| 1fc140a9 | 13-Feb-2019 |
Xiang Huang <xianghuang@gmail.com> |
BRGN: finished tomography test, passed with three test cases: l1dict, l2prox, user (2-norm regularizer ||x||^2) |
| 30eeff36 | 05-Feb-2019 |
Xiang Huang <xianghuang@gmail.com> |
Both tomography and cs1 examples work with both l2prox and l1dict |
| a3c390cf | 01-Feb-2019 |
Alp Dener <adener@anl.gov> |
algorithm structure changes to combine old L2 reg with new L1 reg, and permit user regularizers |
| ef73bedd | 11-Jan-2019 |
Xiang Huang <xianghuang@gmail.com> |
added bounded constraint, non-negative x |
| 8d7b1102 | 11-Jan-2019 |
Xiang Huang <xianghuang@gmail.com> |
Tomography example works: Consolidated A,b, xGT into one file (was three files). |
| c73dea25 | 10-Jan-2019 |
Xiang Huang <xianghuang@gmail.com> |
read from binary Mat/Vec, eliminate the array datastructure, works for a small example now! |
| 3ca75ad3 | 08-Jan-2019 |
Xiang Huang <xianghuang@gmail.com> |
adding tomography L1 reconstruction example, done for generating data in matlab and convert to petsc binary format, todo read data and test reconstruction |
| 8e85b1b3 | 11-Dec-2018 |
Xiang Huang <xianghuang@gmail.com> |
added D matrix for L1 norm |
| 6e9726d0 | 06-Dec-2018 |
Xiang Huang <xianghuang@gmail.com> |
cleared comments of matrix vector dimension, and targeted problem to solve |
| 20fe612c | 06-Dec-2018 |
Xiang Huang <xianghuang@gmail.com> |
fixed the initialize x, added template for inserting D matrix |
| 7cea06e1 | 01-Nov-2018 |
Xiang Huang <xianghuang@gmail.com> |
the compressive sensing (sparse regression) test example works with brgn of the smoothed L1 form |
| 62974d9a | 01-Nov-2018 |
Xiang Huang <xianghuang@gmail.com> |
debug cs1 with intermediate output |
| 70902e93 | 01-Nov-2018 |
Xiang Huang <xianghuang@gmail.com> |
add compressive sensing test example cs1.c |
| 8ac80d48 | 30-Oct-2018 |
Xiang Huang <xianghuang@gmail.com> |
tao-brgn add feature smoothed L1 norm algorithm, after meeting with Todd |
| 0d71dc2b | 23-Oct-2018 |
Xiang Huang <xianghuang@gmail.com> |
brgn modified with approximation of L1 regularizer |
| d61acda5 | 15-Feb-2019 |
Karl Rupp <me@karlrupp.net> |
Merge branch 'denera/tao-brgn-matlab-segfault-fix' [PR #1184]
* denera/tao-brgn-matlab-segfault-fix: MATLAB Leastsquares examples segfault fix for BRGN. Trust region initialization was producing a s
Merge branch 'denera/tao-brgn-matlab-segfault-fix' [PR #1184]
* denera/tao-brgn-matlab-segfault-fix: MATLAB Leastsquares examples segfault fix for BRGN. Trust region initialization was producing a segfault on MatMult with the Jacobian.
Evaluating the Jacobian in MATLAB using a SEQAIJ PETSc type, but then converting it to SEQDENSE before using it in Tao operations resolved the problem.
show more ...
|
| 7faa94c2 | 12-Feb-2019 |
Alp Dener <adener@anl.gov> |
adding the matlab test to the automated test suite |
| 3850be85 | 13-Feb-2019 |
Alp Dener <adener@anl.gov> |
TaoCreate_XYZ() docstring updates |
| 2c9dcad7 | 10-Feb-2019 |
Barry Smith <bsmith@mcs.anl.gov> |
Change CHKERRQ(ierr) after PetscInitialize() calls in examples to if (ierr) return ierr;
Since the PETSc initialization may not have been completed in the failed PetscInitialize() it is not propper
Change CHKERRQ(ierr) after PetscInitialize() calls in examples to if (ierr) return ierr;
Since the PETSc initialization may not have been completed in the failed PetscInitialize() it is not propper to call the PETSc error handling routines
Commit-type: bug-fix, testing-fix, example Reported-by: jed Brown <jed@jedbrown.org>
show more ...
|
| df4be7ee | 16-Jan-2019 |
Matthew G. Knepley <knepley@gmail.com> |
Merge branch 'master' into knepley/feature-multiple-ds
* master: (156 commits) Added missing CHKERRQ() KSP ex56: Added alt output file for ex56_2. Revert "KSP ex56: Updated test output and dom
Merge branch 'master' into knepley/feature-multiple-ds
* master: (156 commits) Added missing CHKERRQ() KSP ex56: Added alt output file for ex56_2. Revert "KSP ex56: Updated test output and domain size for test _2." DMStag tutorial ex4: fix stencil errors Hierarchical partitioner: added an alt output file Add support for DMGetMatrix() with periodic boundary conditions in one dimension for DMDA. Fixed incorrect 'ntext' argument in PetscOptionsEList call in SNES ex12.c ISCreateSubIS: Initialize variables owner and lidx to avoid compiler warnings metis: provide -download-metis-use-doubleprecision option KSP ex56: Updated test output and domain size for test _2. Hierarchical partitioner: add a bit documentation ISCreateSubIS: Added checkers for parameters Added a test for hierarchical partitioning vertex weights Plex+Doc: Document options for DMPlexCreateBoxMesh() Hierarchical partitioner supports vertex weights Create an API ISCreateSubIS that extracts Add an alt output file Plex: Can now give bounds to DMPlexCreateBoxMesh() from options - Use -dm_plex_box_lower and dm_plex_box_upper PCMG: These methods should be optional PCMG: Make PCMGGet/SetLevels() proper member functions - Also need to compose them in GAMG ...
Conflicts: src/dm/impls/plex/plexsection.c
show more ...
|
| c107311a | 05-Jan-2019 |
Matthew G. Knepley <knepley@gmail.com> |
Tao ex3: Updated to new interface |
| e5e52638 | 04-Jan-2019 |
Matthew G. Knepley <knepley@gmail.com> |
DM+DS: Reorganized DS handling completely do that multiple regions can be supported - A region is labeled portion of the mesh which has the same approximation space for the solution. Note that this i
DM+DS: Reorganized DS handling completely do that multiple regions can be supported - A region is labeled portion of the mesh which has the same approximation space for the solution. Note that this is NOT the same as the label indicating the support of a field. Regions are the refinement of the field supports into unique approximation space pieces. Right now, we only support a single strategy, namely one space on the regular mesh and one on the hybrid cells, but we only have to change DMCreateDS() to extend the capability. - Now instead of using PetscDSSetDiscretization() and DMSetDS(), you use DMSetField() and DMCreateDS(). Also, you now must use DMCopyDisc(), instead of DMSetDS(), which copies both the field and DS structure to another DM. These changes have been made in all examples. - DMGetCellDS() returns the approximation space for that cell - DMGetDS() return the default space since most people will not use regions. There is still a lot of work left t generalize all methods to multiple DSes, but all old code still functions. - DMGetRegion/RegionNumDS() gets spaces by label and region number - DMCreateDS() is the setup method which makes the regions and corresponding DSes. With some work, this could be converted to a lazy initialization. - Cleaned up access to DS since now we can have more than one - Projection was improved to extract the DS for the particular cells it is iterating over - The DS are set from options during Section creation. The idea is that people using DS would also use automatic Section creation, or they would manage things themselves and call SetFromOptions manually. - I updated the subDM and superDM constructors, but they only work for a single region
show more ...
|
| b31a704f | 02-Jan-2019 |
Barry Smith <bsmith@mcs.anl.gov> |
Fix MatSeqAIJFromMatlab() it assumed that the inode structure could change only if the number of nonzeros changed in the matrix but, in fact, the struture of the matrix could change while the number
Fix MatSeqAIJFromMatlab() it assumed that the inode structure could change only if the number of nonzeros changed in the matrix but, in fact, the struture of the matrix could change while the number of nonzeros remained the same.
Fixed memory leak in example and missing CHKERRQ(ierr);
Commit-type: bug-fix, testing-fix Reported-by: Alp Dener <adener@anl.gov>
show more ...
|
| 04f58fc4 | 19-Oct-2018 |
Alp Dener <adener@anl.gov> |
matlab LS problem residual jacobian API updated to latest |
| 494bef23 | 18-Oct-2018 |
Alp Dener <adener@anl.gov> |
Fixed double-printing of solver header in trust region initialization |