2e4e842 | AustinRochford | 20 November 2016, 19:23:57 UTC | Use marginalized mixture classes for DP example | 20 November 2016, 19:23:57 UTC |
d57000d | Thomas Wiecki | 18 November 2016, 14:11:30 UTC | DOC Add None option for NUTS intialization. | 18 November 2016, 14:11:30 UTC |
dbf0654 | Thomas Wiecki | 18 November 2016, 13:59:18 UTC | ENH Add ADVI initializing for continuous models sampled using NUTS. | 18 November 2016, 13:59:18 UTC |
ad7dc8a | Chris Fonnesbeck | 18 November 2016, 10:21:34 UTC | Replaced fixed kappa with passed value (#1531) | 18 November 2016, 10:21:34 UTC |
59f500c | Thomas Wiecki | 18 November 2016, 08:33:20 UTC | TST Fix Rhat bad test by changing seed and starting point. | 18 November 2016, 08:33:20 UTC |
34e3f45 | Thomas Wiecki | 17 November 2016, 15:04:41 UTC | MAINT Convert some examples to new sample() api. Adapt tests. | 17 November 2016, 15:04:41 UTC |
b91d3be | Thomas Wiecki | 17 November 2016, 13:32:00 UTC | Merge pull request #1528 from taku-y/fix_sbtrans Fix stickbreaking transform dtype | 17 November 2016, 13:32:00 UTC |
34a8bd4 | taku-y | 17 November 2016, 10:48:17 UTC | Fix stickbreaking transform dtype | 17 November 2016, 10:48:17 UTC |
85e0f9d | Thomas Wiecki | 16 November 2016, 09:46:05 UTC | REF Make all_continuous() a general purpose function. | 16 November 2016, 09:46:05 UTC |
4d53a36 | Thomas Wiecki | 16 November 2016, 09:37:51 UTC | WIP Perform initializing inside of sample instead of adding new sample_init() as per @fonnesbeck suggestion. | 16 November 2016, 09:37:51 UTC |
e3c8b63 | Thomas Wiecki | 11 November 2016, 15:04:26 UTC | DOC Add better docs and add advi_map. | 16 November 2016, 07:17:36 UTC |
85b46f6 | Thomas Wiecki | 11 November 2016, 10:51:57 UTC | MAINT Remove Metropolis init as it does not seem to work well. | 16 November 2016, 07:17:36 UTC |
dc203d9 | Thomas Wiecki | 11 November 2016, 08:33:23 UTC | DOC Convert more NBs to use sample_init(). | 16 November 2016, 07:17:36 UTC |
77b67d0 | Thomas Wiecki | 11 November 2016, 06:20:15 UTC | ENH Add sample_init() function and update examples. | 16 November 2016, 07:17:36 UTC |
1770be2 | Thomas Wiecki | 11 November 2016, 06:12:57 UTC | ENH Infer variable names directly from model in covariance estimation from trace. | 16 November 2016, 07:17:36 UTC |
4f54823 | Thomas Wiecki | 14 November 2016, 20:41:46 UTC | ENH: Add Dockerfile for development inside a container (#1527) * Add Dockerfile for development inside a container * Use /Users/austin for portability * Put Dockerfile with scripts * Add script for starting development Docker container * Really add script this time * Modify options to create_testenv.sh * Shift while parsing arguments * Update dockerfile to run test suite * Add documentation for docker * Typo | 14 November 2016, 20:41:46 UTC |
7367f20 | colin | 14 November 2016, 20:39:54 UTC | Typo | 14 November 2016, 20:39:54 UTC |
d8800bf | colin | 13 November 2016, 00:01:34 UTC | Add documentation for docker | 13 November 2016, 18:38:58 UTC |
75dac62 | colin | 13 November 2016, 00:01:14 UTC | Update dockerfile to run test suite | 13 November 2016, 18:38:58 UTC |
08fbeeb | AustinRochford | 01 November 2016, 00:44:48 UTC | Shift while parsing arguments | 13 November 2016, 18:38:58 UTC |
9d38375 | AustinRochford | 01 November 2016, 00:08:01 UTC | Modify options to create_testenv.sh | 13 November 2016, 18:38:58 UTC |
c489ea6 | AustinRochford | 23 October 2016, 22:10:35 UTC | Really add script this time | 13 November 2016, 18:38:58 UTC |
655a4db | AustinRochford | 23 October 2016, 21:37:43 UTC | Add script for starting development Docker container | 13 November 2016, 18:38:58 UTC |
f60492c | AustinRochford | 23 October 2016, 15:46:01 UTC | Put Dockerfile with scripts | 13 November 2016, 18:38:58 UTC |
3180376 | AustinRochford | 23 October 2016, 15:45:37 UTC | Use /Users/austin for portability | 13 November 2016, 18:38:58 UTC |
231e30f | AustinRochford | 23 October 2016, 13:09:05 UTC | Add Dockerfile for development inside a container | 13 November 2016, 18:38:58 UTC |
082c982 | Thomas Wiecki | 12 November 2016, 13:40:11 UTC | Nuts speed up ~1.5x (#1522) * PERF Make NUTS 1.5x faster by moving E0 calculation outside of buildtree inner loop. * STY Refactor creation of functions. * DOC Adapt doc-string. * STY Pep8. * STY Only return scalar * DOC Typo * MAINT Move theano.tensor to tt. Remove obsolete p0 and q0 args from build_tree. | 12 November 2016, 13:40:11 UTC |
0d5f292 | Maxim | 12 November 2016, 09:31:01 UTC | MAINT glm fixes and refactoring (#1513) * Fixed a bug when it is not possible to run more than one glm models in a model * names are as were in previous versions by default * Added glm.from_xy constructor * Fixed confusing docstrings * Refactored Code | 12 November 2016, 09:31:01 UTC |
3f300e9 | Osvaldo Martin | 12 November 2016, 08:40:18 UTC | change circular transform from ElemwiseTransform to Transform. add circular domain for vonmises tests (#1520) | 12 November 2016, 08:40:18 UTC |
ebf26d5 | colin | 10 November 2016, 01:35:06 UTC | Remove `verbose` argument | 10 November 2016, 13:42:41 UTC |
c04da31 | Colin | 09 November 2016, 16:42:06 UTC | Merge pull request #1516 from pymc-devs/transform_verbosity Changed verbosity threshold for variable transformation | 09 November 2016, 16:42:06 UTC |
1f5480f | Chris Fonnesbeck | 09 November 2016, 15:39:16 UTC | Used logging.debug instead of verbosity level | 09 November 2016, 15:39:16 UTC |
4ca4ef9 | Chris Fonnesbeck | 09 November 2016, 15:26:45 UTC | Changed verbosity threshold for variable transformation | 09 November 2016, 15:26:45 UTC |
51f196d | Colin | 08 November 2016, 07:23:44 UTC | Accept `extent` as kwarg to kde_plot (#1509) | 08 November 2016, 07:23:44 UTC |
1e8549e | Thomas Wiecki | 08 November 2016, 06:29:20 UTC | Merge pull request #1508 from ColCarroll/unpin Unpin theano | 08 November 2016, 06:29:20 UTC |
4ca37f7 | colin | 08 November 2016, 04:21:06 UTC | Unpin theano | 08 November 2016, 04:37:38 UTC |
9be2a02 | springcoil | 07 November 2016, 14:29:19 UTC | ENH: Changed the notebook slightly adding Taku as an author and removing the branch reference | 07 November 2016, 23:57:32 UTC |
2fb658a | Colin | 07 November 2016, 14:21:27 UTC | Merge pull request #1504 from hvasbath/kdekwargs kde plot added kwargs | 07 November 2016, 14:21:27 UTC |
e494d1c | Eigenblutwurst | 07 November 2016, 13:35:21 UTC | move cmap to kwargs | 07 November 2016, 13:35:21 UTC |
4ce9e2c | Eigenblutwurst | 07 November 2016, 10:19:52 UTC | kde plot added kwargs | 07 November 2016, 10:19:52 UTC |
433a9c7 | Taku Yoshioka | 06 November 2016, 23:29:23 UTC | Changes on ADVI functions (#1499) * pep8 * Change the definition of optimizers * Use theano float type * ENH Support models without global RVs * Fix ADVI code for tests * Update notebooks that use ADVI * Add a convolutional vae notebook as an example | 06 November 2016, 23:29:23 UTC |
aa72167 | Hannes Vasyura-Bathke | 06 November 2016, 15:31:30 UTC | added cmap option to kde plots (#1500) | 06 November 2016, 15:31:30 UTC |
113e720 | Colin | 02 November 2016, 11:57:57 UTC | Informative error message in find_MAP (#1493) | 02 November 2016, 11:57:57 UTC |
97f11f4 | Colin | 02 November 2016, 09:24:37 UTC | Pin theano version for now (#1494) | 02 November 2016, 09:24:37 UTC |
587828c | Thomas Wiecki | 02 November 2016, 07:08:38 UTC | Merge pull request #1489 from pymc-devs/arbitrary_deterministic_fix Fixed import error in disaster_model_arbitrary_deterministic | 02 November 2016, 07:08:38 UTC |
bef8df4 | Chris Fonnesbeck | 31 October 2016, 15:24:57 UTC | Fixed import error in disaster_model_arbitrary_deterministic | 31 October 2016, 15:24:57 UTC |
719cd3e | marmaduke woodman | 28 October 2016, 17:58:02 UTC | IPyNB and unit test for EM SDE (#1486) * add original example * restart nb & convert oscillator * add lin sde as test | 28 October 2016, 17:58:02 UTC |
1327069 | Thomas Wiecki | 28 October 2016, 14:38:44 UTC | Merge pull request #1485 from ColCarroll/global_env Add --global flag to create_testenv for docker build | 28 October 2016, 14:38:44 UTC |
5577f40 | colin | 27 October 2016, 23:43:53 UTC | Add --global flag to create_testenv for docker build | 27 October 2016, 23:43:53 UTC |
ed5245b | Austin Rochford | 26 October 2016, 20:50:09 UTC | Merge pull request #1484 from AustinRochford/flexible-multinomial Support 1D n and 2D p in Multinomial | 26 October 2016, 20:50:09 UTC |
96093c6 | Chris Fonnesbeck | 26 October 2016, 18:22:23 UTC | Merge pull request #1482 from pymc-devs/assert_negative Changed check for non-positive support to check for negative support | 26 October 2016, 18:22:23 UTC |
51ee020 | AustinRochford | 26 October 2016, 17:35:18 UTC | Change 2d to 1d where appropriate in tests | 26 October 2016, 17:35:18 UTC |
601b0f3 | AustinRochford | 26 October 2016, 17:32:40 UTC | Support 1D n and 2D p in Multinomial | 26 October 2016, 17:32:40 UTC |
5e2e466 | Colin | 26 October 2016, 06:29:43 UTC | Fix parallel sampling (#1481) * Fix parallel sampling * Everything is good again | 26 October 2016, 06:29:43 UTC |
09fba92 | Chris Fonnesbeck | 26 October 2016, 00:23:43 UTC | Fixed typos | 26 October 2016, 00:23:43 UTC |
77d8715 | Chris Fonnesbeck | 25 October 2016, 23:19:01 UTC | Fixed typo | 25 October 2016, 23:19:01 UTC |
f52d411 | Chris Fonnesbeck | 25 October 2016, 23:17:24 UTC | Changed check for non-positive support to check for negative support | 25 October 2016, 23:17:24 UTC |
954b01d | Chris Fonnesbeck | 25 October 2016, 17:49:43 UTC | Merge pull request #1478 from AustinRochford/bugfix-multinomial-vector Bugfix multinomial vector | 25 October 2016, 17:49:43 UTC |
1a21db4 | AustinRochford | 25 October 2016, 03:17:11 UTC | Fix Multinomial logp when there is more than one observation | 25 October 2016, 03:17:11 UTC |
b7fc184 | AustinRochford | 25 October 2016, 02:46:52 UTC | Add failing test for multinomial with many observations | 25 October 2016, 02:46:52 UTC |
540da5f | Austin Rochford | 23 October 2016, 21:45:12 UTC | Merge pull request #1475 from AustinRochford/git-ignore-test-artifacts Ignore SQLite file generated by tests | 23 October 2016, 21:45:12 UTC |
39637ed | AustinRochford | 23 October 2016, 21:44:18 UTC | Ignore SQLite file generated by tests | 23 October 2016, 21:44:18 UTC |
2a659d8 | Colin | 23 October 2016, 18:17:59 UTC | Merge pull request #1474 from pymc-devs/generalize_fmin Generalize find_MAP to accept more scipy.optimize functions | 23 October 2016, 18:17:59 UTC |
1e845bf | Chris Fonnesbeck | 23 October 2016, 17:28:13 UTC | Generalize find_MAP to accept more scipy.optimize functions | 23 October 2016, 17:28:40 UTC |
968f341 | Chris Fonnesbeck | 23 October 2016, 13:06:37 UTC | Added warnings for non-positive support in positive parameters (#1472) * Added warnings for non-positive support in positive parameters for continuous distributions * Improved message from assert_nonpos_support | 23 October 2016, 13:06:37 UTC |
362c8ee | Chris Fonnesbeck | 22 October 2016, 14:55:17 UTC | More robust transformations and scaling adjustments (#1461) * More robust transformations and scaling adjustments * Fixed bad name for scaling bound * Tests sd rather than tau (or both) in normal logp * Added default scaling_bound for adjust_precision | 22 October 2016, 14:55:17 UTC |
854a6ab | Chris Fonnesbeck | 20 October 2016, 18:36:34 UTC | Replaced multinomial probability bounds test in logp (#1462) | 20 October 2016, 18:36:34 UTC |
3de8330 | Mehmet Ozan Kabak | 19 October 2016, 05:13:24 UTC | Fixed minor backend bugs. (#1456) * Fixed minor backend bugs. * Before this fix, new samples were not appended to a SQLite backend that was loaded using the function sqlite.load. * Fixed a Pandas related deprecation warning. * Load functions now raise sensible exceptions if the requested backend can not be loaded. * Fixes as requested by @twiecki. | 19 October 2016, 05:13:24 UTC |
49ef10a | Thomas Wiecki | 18 October 2016, 19:55:23 UTC | DOC Fix math notation. Closes #1453. | 18 October 2016, 19:55:39 UTC |
7cd5166 | Thomas Wiecki | 18 October 2016, 11:55:35 UTC | DOC Link to github Quantopian logo rather than locally. | 18 October 2016, 11:55:35 UTC |
4db0184 | Thomas Wiecki | 18 October 2016, 11:17:41 UTC | DOC Add all backend submodules. | 18 October 2016, 11:17:41 UTC |
aa8e667 | Thomas Wiecki | 18 October 2016, 09:41:34 UTC | DOC Fix formatting in readme. Add links for further resources. | 18 October 2016, 09:41:34 UTC |
1889b46 | Thomas Wiecki | 18 October 2016, 09:24:23 UTC | DOC Add advi-minibatch, diagnostics and backends to docs. Fix some doc string math problems. | 18 October 2016, 09:24:23 UTC |
ce225df | Thomas Wiecki | 18 October 2016, 08:53:39 UTC | DOC Restructure docs and fix some heading problems. Also include new NUTS scaling NB. | 18 October 2016, 08:53:39 UTC |
370b5b9 | Jonathan Sedar | 18 October 2016, 08:36:57 UTC | DOC Adding an example Notebook for NUTS scaling using ADVI stdevs (#1457) It seems to work, so submitting an example Notebook in the hope it helps someone else. Please see Notebook for method, results and system details too. | 18 October 2016, 08:36:57 UTC |
2572852 | Austin Rochford | 18 October 2016, 08:33:52 UTC | ENH Mixture Models (#1437) * First pass at mixture modelling * No longer necessary to reference self.comp_dists directly in logp * Add dimension internally (when necessary) * Import get_tau_sd * Misc bugfixes * Add sampling to Mixtures * Differentiate between Discrete and Continuous mixtures when possible * Add support for 2D weights * Gracefully try to calculate mean and mode defaults * Add docstrings for Mixture classes * Export mixture models * Reference self.comp_dists * Remove unnecessary pm. * Add Mixture tests * Add missing imports * Add marginalized Gaussian mixture model example * Calculate the mode of the mixture distribution correctly | 18 October 2016, 08:33:52 UTC |
f1e622f | Thomas Wiecki | 13 October 2016, 15:20:20 UTC | Fix notebooks to run with current api | 18 October 2016, 08:21:24 UTC |
07cba62 | colin | 13 October 2016, 01:20:03 UTC | Update notebooks with current API, and add kwarg to tracetab | 18 October 2016, 08:21:24 UTC |
e75bbf6 | Thomas Wiecki | 17 October 2016, 22:18:56 UTC | DOC Add raw link to logo. | 17 October 2016, 22:18:56 UTC |
68e488a | Colin | 17 October 2016, 19:59:29 UTC | Merge pull request #1455 from pymc-devs/logo Added new logo | 17 October 2016, 19:59:29 UTC |
85e3f4f | Chris Fonnesbeck | 17 October 2016, 17:08:55 UTC | Changed image url to point to master | 17 October 2016, 17:08:55 UTC |
51a06c0 | Chris Fonnesbeck | 17 October 2016, 17:08:06 UTC | Replaced title with logo | 17 October 2016, 17:08:06 UTC |
f7fe7df | Chris Fonnesbeck | 17 October 2016, 17:06:07 UTC | Added new logo to README | 17 October 2016, 17:06:07 UTC |
345fcbc | Chris Fonnesbeck | 17 October 2016, 16:52:24 UTC | Added logo for use in README | 17 October 2016, 16:52:24 UTC |
05473eb | Colin | 17 October 2016, 02:36:07 UTC | Merge pull request #1454 from AustinRochford/sampling-progressbars Add progress bar option to PPC and VP sampling | 17 October 2016, 02:36:07 UTC |
b8fd064 | AustinRochford | 17 October 2016, 02:20:15 UTC | Add progress bar option to PPC and VP sampling | 17 October 2016, 02:20:15 UTC |
1b7b4bb | Austin Rochford | 14 October 2016, 13:49:49 UTC | Don't broadcast in alltrue (#1452) | 14 October 2016, 13:49:49 UTC |
4a21624 | Colin | 14 October 2016, 06:34:09 UTC | Make diagnostic tests reproducible (#1451) | 14 October 2016, 06:34:09 UTC |
100dbb1 | Thomas Wiecki | 13 October 2016, 15:20:20 UTC | DOC NUTS requires covariance matrix, so square the standard-deviations from ADVI. Closes #1448. | 13 October 2016, 15:20:20 UTC |
b43faec | Osvaldo Martin | 13 October 2016, 14:32:52 UTC | fix broken link (#1446) | 13 October 2016, 14:32:52 UTC |
6e5dcba | Colin | 13 October 2016, 14:32:07 UTC | Update pmf-pymc example (#1447) | 13 October 2016, 14:32:07 UTC |
6f0013f | Mehmet Ozan Kabak | 13 October 2016, 07:28:54 UTC | Implemented CategoricalGibbsMetropolis, optimized BinaryGibbsMetropolis. (#1439) * Implemented CategoricalGibbsMetropolis, optimized BinaryGibbsMetropolis. * Added unit test for CategoricalGibbsMetropolis. Also Added extra comments requested by PyMC3 devs, and fixed a minor bug in the proportional proposal utilized by CategoricalGibbsMetropolis. * Fix a minor bug in the newly added CategoricalGibbsMetropolis test. | 13 October 2016, 07:28:54 UTC |
2f29af8 | Osvaldo Martin | 12 October 2016, 16:25:50 UTC | Model comparison waic with standard error plot (#1442) * add estimation of standard error * waic standard error computation and plot * remove 0.7 ValueError loo * turn ValueError 0.7 into warning * prevent showing both warnings, fix typo | 12 October 2016, 16:25:50 UTC |
5b2b7f0 | Colin | 12 October 2016, 14:46:17 UTC | Merge pull request #1440 from pymc-devs/plot_posterior_vector Generalized plot_posterior to plot vector-valued variables | 12 October 2016, 14:46:17 UTC |
092faf2 | Chris Fonnesbeck | 12 October 2016, 13:15:49 UTC | Added test for multidimensional plot_posterior | 12 October 2016, 13:15:49 UTC |
90c7286 | Osvaldo Martin | 12 October 2016, 10:18:16 UTC | add estimation of standard error (#1441) | 12 October 2016, 10:18:16 UTC |
16d63c6 | Chris Fonnesbeck | 11 October 2016, 23:30:13 UTC | Generalized plot_posterior to plot vector-valued variables | 11 October 2016, 23:30:13 UTC |
a92c740 | Thomas Wiecki | 11 October 2016, 09:45:24 UTC | DOC Fix ADVI initialization in stoch vol model. | 11 October 2016, 09:45:24 UTC |
f8079ae | Colin | 10 October 2016, 10:38:15 UTC | Clean up metropolis imports (#1433) | 10 October 2016, 10:38:15 UTC |
67f7a81 | Peadar Coyle | 08 October 2016, 16:31:49 UTC | Revert "WIP robust interval transforms" (#1432) | 08 October 2016, 16:31:49 UTC |