d99f8f9 | GilesColclough | 11 February 2016, 16:35:11 UTC | Merge branch 'develop-MEX' of github.com:OHBA-analysis/HMM-MAR into develop-MEX | 11 February 2016, 16:35:11 UTC |
fbb3253 | GilesColclough | 11 February 2016, 16:34:31 UTC | Updated linux install instructions after testing on hbaws23. | 11 February 2016, 16:34:31 UTC |
6959078 | Giles Colclough | 11 February 2016, 12:39:20 UTC | Updated readme Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 11 February 2016, 12:39:20 UTC |
7739885 | Giles Colclough | 11 February 2016, 12:34:34 UTC | Adjusted hsinference to use mex Mex file now works for all orders. Adjusted main code to take account of this. Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 11 February 2016, 12:34:34 UTC |
1d61002 | Giles Colclough | 11 February 2016, 12:33:40 UTC | Added readme to project Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 11 February 2016, 12:33:40 UTC |
7bef8ad | GilesColclough | 11 February 2016, 12:28:59 UTC | Updated linux mex file to version which can deal with all orders. | 11 February 2016, 12:28:59 UTC |
8ab78fa | Giles Colclough | 11 February 2016, 12:22:45 UTC | Merge branch 'develop-MEX' of ssh://github.com/OHBA-analysis/HMM-MAR into develop-MEX | 11 February 2016, 12:22:45 UTC |
4073b72 | Giles Colclough | 11 February 2016, 12:22:00 UTC | Fix issues with order != 0 The code should now run cleanly, replicating results from matlab, for all values of order parameter Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 11 February 2016, 12:22:00 UTC |
80f3793 | GilesColclough | 11 February 2016, 10:19:25 UTC | Removed documentation User can add the documentation themselves, by running make docs. | 11 February 2016, 10:19:25 UTC |
ec664a2 | GilesColclough | 11 February 2016, 10:13:57 UTC | Revert "Remove docs" Accidentaly removed the linux folder. This commit reverts that process. Desired change will be implemented next. This reverts commit 66b91f72f40e7c06fcdb9d5f3ed40a2c63abd9a7. | 11 February 2016, 10:13:57 UTC |
66b91f7 | Giles Colclough | 10 February 2016, 18:06:46 UTC | Remove docs User can build themselves. Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 10 February 2016, 18:06:46 UTC |
6397461 | Giles Colclough | 10 February 2016, 18:02:05 UTC | Now working mex compilation on linux and mac Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 10 February 2016, 18:02:05 UTC |
1a7daaf | Giles Colclough | 10 February 2016, 17:39:38 UTC | interface fix Mex file only works for order=0 atm. Prevented user accidentally encountering errors Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 10 February 2016, 17:39:38 UTC |
16c2aa1 | GilesColclough | 10 February 2016, 17:29:44 UTC | File now can use compiled mex version | 10 February 2016, 17:29:44 UTC |
bf72503 | GilesColclough | 10 February 2016, 17:29:19 UTC | Finished moving files and setting up compilation on linux | 10 February 2016, 17:29:19 UTC |
f96f5d7 | Giles Colclough | 10 February 2016, 14:41:52 UTC | Doc change Missed a bit Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 10 February 2016, 14:41:52 UTC |
36bee98 | Giles Colclough | 10 February 2016, 14:41:41 UTC | Updated documentation To take account of new test subfolder Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 10 February 2016, 14:41:41 UTC |
b379efe | Giles Colclough | 10 February 2016, 14:28:05 UTC | continuing glnx development Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 10 February 2016, 14:28:05 UTC |
ac7f482 | Giles Colclough | 10 February 2016, 14:27:12 UTC | Attempt to compile on linux Will it work without c++11, and compiled against blas, lapack, arpack? Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 10 February 2016, 14:27:12 UTC |
a35c363 | GilesColclough | 10 February 2016, 12:40:03 UTC | making it work on linux | 10 February 2016, 12:40:03 UTC |
e76c552 | Giles Colclough | 10 February 2016, 11:49:16 UTC | Moving to c++98 Because c++11 is not supported on SLES11 Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 10 February 2016, 11:49:16 UTC |
53b2033 | GilesColclough | 09 February 2016, 13:58:00 UTC | more test file reorg | 09 February 2016, 13:58:00 UTC |
15ce61c | GilesColclough | 09 February 2016, 13:57:09 UTC | Added test files and reorganised | 09 February 2016, 13:57:09 UTC |
fc9bf95 | Giles Colclough | 05 February 2016, 13:30:14 UTC | Added check for mac OS The mexfile only works on MAC atm. Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 05 February 2016, 13:30:14 UTC |
a7a4293 | Giles Colclough | 05 February 2016, 13:26:59 UTC | Hidden state inference mex project Added C++ files, make files and mex files, with documentation in html and tex, to speed up the hidden state inference process. Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 05 February 2016, 13:26:59 UTC |
a9ef366 | Giles Colclough | 05 February 2016, 13:17:31 UTC | Added option to use mex function It looks like this function gives a roughly x8 speedup for small datasets Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 05 February 2016, 13:17:31 UTC |
de009bc | Giles Colclough | 05 February 2016, 13:04:42 UTC | Fixed typo On number of arguments: generalised to account for changes in output argument. Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 05 February 2016, 13:04:42 UTC |
003ed2b | dvidaurre@gmail.com | 05 February 2016, 10:23:34 UTC | Completion of the last commit | 05 February 2016, 10:23:34 UTC |
4cafdc1 | dvidaurre@gmail.com | 05 February 2016, 10:12:30 UTC | Bug in the errors of the MAR spectra | 05 February 2016, 10:12:30 UTC |
476f291 | dvidaurre@gmail.com | 04 February 2016, 18:15:42 UTC | Accelerated multitaper errors' computation applies when options.p>0 | 04 February 2016, 18:15:42 UTC |
af9cf9e | dvidaurre@gmail.com | 04 February 2016, 17:47:11 UTC | Update on the BigHMM algorithm | 04 February 2016, 17:47:11 UTC |
132c80b | dvidaurre@gmail.com | 04 February 2016, 16:45:17 UTC | Fixed bug when mean is not modelled and there is only one channel | 04 February 2016, 16:45:17 UTC |
d309fa8 | Giles Colclough | 04 February 2016, 11:33:57 UTC | hsinference.m Ouput log probabilities Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 04 February 2016, 11:33:57 UTC |
7ad0e55 | Diego Vidaurre | 03 February 2016, 22:24:59 UTC | Delete HMM-MAR Manual.pdf | 03 February 2016, 22:24:59 UTC |
1711f77 | dvidaurre@gmail.com | 03 February 2016, 22:14:09 UTC | Sign disambiguation of M/EEG source-reconstructed data Added an utility (unflipchannels.m) to flip the sign of some channels/subjects, in order to solve the sign ambiguity problem produced by M/EEG source reconstruction. | 03 February 2016, 22:14:09 UTC |
53c9cad | Giles Colclough | 03 February 2016, 14:56:14 UTC | Merge branch 'master' of ssh://github.com/OHBA-analysis/HMM-MAR | 03 February 2016, 14:56:14 UTC |
8640ae7 | Giles Colclough | 03 February 2016, 14:53:28 UTC | DTF Added computation of DTF non-parametrically Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 03 February 2016, 14:53:28 UTC |
852de02 | Giles Colclough | 03 February 2016, 14:53:08 UTC | Added manual Signed-off-by: Giles Colclough <giles.colclough@gmail.com> | 03 February 2016, 14:53:08 UTC |
7c70b8d | dvidaurre@gmail.com | 01 February 2016, 14:04:45 UTC | little mistake in the PDC error bar calculations of the parametric estiamtion | 01 February 2016, 14:04:45 UTC |
173138e | dvidaurre@gmail.com | 01 February 2016, 11:47:31 UTC | removing debugging output | 01 February 2016, 11:47:31 UTC |
b509506 | dvidaurre@gmail.com | 01 February 2016, 11:40:25 UTC | Fixed a bug in the BigHMM, in the computation of the covariance matrix of the coefficients (and other stuff about style) | 01 February 2016, 11:40:25 UTC |
b284f51 | dvidaurre@gmail.com | 30 January 2016, 18:06:56 UTC | Changed MAR spectra so that the diagonals of the PSD are presented in its absolute value, and change of default value for completelags to 1 | 30 January 2016, 18:06:56 UTC |
6402933 | dvidaurre@gmail.com | 25 January 2016, 18:40:30 UTC | Added some regularisation in the the initialisation to prevent overflow in the first iteration when data is too extreme | 25 January 2016, 18:40:30 UTC |
9ccc216 | dvidaurre@gmail.com | 19 January 2016, 18:14:56 UTC | Introduced a stochastic variational optimization strategy to make possible the computation of the Gaussian (order=0) HMM for massive data sets, with very high number of subjects | 19 January 2016, 18:14:56 UTC |
7c31eb7 | dvidaurre@gmail.com | 19 January 2016, 18:13:24 UTC | The normalisation in the entropy calculation of P(X_t |X_t-1) had a little error - fixed | 19 January 2016, 18:13:24 UTC |
6fbe7a9 | dvidaurre@gmail.com | 18 January 2016, 17:58:43 UTC | remove annoying output introduced in the last commit | 18 January 2016, 17:58:43 UTC |
2093572 | dvidaurre@gmail.com | 18 January 2016, 17:54:59 UTC | Fixed little mistake in the computation of the state time course entropy (makes little difference) | 18 January 2016, 17:54:59 UTC |
703aa87 | dvidaurre@gmail.com | 17 January 2016, 12:29:08 UTC | Added the possibility to complete to the full set of lags when computing the parametric spectra | 17 January 2016, 12:29:08 UTC |
a311243 | dvidaurre@gmail.com | 30 December 2015, 18:34:03 UTC | Fixed a couple of things when the covariance matrix is not subject-specific | 30 December 2015, 18:34:08 UTC |
3bff6b8 | woolrich | 18 December 2015, 19:31:15 UTC | statepath is returned as empty if Gamma is passed in | 18 December 2015, 19:31:15 UTC |
f868c2d | woolrich | 18 December 2015, 17:56:38 UTC | Added option sim_state_tcs_only to avoid computing X if not needed, and also now returns the statepath | 18 December 2015, 17:56:43 UTC |
1d8b955 | dvidaurre@gmail.com | 10 December 2015, 11:28:45 UTC | Included tool to compute the transition probability matrix for separated segments of the data | 10 December 2015, 11:28:45 UTC |
e413a78 | dvidaurre@gmail.com | 01 December 2015, 13:59:15 UTC | GMM initialisation has been modified so that it doesn't break when the data is poorly conditioned | 01 December 2015, 13:59:15 UTC |
d0a3137 | dvidaurre@gmail.com | 23 November 2015, 16:31:43 UTC | Introduced the possibility of using the multitaper/MAR frequency analysis for stationary signals, where no segmentation has been performed | 23 November 2015, 16:31:43 UTC |
d6b8472 | dvidaurre@gmail.com | 18 November 2015, 13:58:17 UTC | Corrected dumb typo introduced in last commit | 18 November 2015, 13:58:17 UTC |
dd93c64 | dvidaurre@gmail.com | 18 November 2015, 11:08:46 UTC | Small bug corrected in the computation of the state time courses entropy (free energy) | 18 November 2015, 11:08:46 UTC |
cafe533 | dvidaurre@gmail.com | 18 November 2015, 11:08:08 UTC | Small bug corrected in the kullback leibler divergence between gamma distributions | 18 November 2015, 11:08:08 UTC |
f2938ed | dvidaurre@gmail.com | 18 November 2015, 10:49:16 UTC | tweak to make computations faster when the number of channels is high | 18 November 2015, 10:49:16 UTC |
948f5f1 | dvidaurre@gmail.com | 18 November 2015, 10:47:31 UTC | Netlab modified to suppress annoying screen output | 18 November 2015, 10:47:31 UTC |
763b519 | dvidaurre@gmail.com | 17 November 2015, 13:19:36 UTC | Delete temporal files | 17 November 2015, 13:19:36 UTC |
c013f20 | dvidaurre@gmail.com | 14 November 2015, 09:52:33 UTC | Fixed small bug in the computation of the entropy of the state time courses (evaluation of the free energy) | 14 November 2015, 09:52:33 UTC |
0016d95 | dvidaurre@gmail.com | 06 November 2015, 19:55:50 UTC | Corrected bug when the number of channels is 1 | 06 November 2015, 19:55:50 UTC |
b38a86e | dvidaurre@gmail.com | 06 November 2015, 18:47:06 UTC | Removed a pedantic control of the type of covariance matrix when the number of channels is 1 (when it is specified to be full or uniquefull it does not give an error, and just corrects it to be diagonal) | 06 November 2015, 18:47:06 UTC |
de80c34 | dvidaurre@gmail.com | 05 November 2015, 15:14:43 UTC | Solved bug in hmmspectramar | 05 November 2015, 15:14:43 UTC |
5db1079 | dvidaurre@gmail.com | 05 November 2015, 12:20:10 UTC | Added the possibililty of using a Bayesian estimation of the MARs in hmmspectramar | 05 November 2015, 12:20:10 UTC |
70951b1 | dvidaurre@gmail.com | 05 November 2015, 11:17:36 UTC | Small change in mlhmmmar to ease compatibility with older versions of the software | 05 November 2015, 11:17:36 UTC |
27c7527 | dvidaurre@gmail.com | 04 November 2015, 17:29:38 UTC | Added a missing scaling factor in the parametric spectra estimation - function hmmspectramar | 04 November 2015, 17:29:38 UTC |
2bb3ad5 | dvidaurre@gmail.com | 03 November 2015, 09:37:44 UTC | Add .gitignore, to ignore *.m~ files | 03 November 2015, 09:37:44 UTC |
dc32f59 | dvidaurre@gmail.com | 02 November 2015, 18:49:25 UTC | Fixed typo in hmmspectramar | 02 November 2015, 18:49:25 UTC |
f920970 | vidaurre | 02 November 2015, 16:36:35 UTC | remove clear all in example script | 02 November 2015, 16:36:35 UTC |
784f2b3 | dvidaurre@gmail.com | 02 November 2015, 16:35:24 UTC | clear all in example script | 02 November 2015, 16:35:24 UTC |
0d54fd3 | dvidaurre@gmail.com | 02 November 2015, 16:30:45 UTC | Add all existing files | 02 November 2015, 16:30:45 UTC |