a8e182d | ST John | 06 April 2021, 10:21:54 UTC | Merge branch 'develop' of github.com:GPflow/GPflow into st/new_svgp-redesign | 06 April 2021, 10:21:54 UTC |
21e5317 | John Mcleod | 06 April 2021, 10:15:35 UTC | Proposed fix for the fully_correlated_conditional_repeat issue (#1652) Fixes #1651 | 06 April 2021, 10:15:35 UTC |
5b76536 | John Mcleod | 06 April 2021, 09:31:52 UTC | Fix fallback inducing points with multioutput Kuf. (#1654) Fixes #1653 | 06 April 2021, 09:31:52 UTC |
405eb97 | st-- | 19 March 2021, 14:04:42 UTC | Update README.md (#1603) * reorganization of structure * update maintainers * update project list, including link to signature covariances project * table of contents | 19 March 2021, 14:04:42 UTC |
5c7e091 | ST John | 15 March 2021, 18:33:34 UTC | format | 15 March 2021, 18:33:34 UTC |
a9d889b | ST John | 15 March 2021, 18:23:31 UTC | Experiment: use the caching rather than fused predict_f in SVGP | 15 March 2021, 18:23:31 UTC |
5103ece | ST John | 15 March 2021, 17:17:51 UTC | _get_kernels was only used by IndependentPosterior; turn it into specialized _get_Kff method in that class | 15 March 2021, 17:17:51 UTC |
2f52c3f | ST John | 15 March 2021, 17:16:46 UTC | Merge branch 'develop' of github.com:GPflow/GPflow into st/new_svgp-redesign | 15 March 2021, 17:16:46 UTC |
629e7b2 | st-- | 15 March 2021, 17:16:31 UTC | fix bug (that masked test failure!) in test_method_equivalence (#1649) * fix bug (that masked test failure!) in test_method_equivalence and improve collection speed * optimize non-whitened SVGP model more strongly to get test_equivalence to pass * remove --use-feature=2020-resolver (will become an error in pip 21.0) | 15 March 2021, 17:16:31 UTC |
380ce0f | ST John | 15 March 2021, 15:56:14 UTC | update vff notebook | 15 March 2021, 15:56:14 UTC |
984b400 | ST John | 15 March 2021, 15:44:16 UTC | Merge branch 'st/fix_test_method_equivalence' into st/new_svgp-redesign | 15 March 2021, 15:44:16 UTC |
95bf01c | ST John | 15 March 2021, 15:00:40 UTC | remove --use-feature=2020-resolver (will become an error in pip 21.0) | 15 March 2021, 15:43:36 UTC |
f691dbd | ST John | 15 March 2021, 14:31:27 UTC | optimize non-whitened SVGP model more strongly to get test_equivalence to pass | 15 March 2021, 15:43:29 UTC |
78ecd71 | ST John | 15 March 2021, 15:42:52 UTC | better handling of mean function (fixes mcmc notebook) | 15 March 2021, 15:42:52 UTC |
5a85cde | ST John | 15 March 2021, 15:19:18 UTC | minimal change to get VFF notebook to work with new Posterior objects | 15 March 2021, 15:19:18 UTC |
ef3f26a | ST John | 15 March 2021, 14:34:46 UTC | Merge branch 'st/fix_test_method_equivalence' into st/new_svgp-redesign | 15 March 2021, 14:34:46 UTC |
3d42bbf | ST John | 15 March 2021, 14:31:27 UTC | optimize non-whitened SVGP model more strongly to get test_equivalence to pass | 15 March 2021, 14:31:57 UTC |
7f86e63 | ST John | 15 March 2021, 14:10:52 UTC | fix bug (that masked test failure!) in test_method_equivalence and improve collection speed | 15 March 2021, 14:11:06 UTC |
ae9e692 | ST John | 15 March 2021, 14:06:04 UTC | add missing mean_function calls in posterior predict f | 15 March 2021, 14:06:04 UTC |
d4d8a44 | ST John | 15 March 2021, 12:50:29 UTC | another fix via mypy | 15 March 2021, 12:50:29 UTC |
a11e83f | ST John | 15 March 2021, 12:45:56 UTC | fix test_method_equivalence to only optimise *in test*, not during collection, to trace down bug (and make collecting faster!) | 15 March 2021, 12:45:56 UTC |
ad5a758 | ST John | 15 March 2021, 11:17:01 UTC | Merge branch 'st/new_svgp' into st/new_svgp-redesign | 15 March 2021, 11:17:01 UTC |
646ca7a | ST John | 15 March 2021, 11:16:47 UTC | format | 15 March 2021, 11:16:47 UTC |
d0b4556 | ST John | 15 March 2021, 11:15:58 UTC | Merge branch 'develop' of github.com:GPflow/GPflow into st/new_svgp | 15 March 2021, 11:15:58 UTC |
e68e2fe | ST John | 15 March 2021, 11:05:22 UTC | only run mypy on gpflow/tests (not docs) | 15 March 2021, 11:05:22 UTC |
9f3410b | ST John | 15 March 2021, 10:41:06 UTC | fixes via mypy | 15 March 2021, 10:41:06 UTC |
ac700bb | ST John | 15 March 2021, 10:22:50 UTC | Merge branch 'develop' of github.com:GPflow/GPflow into st/new_svgp-redesign | 15 March 2021, 10:22:50 UTC |
1831a5d | st-- | 07 March 2021, 14:05:42 UTC | Update README and require TensorFlow>=2.2 (#1647) Add explanation of TensorFlow/TFP dependencies (as brought up by https://github.com/GPflow/GPflow/issues/1623) | 07 March 2021, 14:05:42 UTC |
94fa390 | st-- | 25 February 2021, 16:12:48 UTC | New _add_noise_cov method for GPR model class to make it more easily extensible (#1645) | 25 February 2021, 16:12:48 UTC |
3ea8731 | Vatsal Aggarwal | 18 February 2021, 16:29:39 UTC | Compatibility between Monitor API and SciPy optimizer (#1642) Co-authored-by: st-- <st--@users.noreply.github.com> | 18 February 2021, 16:29:39 UTC |
5ffd5af | ST John | 03 February 2021, 14:32:05 UTC | use fused_predict_f in NewSVGP | 03 February 2021, 14:32:05 UTC |
406d2d0 | Samuel Willis | 03 February 2021, 13:45:29 UTC | fix ModelToTensorBoard max_size = -1 (#1619) | 03 February 2021, 13:45:29 UTC |
797fed1 | st-- | 03 February 2021, 12:38:37 UTC | Mypy build fix (#1637) | 03 February 2021, 12:38:37 UTC |
627c39b | ST John | 03 February 2021, 10:03:56 UTC | commit missing types module | 03 February 2021, 10:03:56 UTC |
7fadfd7 | ST John | 03 February 2021, 09:54:18 UTC | fix linearcoreg posterior fused | 03 February 2021, 09:54:18 UTC |
7e39e3f | ST John | 02 February 2021, 18:19:22 UTC | more fixes | 02 February 2021, 18:19:22 UTC |
65d01fe | ST John | 02 February 2021, 18:08:51 UTC | fix imports and q_sqrt=None case | 02 February 2021, 18:08:51 UTC |
8523f4d | ST John | 02 February 2021, 17:58:16 UTC | route all conditionals() through Posterior objects: fused_predict_f | 02 February 2021, 17:58:16 UTC |
3ff7923 | ST John | 28 January 2021, 17:14:37 UTC | incorporate code from posterior_conditional.py | 28 January 2021, 17:14:37 UTC |
b83864a | ST John | 28 January 2021, 12:35:20 UTC | LinearOperatorBasePosterior -- WIP | 28 January 2021, 12:35:20 UTC |
99a8ab7 | ST John | 28 January 2021, 12:35:01 UTC | introduce _set_qdist | 28 January 2021, 12:35:01 UTC |
2e08fad | ST John | 28 January 2021, 12:34:24 UTC | bugfix | 28 January 2021, 12:34:24 UTC |
25d9abd | ST John | 28 January 2021, 12:33:39 UTC | bugfix in VFF notebook (set is_Positive_definite) | 28 January 2021, 12:33:39 UTC |
5eae47a | ST John | 28 January 2021, 12:29:04 UTC | introduce Abstract base class | 28 January 2021, 12:29:04 UTC |
bc04306 | Vincent Dutordoir | 13 January 2021, 16:31:51 UTC | Quickfix dynamic shapes in quadrature (#1626) | 13 January 2021, 16:31:51 UTC |
692ab2f | ST John | 13 January 2021, 14:22:43 UTC | Dispatcher instead of if-elif for posterior_class | 13 January 2021, 14:22:43 UTC |
696b30e | ST John | 13 January 2021, 14:22:13 UTC | move out fully_correlated path | 13 January 2021, 14:22:13 UTC |
48d2050 | ST John | 12 January 2021, 17:00:49 UTC | less adjoints | 12 January 2021, 17:00:49 UTC |
4fd77b4 | ST John | 12 January 2021, 16:51:32 UTC | clean up FUllyCorrelated branching | 12 January 2021, 16:51:32 UTC |
cb9ad16 | stefanos | 12 January 2021, 16:21:26 UTC | creating additional posterior objects for cleaner structure | 12 January 2021, 16:21:26 UTC |
9f5b79b | ST John | 12 January 2021, 14:39:51 UTC | write out iv -> inducing_variable | 12 January 2021, 14:39:51 UTC |
fe731ba | ST John | 12 January 2021, 14:14:37 UTC | Merge branch 'develop' of github.com:GPflow/GPflow into st/new_svgp | 12 January 2021, 14:14:37 UTC |
0a2c2e3 | ST John | 12 January 2021, 14:13:59 UTC | add comments | 12 January 2021, 14:13:59 UTC |
721756a | jakoblindinger | 12 January 2021, 12:31:17 UTC | list structured-dgp as downstream project (#1621) Co-authored-by: st-- <st--@users.noreply.github.com> | 12 January 2021, 12:31:17 UTC |
7688477 | st-- | 12 January 2021, 11:59:48 UTC | Fix tests for TF 2.4/TFP 0.12 (#1625) * fix test_deepcopy * increase test_scipy_jit tolerance to make it pass | 12 January 2021, 11:59:48 UTC |
f6a8bf4 | ST John | 11 January 2021, 15:06:27 UTC | T -> ᵀ in math comments | 11 January 2021, 15:06:27 UTC |
c7d3b76 | ST John | 21 December 2020, 12:16:09 UTC | revert to posterior() method to fix printing | 21 December 2020, 12:16:09 UTC |
19d735f | ST John | 21 December 2020, 12:05:56 UTC | format | 21 December 2020, 12:05:56 UTC |
56c5aff | ST John | 21 December 2020, 12:05:23 UTC | code review suggestions | 21 December 2020, 12:05:23 UTC |
870baaf | st-- | 21 December 2020, 12:03:17 UTC | Apply suggestions from code review Co-authored-by: Fergus <fergus2@gmail.com> | 21 December 2020, 12:03:17 UTC |
f4cfcc9 | ST John | 21 December 2020, 12:01:33 UTC | slightly increase atol/rtol in test_scipy following Posterior / math reordering changes | 21 December 2020, 12:01:33 UTC |
47e5b78 | stefanos | 18 December 2020, 03:29:03 UTC | fully correlated mok works, all multioutput tests pass. needs clean-up | 18 December 2020, 03:29:03 UTC |
735f6b0 | stefanos | 17 December 2020, 20:42:28 UTC | fully correlated mok with full_cov==full_output_cov works | 17 December 2020, 20:42:28 UTC |
ff07c9c | stefanos | 17 December 2020, 19:07:39 UTC | wip for more general predict_f | 17 December 2020, 19:07:39 UTC |
23e532a | ST John | 17 December 2020, 14:56:20 UTC | enable gradient flow through posterior for training | 17 December 2020, 14:56:20 UTC |
1337995 | stefanos | 17 December 2020, 14:21:27 UTC | correct argument passing to Posterior | 17 December 2020, 14:21:27 UTC |
d632732 | stefanos | 17 December 2020, 14:04:29 UTC | removing initial None value for the cached variables in posterior class | 17 December 2020, 14:04:29 UTC |
2fd59c6 | stefanos | 17 December 2020, 13:56:04 UTC | make posterior work with parameter object w/o recreating variables | 17 December 2020, 13:56:04 UTC |
ad3b2ba | stefanos | 17 December 2020, 12:33:04 UTC | black format | 17 December 2020, 12:33:04 UTC |
ba71ff1 | stefanos | 17 December 2020, 11:52:36 UTC | added posterior object to new module | 17 December 2020, 11:54:17 UTC |
00c4b20 | ST John | 16 December 2020, 21:21:19 UTC | fix printing tests | 16 December 2020, 21:21:19 UTC |
ac8c585 | ST John | 16 December 2020, 18:30:07 UTC | avoid recreating parameter | 16 December 2020, 18:30:07 UTC |
b15aee1 | ST John | 16 December 2020, 18:17:57 UTC | NewSVGP posterior test | 16 December 2020, 18:17:57 UTC |
92e3176 | ST John | 16 December 2020, 18:17:19 UTC | fix imports | 16 December 2020, 18:17:19 UTC |
148a92e | ST John | 16 December 2020, 18:05:32 UTC | OldSVGP -> NewSVGP | 16 December 2020, 18:05:32 UTC |
dd2e0a5 | ST John | 16 December 2020, 18:03:31 UTC | NewSVGP | 16 December 2020, 18:04:02 UTC |
60e19f8 | st-- | 01 December 2020, 11:59:33 UTC | bump version to 2.1.4 (#1612) | 01 December 2020, 11:59:33 UTC |
61fc327 | Joel Berkeley | 26 November 2020, 17:14:04 UTC | list trieste as downstream project (#1611) * list trieste as downstream project Co-authored-by: Joel Berkeley <joel.berkeley@secondmind.ai> Co-authored-by: st-- <st--@users.noreply.github.com> | 26 November 2020, 17:14:04 UTC |
282d1cd | polaschwoebel | 15 October 2020, 17:54:23 UTC | HeteroskedasticTFPConditional should construct tensors at class-construction, not at module-import time (#1598) | 15 October 2020, 17:54:23 UTC |
8562503 | st-- | 08 October 2020, 15:04:55 UTC | Replace len(inducing_variable) with inducing_variable.num inducing property (#1594) Adds support for inducing variables with dynamically changing shape. Change usage from `len(inducing_variable)` to `inducing_variable.num_inducing` instead. Resolves #1578. | 08 October 2020, 15:04:55 UTC |
9f95d87 | st-- | 07 October 2020, 12:01:42 UTC | bump version to 2.1.3 (#1590) | 07 October 2020, 12:01:42 UTC |
cfd1830 | st-- | 07 October 2020, 11:46:04 UTC | Doc requirements update (#1592) * update doc requirements * update copyright | 07 October 2020, 11:46:04 UTC |
638c4ed | st-- | 07 October 2020, 10:07:35 UTC | Bugfix for independent_interdomain_conditional (#1583) Resolves #818 Co-authored-by: ST John <st@secondmind.ai> | 07 October 2020, 10:07:35 UTC |
39c8a76 | st-- | 06 October 2020, 15:25:12 UTC | enable trigger-docs-generation for master branch as well (#1580) Co-authored-by: ST John <st@secondmind.ai> | 06 October 2020, 15:25:12 UTC |
3baf368 | st-- | 06 October 2020, 15:03:54 UTC | Remove tiling workaround for lack of broadcasting in previous TensorFlow versions (#1585) We used to tile in gauss_kl() to work around TensorFlow's lack of broadcasting. Since TensorFlow 2.2, this is finally working, and this PR removes the tiling for TensorFlow>=2.2. Resolves #1321. Co-authored-by: ST John <st@secondmind.ai> | 06 October 2020, 15:03:54 UTC |
e248051 | st-- | 06 October 2020, 14:12:33 UTC | update intro_to_gpflow2 notebook to account for saved_model now working (#1586) Co-authored-by: ST John <st@secondmind.ai> Co-authored-by: Sandeep Tailor <s.tailor@insysion.net> | 06 October 2020, 14:12:33 UTC |
98f7997 | st-- | 06 October 2020, 13:30:52 UTC | fix broken API docs (#1587) * move Version check inside reset_cache_bijectors to not confuse Sphinx * revert non-working "projects: 7" line in bugs issue template * fix code-block type in intro.rst | 06 October 2020, 13:30:52 UTC |
8dc912e | st-- | 06 October 2020, 12:37:15 UTC | Fix notebook heading levels and add links (#1579) * fix heading levels * add cross-references | 06 October 2020, 12:37:15 UTC |
7d6a6da | st-- | 05 October 2020, 15:50:31 UTC | Update issue templates (#1581) * automatically add bugs to bug project board * disable blank issues Co-authored-by: ST John <st@secondmind.ai> | 05 October 2020, 15:50:31 UTC |
b74e19e | st-- | 05 October 2020, 10:48:01 UTC | Explicit NotImplementedError on predict_y/predict_log_density with full_(output_)cov=True (#1582) Closes #1569 Co-authored-by: ST John <st@secondmind.ai> | 05 October 2020, 10:48:01 UTC |
8e0744a | st-- | 30 September 2020, 16:57:30 UTC | bump version to 2.1.2 (#1575) | 30 September 2020, 16:57:30 UTC |
65a79e4 | st-- | 30 September 2020, 16:42:05 UTC | isort for consistent imports (#1572) * GPflow now includes an isort check in its CI build. Run with `make format` / `make format-check`. | 30 September 2020, 16:42:05 UTC |
d618fe5 | st-- | 30 September 2020, 13:54:44 UTC | Likelihood quadrature cleanup (#1571) * Global constant for num_gauss_hermite_points default * QuadratureLikelihood with dependency injection (in preparation for unifying with MonteCarloLikelihood) * ScalarLikelihood as QuadratureLikelihood subclass (with _quadrature_dim/_quadrature_log_prob/_quadrature_reduction) * HeteroskedasticTFPConditional: rename argument to `scale_transform` and remove undocumented scaling * DeprecationWarning for deprecated code | 30 September 2020, 13:54:44 UTC |
12395be | st-- | 29 September 2020, 15:17:18 UTC | fix BijectorCache workaround (#1574) Follow-up to #1553 | 29 September 2020, 15:17:18 UTC |
0d3684b | joelberkeley-pio | 17 September 2020, 15:33:35 UTC | add packaging to dependencies; add 'import gpflow' check (#1568) | 17 September 2020, 15:33:35 UTC |
ad6e031 | Vincent Dutordoir | 15 September 2020, 15:37:25 UTC | Multi Latent Likelihoods using new quadrature Likelihoods (#1559) * HeteroskedasticLikelihood base class draft * fixup * cleanup * cleanup heteroskedastic * multioutput likelihood WIP * Notebook exemplifying HeteroskedasticTFPDistribution usage (#1462) * fixes * typo fix; reshaping fix * notebook showing how to use HeteroskedasticTFPDistribution likelihood * converting to .pct.py format * removed .ipynb * better descriptions * black auto-formatting Co-authored-by: Gustavo Carvalho <gustavo.carvalho@delfosim.com> * note and bugfix * add comment * Adding heteroskedastic tests (#1508) These tests ensure that heteroskedastic likelihood with a constant variance, will give the same results as a Gaussian likelihood with the same variance. * testing * added QuadratureLikelihood to base, refactored ScalarLikelihood to use it * fix * using the first dimension to hold the quadrature summation * adapting ndiagquad wrapper * merged with gustavocmv/quadrature-change-shape * removed unecessary tf.init_scope * removed print and tf.print * removed print and tf.print * Type annotations Co-authored-by: Vincent Dutordoir <dutordoirv@gmail.com> * Work * Fix test * Remove multioutput from PR * Fix notebook * Add student t test * More tests * Copyright * Removed NDiagGHQuadratureLikelihood class in favor of non-abstract QuadratureLikelihood * _set_latent_and_observation_dimension_eagerly * n_gh ---> num_gauss_hermite_points * removed NDiagGHQuadratureLikelihood from test * black * bugfix * removing NDiagGHQuadratureLikelihood from test * fixed bad commenting * black * refactoring scalar likelihood * adding dtype casts to quadrature * black * small merging fixes * DONE: swap n_gh for num_gauss_hermite_points * black Co-authored-by: ST John <st@prowler.io> Co-authored-by: gustavocmv <47801305+gustavocmv@users.noreply.github.com> Co-authored-by: Gustavo Carvalho <gustavo.carvalho@delfosim.com> Co-authored-by: st-- <st--@users.noreply.github.com> Co-authored-by: joshuacoales-pio <47976939+joshuacoales-pio@users.noreply.github.com> | 15 September 2020, 15:37:25 UTC |
799b659 | gustavocmv | 15 September 2020, 12:36:38 UTC | Scalar Likelihood Refactoring (#1566) * refactoring scalar likelihood * adding dtype casts to quadrature * black Co-authored-by: Gustavo Carvalho <gustavo.carvalho@delfosim.com> Co-authored-by: Vincent Dutordoir <dutordoirv@gmail.com> | 15 September 2020, 12:36:38 UTC |
0336b27 | st-- | 14 September 2020, 17:00:04 UTC | Merge pull request #1564 from GPflow/st/remerge_master_210 Merge master back into develop | 14 September 2020, 17:00:04 UTC |
d51c487 | st-- | 14 September 2020, 16:39:27 UTC | Merge branch 'develop' into st/remerge_master_210 | 14 September 2020, 16:39:27 UTC |
234a734 | ST John | 14 September 2020, 16:38:41 UTC | Merge remote-tracking branch 'origin/master' into st/remerge_master_210 | 14 September 2020, 16:38:41 UTC |