15:09 +02:00
kalmar 9d8f001363 use sign(::perm) 2017-07-12 20:11:43 +02:00
kalmar 78490e6ca8 visual changes to central projection 2017-07-09 14:55:19 +02:00
kalmar 94fd8f0bf9 fix: can not assign in short-circut eval 2017-07-06 09:03:05 +02:00
kalmar 7a39775fff Merge branch 'master' into enh/sparsification 2017-07-05 13:24:21 +02:00
kalmar 94607bfaac don't redirect errors to full_log 2017-07-05 13:23:54 +02:00
kalmar 8810836fab specialise that mreps should be a dict 2017-07-05 13:23:26 +02:00
kalmar ca3e606ab6 use the dot-syntax to broadcast abs 2017-07-05 13:22:59 +02:00
kalmar b3f596d8d8 save sparsified U\pis for futher experiments 2017-07-05 13:19:41 +02:00
kalmar e648d2c370 another sparsify 2017-07-05 13:19:12 +02:00
kalmar c0e2577684 sparsify by default 2017-07-05 13:14:58 +02:00
kalmar 760270d2e0 new, unified sparsify 2017-07-05 13:14:34 +02:00
kalmar f3874828e7 increase the default precision in check_distance_to_positive_cone 2017-07-05 13:03:25 +02:00
kalmar dad6576365 better specialised groupring_square 2017-07-05 12:28:29 +02:00
kalmar 6b9d2b5d28 update to the latest version of IntervalArithmetic 2017-07-05 12:24:49 +02:00
kalmar bfdc08e91d require IntervalArithmetic v0.9.1 2017-06-25 09:34:16 +02:00
kalmar 310165b864 broader arrays in signature of distance_to_cone 2017-06-25 09:33:25 +02:00
kalmar 0e89b12e90 move functions 2017-06-25 09:31:52 +02:00
kalmar ac18923631 use IntervalArithmetic instead of ValidatedNumerics 2017-06-25 09:29:31 +02:00
kalmar ff4f9e7857 faster version of reconstruct_sol 2017-06-23 13:38:45 +02:00
kalmar fd32ddbe82 use the given precision as rationalisation precision 2017-06-23 13:30:37 +02:00
kalmar 709f030cb5 Nemo is already in scope 2017-06-22 21:06:51 +02:00
kalmar b7d9731f8b update to IntervalArithmetic v0.10.0 2017-06-22 15:48:06 +02:00
kalmar 1053a28970 remove left-over usings/imports 2017-06-22 15:31:55 +02:00
kalmar 4bde63f580 final integration via includes 2017-06-22 15:15:55 +02:00
kalmar be1c473e4a cleaning using and imports 2017-06-22 15:15:43 +02:00
kalmar 3c0ecd2e73 DirectProducts and WreathProducts are in Groups now 2017-06-22 15:14:15 +02:00
kalmar dca65e571e multiple dispatch for check_property_T 2017-06-22 15:13:37 +02:00
kalmar b881a40721 move functions around 2017-06-22 15:11:14 +02:00
kalmar f6155d8382 turn init_OrbitData into constructor 2017-06-22 15:10:26 +02:00
kalmar 3fa7491cb5 mv oPropertyT -> Orbit-wise 2017-06-22 14:40:55 +02:00
kalmar 77121c32eb add orbit-related code 2017-06-22 14:12:35 +02:00
kalmar e21f97f541 mv sdps -> SDPs 2017-06-22 14:11:33 +02:00
kalmar cd24767691 mv checksolution -> CheckSolution 2017-06-22 14:11:00 +02:00
kalmar 21e6811292 Move sqrtm out of check_distance_to_positive_cone 2017-06-22 11:56:46 +02:00
kalmar 39732bb627 make splaplacian spvector of Float64 by default 2017-06-09 09:36:20 +02:00
kalmar d98d68004b len -> wlen is no longer a kwarg, and has no default value 2017-06-09 09:35:33 +02:00
kalmar aae4ee57c7 deparametrise distance_to_cone to provide generic version 2017-06-08 21:47:09 +02:00
kalmar 98744d4f30 we need Flota64 laplacian 2017-06-08 21:46:07 +02:00
kalmar 3c0ae5baf0 dispatch splaplacian on GroupRing{T} type
with multiplicative unit for Rings
2017-06-08 21:45:46 +02:00
kalmar a68e546b7e pm is saved as pm.jld; Delta.coeffs as delta.jld 2017-06-07 00:23:51 +02:00
kalmar 5b49ae0ca0 groupring_mult -> mul 2017-06-07 00:22:06 +02:00
kalmar f6383eed7e Merge branch 'AutF4' of git.wmi.amu.edu.pl:kalmar/PropertyT.jl into AutF4
# Conflicts:
#	src/sdps.jl
2017-06-06 16:26:31 +02:00
kalmar 0c6ac791c8 basis belongs to RG 2017-06-06 16:24:42 +02:00
kalmar 883f6dcd18 Merge branch 'AutF4' of git.wmi.amu.edu.pl:kalmar/PropertyT.jl into AutF4
# Conflicts:
#	src/PropertyT.jl
#	src/sdps.jl
2017-06-06 11:54:24 +02:00
kalmar aa0f0f3034 add type argument to splaplacian 2017-06-06 11:51:15 +02:00
kalmar 4f14b0eff7 remove old product_matrix code 2017-06-06 11:50:26 +02:00
kalmar 54da1d326c import Nemo: Group, GroupElem 2017-06-06 11:48:52 +02:00
kalmar 73c9e65558 cosmetic 2017-06-05 17:31:35 +02:00
kalmar d9e6f5991e short circut for directory creation 2017-06-05 17:29:05 +02:00
kalmar 8a3e0ae62a fix: S is a generating set, we need its length 2017-06-05 17:28:39 +02:00
kalmar eac9ad2083 fix GroupRingElem creation 2017-06-05 17:27:55 +02:00
kalmar 5a9f103049 can load splaplacian as GroupRingElem 2017-06-05 17:26:55 +02:00
kalmar 3420aa5504 symmetry assertion is not necessary 2017-06-05 17:25:00 +02:00
kalmar 936c285155 less unicode 2017-06-05 17:24:45 +02:00
kalmar 54b6fda9e8 @variable(m, ..., SDP) and @SDconstraint(m,... >=0) are equivalent 2017-06-05 13:55:24 +02:00
kalmar ba04aa4f34 cosmetics 2017-06-05 13:49:57 +02:00
kalmar 79e98b07be S is now a honest Vector of groupelems 2017-06-05 13:49:43 +02:00
kalmar 477417bb9a We can again use cached Delta and pm 2017-06-05 13:48:44 +02:00
kalmar 30aea5c043 We need a group to construct group Laplacian 2017-06-05 13:48:02 +02:00
kalmar 04ecbb4a53 Generating set is an ordinary Vector 2017-06-05 13:23:29 +02:00
kalmar 0f3313b290 missing end 2017-06-05 13:23:10 +02:00
kalmar 2471c752e6 bring ΔandSDPconstraints{T<:GroupElem} from specific groups definition 2017-06-05 13:20:32 +02:00
kalmar 148bab5af2 constraints_from_pm now have default total_length 2017-06-05 13:19:57 +02:00
kalmar 8b0033da58 splaplacian_coeff -> splaplacian returns GroupRingElem 2017-06-05 13:18:46 +02:00
kalmar b35ee6bfdb remove (full) laplacian_coeff 2017-06-05 13:18:19 +02:00
kalmar 923b4f9305 simplify \DeltaandSDPconstr... 2017-06-05 13:17:49 +02:00
kalmar 9d72141624 further reduce check_property_T 2017-06-05 12:48:33 +02:00
kalmar 82d175b777 timestamp both log files 2017-06-05 12:48:13 +02:00
kalmar 58c0283a77 typo 2017-06-05 12:15:34 +02:00
kalmar 0b0c13e4f6 use setup_logger 2017-06-05 12:15:09 +02:00
kalmar 887b40b98c setup_logging 2017-06-05 10:27:07 +02:00
kalmar cbb8d5a876 adapt check_property_T to changes
It should be gone soon, anyway
2017-06-04 20:37:52 +02:00
kalmar f87e484d83 rework λandP: sets loggers and solves SDP 2017-06-04 20:37:02 +02:00
kalmar 6b20e37c74 move timed_msg 2017-06-04 20:34:12 +02:00
kalmar cd66f8d914 compute_\lambdaandP does only that 2017-06-04 20:32:40 +02:00
kalmar 25f7b528fc less unicode characters 2017-06-04 20:30:34 +02:00
kalmar e544e94be0 add word length to distance_to_cone 2017-06-04 20:29:44 +02:00
kalmar 9dfcb49d00 solve_SDP just solve 2017-06-04 20:26:05 +02:00
kalmar ae5b37e177 return both model and optimisation variables 2017-06-04 20:13:27 +02:00
kalmar 9aa427fe8e rename spectral_gap to sgap for easier handling 2017-05-28 20:07:01 +02:00
kalmar f987ae7622 typo 2017-05-28 20:06:21 +02:00
kalmar 247287b892 block loading of pm, Delta -- can't reconstruct Delta
bot JLD and serializer does not allow serialising functions;
moreover, serialising E for SOutF4 takes ~2GB
2017-05-28 20:06:04 +02:00
kalmar 5516b1ae64 using GroupRings 2017-05-28 20:03:57 +02:00
kalmar d1fb62cb8c length(::GroupRing) is the size of support, not of coeffs 2017-05-28 20:01:52 +02:00
kalmar 347c09c598 rename \epsilon -> augmentation 2017-05-28 20:01:03 +02:00
kalmar b589fb9844 algebra_square -> groupring_square, using groupring_mult 2017-05-28 20:00:22 +02:00
kalmar 2adcf0bc13 adopt to GroupRings nemoification 2017-05-28 19:59:26 +02:00
kalmar 0ba66ceaa4 remove products and generate_balls: these are in Groups.jl 2017-05-28 19:58:05 +02:00
kalmar 70b98ece5f adaptation to changes in GroupRings (not tested, not working) 2017-05-16 18:53:25 +02:00
kalmar e93ec0c422 add keyword for twisted product_matrix 2017-04-17 15:23:36 +02:00
kalmar c0da44754d early exit if fp_distance <= 0 2017-04-17 15:22:33 +02:00
kalmar d65bb54929 update for latest Memento 2017-04-17 15:20:58 +02:00
kalmar 85e07a5a57 add common routines for (unique) products and generation of balls 2017-04-10 21:46:33 +02:00
kalmar 76a2135063 change view to slice: it's not the bottleneck and thread safe 2017-04-08 13:50:59 +02:00
kalmar 8ccfe09105 correct the final printout 2017-04-08 13:49:57 +02:00
kalmar 10a0977d2e yet another change to printed format 2017-04-01 17:10:01 +02:00
kalmar 5876998cba Change names: kappa -> lambda, A -> P 2017-04-01 15:21:57 +02:00
kalmar 028979bfd1 better printing 2017-04-01 14:22:30 +02:00
kalmar 41e53a598a Finally move all intense interval arithmetic to Interval{Float64} 2017-04-01 14:22:01 +02:00
kalmar dbed75103e allow check_distance_to_positive_cone return Interval 2017-04-01 14:21:01 +02:00
kalmar b20cd02d5b Revert "tad better try...catch syntax"
This reverts commit 69e74190af.
2017-04-01 08:32:34 +02:00
kalmar a827f8c425 Revert "fix: don't throw exception in kandA"
This reverts commit 837988c381.
2017-04-01 08:27:54 +02:00
kalmar 450d43ed14 ValidatedNumerics parameters -- to be checked 2017-03-31 22:38:51 +02:00
kalmar 2b1bb7b765 break after checking in floats return negatives (commented out) 2017-03-31 22:38:18 +02:00
kalmar 98cdeedc7a make A_sqrt_Q_aug \pm \delta a matrix of Float64 intervals 2017-03-31 22:36:18 +02:00
kalmar 5aa16314ec do not force @bigintervals when we use Floats64 in a moment 2017-03-31 22:35:30 +02:00
kalmar 69e74190af tad better try...catch syntax 2017-03-31 22:34:56 +02:00
kalmar 837988c381 fix: don't throw exception in kandA
k <=0 is handled in the outside loop
2017-03-31 22:34:28 +02:00
kalmar ada830f5d5 fix: don't return after solving, save kappa and A first 2017-03-31 22:33:17 +02:00
kalmar e08e787c1a incorporate radius parameter 2017-03-31 22:32:36 +02:00
kalmar 5c36eccc7b cosmetics 2017-03-31 16:12:00 +02:00
kalmar f487c4635c Cleanup of the check_propertyT function 2017-03-31 16:07:08 +02:00
kalmar a180025dc0 log time of projecting A onto augmentation ideal 2017-03-31 15:21:45 +02:00
kalmar 9b07430b5e type parametrised create_product_matrix (works with Nemo) 2017-03-26 16:22:51 +02:00
kalmar 31393cab34 solver logger console format 2017-03-26 16:22:07 +02:00
kalmar 58217aaaf6 generating_set now returns both S and ID 2017-03-26 16:21:47 +02:00
kalmar a3779fb50a redirect stderr to full.log 2017-03-22 13:24:36 +01:00
kalmar 7796b0b028 finally fix for unbuffered output to solver-log 2017-03-22 13:24:17 +01:00
kalmar 3dc571a47c variables in blocks are reused; define A before try
or face UndefVarError: A not defined
2017-03-22 12:24:36 +01:00
kalmar 6fe0f9bda0 unredirect stderr 2017-03-20 22:12:10 +01:00
kalmar 65f2736588 logging keyword is "solver_log" 2017-03-20 22:11:57 +01:00
kalmar 87cfb742bf add and remove log Handler in the same function (solve_SDP) 2017-03-20 22:11:26 +01:00
kalmar adb9d1b273 Workaround to SCS ERROR: A->p (column pointers) decreasing
loop solver when SCS gives the error (works in Jupyter)
2017-03-20 21:48:22 +01:00
kalmar 07877556db update ValidatedNumerics to 0.8 2017-03-20 21:41:34 +01:00
kalmar 8ddb60a400 Shuffling of JuMP vars, constraints 2017-03-20 21:41:12 +01:00
kalmar 1368dfed5e errors are redirected in SL.jl 2017-03-17 17:32:49 +01:00
kalmar 0abcf2a10d report time of JuMP.solve 2017-03-17 16:32:20 +01:00
kalmar e8d3201390 no need for const inside function 2017-03-17 16:27:01 +01:00
kalmar 3517d28990 if spectral_gap is < 0 don't compute Kazhdan constant 2017-03-17 16:24:55 +01:00
kalmar 7335a94fa6 log more computation specific information 2017-03-17 16:23:36 +01:00
kalmar 220d2302bd fix true_kappa -> Kazhdan_\kappa 2017-03-16 18:20:09 +01:00