1a5cc67 | asardaes | 20 March 2016, 15:04:46 UTC | More fixes for fuzzy clustering | 20 March 2016, 15:04:46 UTC |
1f1796d | asardaes | 19 March 2016, 23:11:50 UTC | Some more changes to use lengths() | 19 March 2016, 23:11:50 UTC |
0fd0d59 | asardaes | 12 March 2016, 16:17:13 UTC | Add R version to description Function 'lengths' was introduced in R 3.2.0 | 12 March 2016, 16:17:13 UTC |
7a6dcd6 | asardaes | 24 February 2016, 08:34:03 UTC | Fixed predict for fuzzy clustering | 24 February 2016, 08:34:03 UTC |
9fe2325 | Alexis Sarda | 21 February 2016, 12:10:07 UTC | Documentation fixes | 21 February 2016, 12:10:07 UTC |
ba2d267 | Alexis Sarda | 20 February 2016, 19:52:58 UTC | Update main documentation | 20 February 2016, 19:52:58 UTC |
782ef73 | Alexis Sarda | 20 February 2016, 17:23:21 UTC | Minor cleaning | 20 February 2016, 17:23:21 UTC |
8e40b99 | Alexis Sarda | 20 February 2016, 16:52:05 UTC | Added show.centers parameter to plot method | 20 February 2016, 16:52:05 UTC |
678bd37 | Alexis Sarda | 20 February 2016, 15:56:34 UTC | Clean SBD and REFERENCES.bib | 20 February 2016, 15:56:34 UTC |
b15b449 | asardaes | 20 February 2016, 13:29:29 UTC | Update package description | 20 February 2016, 13:29:29 UTC |
d36d9d9 | asardaes | 17 February 2016, 18:40:26 UTC | Added missing fclusters | 17 February 2016, 18:40:26 UTC |
0af9545 | asardaes | 15 February 2016, 09:07:00 UTC | Fixed fuzzy clustering algorithm The definition was wrong, the distance should be squared by me and the objective function was also not being considered. | 15 February 2016, 09:07:00 UTC |
7d27efd | Alexis Sarda | 14 February 2016, 21:42:37 UTC | Update namespace and news | 14 February 2016, 21:42:37 UTC |
3daa852 | Alexis Sarda | 14 February 2016, 21:12:34 UTC | Support for a vector of k values And quite a few bug fixes | 14 February 2016, 21:12:34 UTC |
dec3664 | asardaes | 14 February 2016, 10:39:51 UTC | Modified show method | 14 February 2016, 10:39:51 UTC |
8cea4ac | Alexis Sarda | 13 February 2016, 19:29:57 UTC | Added fuzzy c-means | 13 February 2016, 19:29:57 UTC |
83e8256 | asardaes | 07 February 2016, 14:33:39 UTC | Moved flexclust to Imports | 07 February 2016, 14:33:39 UTC |
92c7965 | asardaes | 07 February 2016, 14:06:40 UTC | C helper for DBA I think I have to do the update in C in order to prevent R from allocating memory at each iteration. | 07 February 2016, 14:06:40 UTC |
d69bf81 | asardaes | 06 February 2016, 22:47:23 UTC | Update documentation | 06 February 2016, 22:47:23 UTC |
174cbb3 | asardaes | 05 January 2016, 17:41:14 UTC | Fix minor typos | 05 January 2016, 17:41:14 UTC |
df5607b | Alexis Sarda | 03 January 2016, 22:17:37 UTC | Skipping x32-specific tests on CRAN | 03 January 2016, 22:17:37 UTC |
5fbc55b | asardaes | 03 January 2016, 15:07:08 UTC | Misc. cleaning | 03 January 2016, 15:07:08 UTC |
749ac25 | asardaes | 02 January 2016, 15:42:18 UTC | Preserve rng attribute If several repetitions are made, the %dorng% operator returns seed sequences in the "rng" attribute. | 02 January 2016, 15:42:18 UTC |
026e000 | asardaes | 31 December 2015, 16:46:12 UTC | Updated news | 31 December 2015, 16:46:12 UTC |
9e875be | asardaes | 30 December 2015, 19:12:48 UTC | Added NEWS.Rd | 30 December 2015, 19:12:48 UTC |
df7c292 | asardaes | 30 December 2015, 17:51:52 UTC | Cleaned some parallel functions The doSEQ backend is used if no parallel backend is provided, so 'dopar' will also work with a 1-object list. | 30 December 2015, 17:51:52 UTC |
f60adc2 | asardaes | 30 December 2015, 17:17:03 UTC | Cleaned x32 tests Only 3 tests vary between architectures, apparently. Probably because SBD depends on FFT and changes accuracy. | 30 December 2015, 17:17:03 UTC |
1a8cab0 | asardaes | 30 December 2015, 17:15:51 UTC | Pairwise tests | 30 December 2015, 17:15:51 UTC |
200263d | asardaes | 30 December 2015, 16:23:36 UTC | Parallel proxy lower bounds | 30 December 2015, 16:23:36 UTC |
4eb8968 | asardaes | 30 December 2015, 15:47:49 UTC | Parallel SBD.proxy | 30 December 2015, 15:47:49 UTC |
80ce6d1 | asardaes | 30 December 2015, 11:46:15 UTC | Minor misc. | 30 December 2015, 11:46:15 UTC |
0bd0d22 | Alexis Sarda | 30 December 2015, 00:32:54 UTC | Added missing new i386 file Re-computed with reinterpolated data. Also some code cleaning. | 30 December 2015, 00:32:54 UTC |
b4e9aa2 | asardaes | 29 December 2015, 23:56:36 UTC | Simplified dtwdistfun Which allowed me to simplify check_parallel too. | 29 December 2015, 23:56:36 UTC |
593b97c | asardaes | 29 December 2015, 23:55:08 UTC | Test for symmetric optimization | 29 December 2015, 23:55:08 UTC |
5b4798e | asardaes | 29 December 2015, 12:11:33 UTC | Testing pairwise workaround As suggested by the maintainer of proxy, a PREFUN function is able of detecting and passing the pairwise argument, but right now using that workaround results in too many warnings. Hopefully it'll be fixed in a new version of proxy. | 29 December 2015, 12:11:33 UTC |
5f8fdf1 | asardaes | 28 December 2015, 16:12:12 UTC | Simplified split_parallel | 28 December 2015, 16:12:12 UTC |
7e5b7d9 | Alexis Sarda | 28 December 2015, 10:51:59 UTC | Tests without data Changed save.data to FALSE to save some space on disk | 28 December 2015, 10:51:59 UTC |
c12a139 | asardaes | 28 December 2015, 10:28:15 UTC | Added repetitions tests | 28 December 2015, 10:28:15 UTC |
55b1d93 | Alexis Sarda | 28 December 2015, 00:09:33 UTC | Added i386 testhat tests I created a function to detect architecture and compare the appropriate files. However, now I can't run Check within RStudio in Windows, because it doesn't skip the parallel tests and results in an error. The tests alone run fine, but the Check only succeeds if done in a cmd window. Linux should be fine. | 28 December 2015, 00:09:33 UTC |
9544030 | asardaes | 27 December 2015, 22:32:51 UTC | Added testthat tests I was already doing some checks to make sure things were consistent, now it should be easier. | 27 December 2015, 22:32:51 UTC |
55770af | asardaes | 27 December 2015, 21:36:51 UTC | Strict check_parallel For some reason the previous method worked fine for distance calculations, but it gave weird results with centroid calculations. | 27 December 2015, 21:36:51 UTC |
3a7428f | asardaes | 27 December 2015, 21:05:56 UTC | check_parallel now allows doSEQ In order to be able to take advantage of symmetric distances. | 27 December 2015, 21:05:56 UTC |
2d52dc3 | asardaes | 27 December 2015, 20:49:07 UTC | Added symmetric slot to control If the distance matrix is symmetric, some time can be saved during calculation. As a result, parallel dtw and dtw2 now check for series lengths. | 27 December 2015, 20:49:07 UTC |
adf602d | asardaes | 26 December 2015, 22:31:21 UTC | Fixed Check warning I think I have to import the update generic to be able to export a corresponding method. | 26 December 2015, 22:31:21 UTC |
785c595 | asardaes | 26 December 2015, 21:02:10 UTC | Minor cleaning | 26 December 2015, 21:02:10 UTC |
f58050f | asardaes | 26 December 2015, 14:02:12 UTC | Improved PAM centroids The custom function now uses attributes and names to keep track of which series are the current centers. | 26 December 2015, 14:02:12 UTC |
a3843d1 | asardaes | 26 December 2015, 12:50:42 UTC | Fixed centroid random re-start Also modified the functions so that they all recompute only the necessary centers. | 26 December 2015, 12:50:42 UTC |
5b93ca9 | asardaes | 26 December 2015, 02:09:53 UTC | Improved centroid functions I refactored them and at the end I check if a cluster is empty, in which case I randomly re-initialize a new one. | 26 December 2015, 02:09:53 UTC |
994c91a | asardaes | 25 December 2015, 17:00:33 UTC | Workaround for proxy's entry_exists I think the pr_DB$entry_exists function is not working correctly, so I made a workaround with the try() function. | 25 December 2015, 17:00:33 UTC |
904299f | asardaes | 23 December 2015, 20:49:25 UTC | Update methods' docu | 23 December 2015, 20:49:25 UTC |
e07975f | asardaes | 23 December 2015, 17:27:43 UTC | Update classes' docu | 23 December 2015, 17:27:43 UTC |
56095d5 | asardaes | 23 December 2015, 17:09:55 UTC | Dropped support for non-proxy distances Otherwise, pam.precompute and all the parallel optimizations would never work. | 23 December 2015, 17:09:55 UTC |
9908aff | asardaes | 23 December 2015, 16:34:05 UTC | Update docu | 23 December 2015, 16:34:05 UTC |
3a827d3 | Alexis Sarda-Espinosa | 23 December 2015, 11:37:23 UTC | Improved dtwdistfun Now it should detect extra parameters better, either for distance functions or proxy. | 23 December 2015, 11:37:23 UTC |
cd4215e | Alexis Sarda-Espinosa | 23 December 2015, 11:24:21 UTC | Parallel pairwise in dtw_lb | 23 December 2015, 11:24:21 UTC |
c8cc9c4 | Alexis Sarda | 22 December 2015, 09:38:25 UTC | Added missing elements to TADPole result It will slow it down a little bit though. And it won't take advantage of parallel backend for now. | 22 December 2015, 09:38:25 UTC |
b28e4f0 | Alexis Sarda | 22 December 2015, 08:58:39 UTC | Fixed for predict generic when TADPole was used | 22 December 2015, 08:58:39 UTC |
274d5e3 | Alexis Sarda | 22 December 2015, 08:45:17 UTC | Fix minor dtwclust.R bugs Distance would always be a character in the end, so I should check that it actually exists in proxy. I'm not sure the predict generic would work if TADPole was used. | 22 December 2015, 08:45:17 UTC |
65b9391 | asardaes | 21 December 2015, 22:18:45 UTC | Testing faster computation of envelopes | 21 December 2015, 22:18:45 UTC |
8ab1712 | asardaes | 20 December 2015, 22:49:35 UTC | Minor TADPole optimizations | 20 December 2015, 22:49:35 UTC |
55d4b11 | asardaes | 20 December 2015, 15:32:44 UTC | Update documentation | 20 December 2015, 15:32:44 UTC |
e1b78b4 | asardaes | 20 December 2015, 13:48:00 UTC | Pairwise bug workaround | 20 December 2015, 13:48:00 UTC |
fb2b3cb | Alexis Sarda | 19 December 2015, 15:38:35 UTC | Updated shape_extraction Same functionality, but updated documentation and the way the code checks for special cases. | 19 December 2015, 15:38:35 UTC |
8b6f09a | Alexis Sarda | 19 December 2015, 14:19:07 UTC | Moved save.data to control | 19 December 2015, 14:19:07 UTC |
e94b5b3 | Alexis Sarda | 19 December 2015, 12:27:27 UTC | Fixed mean and median centroids There was a typo from an old version. | 19 December 2015, 12:27:27 UTC |
4e99867 | Alexis Sarda | 19 December 2015, 11:57:57 UTC | Bug fixes Extra parameters for custom distances were being matched twice via dots and dots2. Also, dtw_lb shouldn't be used with pam.precompute = TRUE, so that now throws a warning. Pairs function was also fixed. | 19 December 2015, 11:57:57 UTC |
59d43a2 | Alexis Sarda | 18 December 2015, 20:05:51 UTC | Added control slot to dtwclust class | 18 December 2015, 20:05:51 UTC |
fe29e51 | Alexis Sarda | 18 December 2015, 19:46:24 UTC | Improvements for hierarchical procedures Now several runs can be done in one go, and plotting can also handle extra arguments. Also updated documentation. | 18 December 2015, 19:46:24 UTC |
7ecc602 | Alexis Sarda | 17 December 2015, 23:04:31 UTC | Dropped many formal arguments Replaced them with a formal control class. For now, the old way will still be supported and detected through ... with a message. | 17 December 2015, 23:04:31 UTC |
950c2c4 | Alexis Sarda | 16 December 2015, 23:10:51 UTC | Plot types Option to choose between default or dendrogram plot. Implemented cluster centers for hierarchical methods so default plot works. | 16 December 2015, 23:10:51 UTC |
fa5a225 | Alexis Sarda | 15 December 2015, 23:00:19 UTC | Dropped kccasimple inheritance Ported several slots and methods, and implemented my own family class. | 15 December 2015, 23:00:19 UTC |
8176ee9 | Alexis Sarda | 14 December 2015, 22:56:37 UTC | Major refactor I will probably drop the flexclust dependency. Now all functions expect a list of time series as input. | 14 December 2015, 22:56:37 UTC |
9b4519f | Alexis Sarda | 13 December 2015, 13:05:29 UTC | Update .gitignore Removing results from compilation. | 13 December 2015, 13:05:29 UTC |
7033c9c | Alexis Sarda | 13 December 2015, 10:49:01 UTC | Minor refactor Now a function checks for parallel backend. | 13 December 2015, 10:49:01 UTC |
32fe2a5 | Alexis Sarda | 12 December 2015, 20:32:46 UTC | Shape extraction with different lengths The shape extration algorithm now supports series with different lengths. A reference series must be provided in that case. The clustering routines were also adapted to allow for this. | 12 December 2015, 20:32:46 UTC |
e4b5144 | Alexis Sarda | 12 December 2015, 15:35:15 UTC | Parallel centroids Adapted centroid calculation when using DBA and shape_extraction to support parallelization. DBA itself can also use it, although it will only be helpful if averaging a lot of series. | 12 December 2015, 15:35:15 UTC |
7e04841 | Alexis Sarda | 12 December 2015, 10:25:57 UTC | Bug fix related to parallel task creation If there were more cores than tasks, some indices would be empty, and that could potentially result in an error. | 12 December 2015, 10:25:57 UTC |
d39eaee | Alexis Sarda | 10 December 2015, 22:03:06 UTC | Parallel support for dtw_lb | 10 December 2015, 22:03:06 UTC |
4d592cc | Alexis Sarda | 09 December 2015, 19:44:39 UTC | Parallel support for TADPole Lots of code refactoring | 09 December 2015, 19:44:39 UTC |
0827c89 | Alexis Sarda | 08 December 2015, 22:25:45 UTC | Parallel support for distance matrix calculation Also refactored a lot of the code. | 08 December 2015, 22:25:45 UTC |
7bf74c2 | Alexis Sarda | 05 December 2015, 23:06:28 UTC | Added option for (possibly parallel) repetitions See the Repetitions section and the corresponding new example. | 05 December 2015, 23:06:28 UTC |
1125182 | Alexis Sarda | 18 November 2015, 20:24:40 UTC | Minor class change Slot "proctime" appropriately defined now. | 18 November 2015, 20:24:40 UTC |
c7788ae | Alexis Sarda | 14 November 2015, 15:41:07 UTC | Version 1.2.0 (resubmission) Added missing imports and a quick bug fix of TADPole | 14 November 2015, 15:41:07 UTC |
2c424f8 | Alexis Sarda | 14 November 2015, 10:43:14 UTC | Version 1.2.0 Final layout adjustments. | 14 November 2015, 10:43:14 UTC |
edaa77a | Alexis Sarda | 13 November 2015, 22:44:32 UTC | SBD, DBA and shape_extraction optimizations They should all be a bit faster now. Several other bug fixes included, and more examples in the README. | 13 November 2015, 22:44:32 UTC |
a3e3fb3 | Alexis Sarda | 12 November 2015, 21:44:25 UTC | PAM option to prevent pre-computation, custom distance example, improvements Added an example using a custom distance registered with proxy. This was partly to test some related bug fixes. Also using closures instead of relying on environment detection, that way it doesn't matter if data is changed or preprocessed. | 12 November 2015, 21:44:25 UTC |
5948968 | Alexis Sarda | 11 November 2015, 17:26:00 UTC | Changed step pattern to symmetric1 within TADPole Euclidean distance is only valid as an upper bound of DTW if the symmetric1 step pattern is used. | 11 November 2015, 17:26:00 UTC |
1b74b0f | Alexis Sarda | 05 November 2015, 10:26:34 UTC | Fix missing variable If data is provided through the 'data' argument as a matrix, L was not being calculated. | 05 November 2015, 10:26:34 UTC |
c128dd5 | Alexis Sarda | 12 September 2015, 12:04:42 UTC | Final fixes for v1.1.0 | 12 September 2015, 12:04:42 UTC |
3c06416 | Alexis Sarda | 09 September 2015, 21:16:07 UTC | Further improvements to LBs and DBA Probably release this as v1.1.0 | 09 September 2015, 21:16:07 UTC |
32a9683 | Alexis Sarda | 06 September 2015, 13:43:28 UTC | Fixed LB Improved and added options to DBA Also checking if DBA can be optimized by pre-allocating matrices. | 06 September 2015, 13:43:28 UTC |
6b5849d | Alexis Sarda | 29 August 2015, 21:35:13 UTC | Added more options to the plot method Custom time values and the ability to prevent printing to screen | 29 August 2015, 21:35:13 UTC |
57c85b0 | Alexis Sarda | 23 August 2015, 12:33:02 UTC | Finished adjustments for series of different lengths v1.0.0 | 23 August 2015, 12:33:02 UTC |
8c64852 | Alexis Sarda | 22 August 2015, 23:58:54 UTC | Limited support of time series with different lengths Also optimized some code layout and fixed a couple of bugs. | 22 August 2015, 23:58:54 UTC |
ac63d78 | Alexis Sarda | 22 August 2015, 13:30:45 UTC | Added preproc to class definition Also changed default znorm of SBD to FALSE | 22 August 2015, 13:30:45 UTC |
eebc344 | Alexis Sarda | 22 August 2015, 12:35:54 UTC | Added support for custom preprocessing Also fixed a bug that could occur when calling custom distance/preprocessing functions. Minor updates and code layout changes. | 22 August 2015, 12:35:54 UTC |
f8b9a03 | Alexis Sarda | 13 August 2015, 21:10:23 UTC | Added DBA iterations to dtwclust function The maximum number of allowed iterations for DBA centroids can now be specified when using dtwclust | 13 August 2015, 21:10:23 UTC |
f9caf72 | Alexis Sarda | 12 August 2015, 20:34:20 UTC | Incremented max.iter for DBA | 12 August 2015, 20:34:20 UTC |
3154183 | Alexis Sarda | 05 August 2015, 20:33:07 UTC | ggplot2 is now a dependency, modified the plot method The plot method can now accept another argument to modify title and axis labels. It also returns the gg object. Some more examples and cleaning too. | 05 August 2015, 20:33:07 UTC |
bf2c49c | Alexis Sarda | 02 August 2015, 11:05:37 UTC | Resubmission Added necessary imports and reduced the number of examples being executed | 02 August 2015, 11:05:37 UTC |