https://github.com/GPflow/GPflow

sort by:
Revision Author Date Message Commit Date
0c3e151 Resolve merge conflict. 19 October 2018, 16:42 UTC
703ccb1 Merge branch 'develop' of github.com:GPflow/GPflow into develop 19 October 2018, 16:42 UTC
fe82d3f Release version 1.3.0 (#870) 19 October 2018, 16:38 UTC
377ff5f Merge branch 'develop' of github.com:GPflow/GPflow into develop 19 October 2018, 16:05 UTC
9587e10 Change release version to 1.3.0 19 October 2018, 15:32 UTC
80f6d6f Release v1.3.0 (#869) 19 October 2018, 13:20 UTC
63975b1 Release notes 1.3.0 (#868) 19 October 2018, 12:44 UTC
120b128 Update RELEASE.md 19 October 2018, 11:05 UTC
1707d7b Merge branch 'develop' of github.com:GPflow/GPflow into develop 06 October 2018, 18:12 UTC
3935d6b Merge pull request #854 from GPflow/hughsalimbeni-patch-1 added link to orthogonally decoupled repo 28 September 2018, 16:00 UTC
185cc8f Merge branch 'develop' into hughsalimbeni-patch-1 27 September 2018, 10:14 UTC
e8ba02e Fixing the problem with monitoring ScipyOptimizer (#856) Now lets the monitor context keep track of the optimiser to update optimiser's variables - necessary for obtaining correct traces (of hyperparameters, log-marginal likelihoods etc.) when using ScipyOptimizer. 27 September 2018, 10:14 UTC
f8deb4c clean up GPR docstrings (#859) 27 September 2018, 09:24 UTC
1196600 Merge branch 'develop' of github.com:GPflow/GPflow into develop 20 September 2018, 10:34 UTC
bb3d1f4 Clean up notebooks and documentation (#849) Minor code changes: - adds rich representation to Parameter(ized) objects so `m.as_pandas_table()` can be replaced by `m` for pretty-printing in jupyter notebooks - updates priors to have prettier printing (and more value checks) Notebooks - coreg_demo.ipynb: clean-up, typo fixes - models.ipynb: clean-up, now actually consistent with current (gpflow-1.0) architecture - kernels.ipynb: clean-up Documentation - updates various links in README.md (JH's webpage, readthedocs, tensorflow installation) - adds all currently existing notebooks to the documentation (under new section "advanced tutorials") Stand-alone .py demo scripts in doc/source/notebooks/ - makes them runnable with current gpflow version - adds "unmaintained" warning header 19 September 2018, 13:48 UTC
2a16f04 added link to orthogonally decoupled repo pending arxiv/nips 2018 publication 16 September 2018, 16:30 UTC
83631b7 Update docs_require.txt to supported tensorflow version (#853) 14 September 2018, 16:01 UTC
b94500b Merge branch 'develop' of github.com:GPflow/GPflow into develop 14 September 2018, 15:35 UTC
fa20322 Docs use supported tensorflow version 14 September 2018, 15:34 UTC
460afc7 Sample conditional returns mean and var as well as samples, and generates more than one sample (#836) 12 September 2018, 19:01 UTC
dc0b987 Merge branch 'develop' of github.com:GPflow/GPflow into develop 09 September 2018, 22:18 UTC
c8b86d3 New line in docstrings of Stationary kernels 09 September 2018, 22:18 UTC
2475891 fix active_dims=slice (#840) Fixes #839 05 September 2018, 15:12 UTC
12c7656 add link to github.com/Joshuaalbert/heterogp in lieu of #735 (#773) * add link to github.com/Joshuaalbert/heterogp in lieu of #735 * remove link to gpflow-monitor now that it's in GPflow itself, ok'd by @markvdw 31 August 2018, 17:57 UTC
a2afcee RBF Cross-Covariance (#754) Adds RBF cross covariance in expectations. Previous implementation defaulted to quadrature expectations for cross co-variance when kernel and feature were not the same. 30 August 2018, 09:23 UTC
0f6a5d3 Added MixedKernelSeparateMof (#830) Adds a new feature MixedKernelSeparateMof (SharedIndependentMof subclass) for the most efficient code together with SeparateMixedMok. 29 August 2018, 22:11 UTC
fd6beaa Support broadcasting (#829) Modifies the kernels and conditionals to support broadcasting over (arbitrarily many) extra leading dimensions, similar to most of the tensorflow ops. So e.g. kern.K(X) now supports X being shape `[S, ..., T, N, D]`, and returns `[S, ..., T, N, N]` 22 August 2018, 14:13 UTC
af89b6d Cleaning up stationary kernel implementations (#827) Stationary kernels are now implemented by K_r2 as a function of the scaled square distance or K_r as a function of the scaled Euclidean distance, see the docstrings for details. scaled_square_dist and scaled_euclid_dist are deprecated, and square_dist and euclid_dist are removed. RBF is now an alias for SquaredExponential. 22 August 2018, 10:26 UTC
51e17ae Merge branch 'master' of http://github.com/GPflow/GPflow into develop 17 August 2018, 21:00 UTC
2c4cf39 Adding widedeepnetworks repository link (#825) * Adding repository link * Correction to table 17 August 2018, 19:05 UTC
66aa1f9 Extended ndiagquad docstrings (#823) 16 August 2018, 10:30 UTC
e41d8f3 Bugfix control dependencie in likelihoods (#821) Fixes a bug with control dependencies in likelihoods. 13 August 2018, 23:08 UTC
d9624ae remove references to six module (#816) GPflow no longer supports python 2 and hence references to the six module are redundant. 06 August 2018, 09:49 UTC
54a276a Parameter.trainable was not properly a property, and param.trainable = False did not work. Fixed by this commit. (#814) 19 July 2018, 13:19 UTC
f746253 Fix bug in ndiag_mc re multi-dimensional kwargs (#813) 17 July 2018, 12:58 UTC
1db48f3 Update version.py file to 1.2.0 (#812) 11 July 2018, 12:50 UTC
707b195 Release notes 1.2.0 (#810) 11 July 2018, 11:21 UTC
1ff0a87 Optimizer's step_callback accepts step number (#811) 11 July 2018, 10:58 UTC
b78c1a8 Fixed discovered problem with the monitor, in particular with TensorBoard tasks (#807) 10 July 2018, 11:13 UTC
495b6eb Softmax fix (#804) * PEP8. * PEP8. * Added `SoftMax.conditional_variance()`. * SoftMax test now tests against `Bernoulli` likelihood, including `predict_mean_and_var`. * super * consistency * test comment * bugfix * fix tests 28 June 2018, 12:59 UTC
45fed42 Big re-work of monitor (#792) * New monitor initial commit * Fixed the monitor notebook (checking that the saving directory exists) * Addressed PR comments. * refactored environment * Attempt to avoid using X server when testing the monitor. * Added few more tests. 28 June 2018, 11:57 UTC
1851e0d Monte-Carlo likelihoods (#799) * add @markvdw's stochastic likelihood, including the softmax * initial MC likelihood * remove MonteCarloLikelihood base class from tests * fix test * var of predict_mean_and_var and predict_density for MC likelihood * factor out MC sampling * add comment for variance bias * add tests * fixes * use same integration as for GH quadrature in MonteCarloLikelihood.predict_mean_and_var() * . * increase rtol * move to proper use of super() * move MC integration to quadrature module, similar to ndiagquad * seed to make test deterministic * add Assert for shape of Y * tidy up studentT likelihood * fix for heteroskedastic likelihoods -- requires logp to always call the Y argument Y * fix doc * add assert and equivalence tests for SoftMax * remove erroneously added file * rename "probit" to inv_probit (which is what it actually is) * add assert for num_classes to SoftMax * fix whitespace * Update RELEASE.md * Update RELEASE.md 27 June 2018, 11:19 UTC
1b0f4b0 Minibatch shape is unknow by default (#704) 26 June 2018, 10:18 UTC
9d23839 Notebook to demonstrate advanced usage such as combining GP with Neural Network (#712) * added notebook * added vanilla conv net accuracy * Tidy up notebook * ran notebook * ran notebook #2 * add untested notebooks to test_notebooks * fix bug in natural_gradients notebook * add nb_niter and nb_range to only run a small number of iterations in continuous integration notebook tests * make advanced_usage.ipynb use nb_iter() * don't add natural_gradients.ipynb and upper_bound.ipynb notebook tests in this PR * make advanced_usage run faster in CI * clean up advanced_usage * Update test_notebooks.py * fix notebook_niter * missing input_dim for CI * update notebook * another reset_default_graph_and_session call 20 June 2018, 14:59 UTC
3a16991 Removed jitter in nat grads (#768) 19 June 2018, 15:00 UTC
61088fd Speed up notebooks (#789) Notebooks are slow to execute. Additional utility functions help to detect where notebook are run and control number of iterations (optimization, standard loops and cetera), therefore minimizing spent time on running notebook integration tests. 19 June 2018, 13:19 UTC
bb08f22 Multi-output conditionals (#724) * Introduction of MultiOutputFeatures (Mof) and MultiOutputKernels (Mok). These are used to specify a particular setup of multi-output correlation. * Multiple-dispatch for conditional. This allows GPflow to select the most efficient conditional code depending on your choice of Mof and Mok. * Multiple-dispatch for Kuu and Kuf. Previously Kuu(.) and Kuf(.) were member functions of the feature class. This became cumbersome as the calculation of Kuu and Kuf also depends on the kernel used. In line with conditional we now also use multiple-dispatch to calculate Kuu and Kuf for a particular combination of Mok and Mof. * The actual maths to efficiently calculate the output-correlated conditional (credits to @markvdw ) * sample_conditional function that makes sure that the most efficient code is used to get a sample from the conditional distribution. * Minor: we updated a couple of models to use the new multi-output conditional. 18 June 2018, 17:04 UTC
6baeb43 Likelihood/students t variance scaling (#777) * Add scaling to studentT conditional variance The conditional variance of the Student’s T distributions is proportional to the square of the scale of the distribution. See https://en.wikipedia.org/wiki/Student%27s_t-distribution#In_terms_of_sca ling_parameter_σ,_or_σ2. I’ve incorporated the correct scaling factor. * explicit scale dtype and tensor broadcasting Added an explicit data type for the Student’s T scale parameter, and made the broadcasting in the conditional_variance method explicit. 15 June 2018, 10:31 UTC
916458e ndiagquad logspace feature for improved numerical stability (#747) 13 June 2018, 13:21 UTC
8fa0d57 Fix failing unittest under TensorFlow 1.8 (#788) 13 June 2018, 09:39 UTC
0e2ee24 Support Python 3.5.2 in typing checks (Ubuntu 16.04 default python3) (#787) * Support Python 3.5.2 in typing checks (Ubuntu 16.04 default python3) 12 June 2018, 19:00 UTC
147ce61 Logger in the settings (#764) 22 May 2018, 21:33 UTC
d40222d Updating readthedocs link in readme to point at master (#760) 17 May 2018, 08:58 UTC
e5d2d02 Allow `StoreSession` to take `False` for `restore_path` to prevent restoring. (#759) 16 May 2018, 12:23 UTC
bac9728 fix for Param(shape=(1,)) in ModelTensorBoard (#752) Some parameters of size 1 need to be squeezed to scalars before they can be viewed with TensorBoard. 10 May 2018, 15:52 UTC
aa98c55 quickfix for bug with params_as_tensors_for (#751) * quickfix for bug with params_as_tensors_for * now last-in first-out as it should be * fix bug in bug fix 10 May 2018, 14:24 UTC
17781c2 Inclusion of `gpflow-monitor` (#705) * Start integrating gpflow-monitor in actions framework. * Initial commit for GPfow monitor integration with actions framework. * Quick fix to TriggeredAction. * Saver works. * PrintTimings added to actions. * simple callback action added. Needs to handle session correctly * ModelTensorBoard action seems to be working. More testing needed * Exclude monitor from testing. * Check if notebook coverage is included. * Moving back to removing coverage of monitor. * changed callback action to have access to the model * yet another edit to callback action in order to get access to the context * adding type annotations and cleanning up code * changed condition in the triggered action * Notebook, small bugfixes & LmlTensorBoard. * Add very simple test for monitor. * import of tqdm if not installed is now gracefully handled. * Added seq_exp_lin. * `force_run` now first test. * Fix `force_run` bug. * Updated test. * Actions test now also tests `CallbackAction`. 09 May 2018, 18:21 UTC
f867d66 Improve docstring of Periodic kernel (#745) 08 May 2018, 19:57 UTC
20a006f Copyright notes 05 May 2018, 22:52 UTC
b9e5817 Improved quadrature for likelihoods (#736) * introduce ndiagquad to unify quadrature in likelihoods * ndiagquad that can cope with several dimensions over which to integrate * add pragma: no cover to error checking * small change of signature * fix for multi-parameter likelihoods * preliminary quadrature test * improved quadrature test * fix weight normalisation in ndiagquad * improve test * add some type annotations * more quadrature tests 04 May 2018, 15:50 UTC
38b2e90 Gauss kl method is unable to work with unknown shape of the guassian parameters 03 May 2018, 21:35 UTC
69dbeb1 Extend contributing.md 02 May 2018, 20:30 UTC
da10b35 Copy values of parameters in as_pandas_table method to make them independent from ongoing changes during optimization. 02 May 2018, 11:15 UTC
ee4f774 GPflow scipy optimizer doesn't pass options kwargs to actual scipy optimizer correctly. (#738) 02 May 2018, 09:00 UTC
3b49709 Unify calling signature of (log)densities (#721) * rename densities -> logdensities * unify ordering of (log)densites to have random variable first * adjust logdensities test * forgot to git add * fix Exponential prior * fix Exponential prior #2 * minor change of notation * move to format strings * Uniform.log_height as property * logdensities: rename y -> x 01 May 2018, 09:09 UTC
ae7ed2c Numerically stable scaled squred distance via clipping near to 0. 30 April 2018, 20:59 UTC
aae4f34 Allow the epsilon parameter of the Robustmax likelihood to be trainable (#635) * Robustmax epsilon learnable * Robustmax epsilon a parameter * eps_K1 now a property so that changes with epsilon * test for this property being up to date * updated multiclass notebook to show how it could work * . * cosmetic change * update notebook following merge conflict 30 April 2018, 17:50 UTC
4a13271 Add Exponential distribution as prior (#717) * Support exponential distribution as prior. The exponential distribution here is parametrized using the rate parameter lambda and always has its mode at 0. * Reuse exponential density for prior distribution Add transformation to test of exponential prior This ensures that it is only evaluated on the positive domain * Add sample method * fix test * fix test again * more fixes * add pragma: no cover to make codecov happy * add sample tests * sample test * revert change to densities.exponential() * fix * pr comments * fix test * removing commented code 30 April 2018, 16:35 UTC
e59ff2f add shape check for ARD hyperparameters to Kernel classes (#694) * add check of lengthscales shape to Stationary kernel * refactor to _validate_ard_shape * add casts to float * fix ArcCosine test * add None to ARD test * remove deprecated num_gauss_hermite_points from kernel * Clean up docstrings & make code a bit shorter. * Typo ARG => ARD. * update docstring * more dtype=float_type 30 April 2018, 10:44 UTC
3ff6110 returning var_list to Optimizer in order - Issue appears to affect only ScipyOptimizer. Does not affect, e.g., AdamOptimizer on doc/source/notebook/classification - Appears to be due to the fact that TensorFlow variables are not stably ordered across across replications, so var_list (and thus random initial values for optimization) is not either. 27 April 2018, 13:36 UTC
913bfe3 Minor clean-up * removing old notebooks dir * add symbolic link to notebooks 27 April 2018, 08:12 UTC
f31223c Merge pull request #731 from GPflow/jameshensman-tidy1 removing errant file 26 April 2018, 13:53 UTC
c7acc2f removing errant file 26 April 2018, 13:28 UTC
5e841b7 Model saver (#660) GPflow model saver. Supports native GPflow models and provides an interface for defining custom savers for user's models. Saver stores GPflow structures and pythonic types as numpy structured arrays and serializes them using HDF5. 19 April 2018, 14:24 UTC
2fb4b6d Merge pull request #723 from GPflow/awav/fix-pytest-import Pytest used by test_util module 17 April 2018, 14:24 UTC
26254a2 Bump patch version to 1.1.1 17 April 2018, 13:54 UTC
753aa2e pytest must be included as a dependency 17 April 2018, 13:53 UTC
8a412f6 Merge pull request #713 from GPflow/awav/github-templates GitHub PR and issue templates. 12 April 2018, 10:05 UTC
46c8784 Python code block as an example. 12 April 2018, 09:24 UTC
e3aba01 Merge branch 'awav/github-templates' of github.com:GPflow/GPflow into awav/github-templates 12 April 2018, 09:22 UTC
ec43181 Content table had broken links 12 April 2018, 09:22 UTC
ee5c897 Merge branch 'master' into awav/github-templates 12 April 2018, 09:08 UTC
eb6a269 Merge pull request #714 from GPflow/awav/optimizers-fix Optimizers incompatibility with TensorFlow >= 1.6 12 April 2018, 09:08 UTC
c5f8520 Update ISSUE_TEMPLATE.md 11 April 2018, 23:13 UTC
9ccf2b4 Update ISSUE_TEMPLATE.md 11 April 2018, 23:11 UTC
d15a401 Change tensorflow version on 1.6 11 April 2018, 22:14 UTC
745dcf8 Bump tensorflow version requirement to 1.5 11 April 2018, 22:01 UTC
5e57df5 TensorFlow optimize initilizer. 11 April 2018, 21:34 UTC
d05309e Fixes after code review. 11 April 2018, 20:20 UTC
ebc8d56 Merge branch 'awav/github-templates' of github.com:GPflow/GPflow into awav/github-templates 11 April 2018, 16:01 UTC
96e918d Update contributing.md 11 April 2018, 16:00 UTC
a6f48fa Fix typo. 11 April 2018, 15:31 UTC
1c5c7fb Update README.md Small changes inline 11 April 2018, 15:27 UTC
6d58de8 Update PULL_REQUEST_TEMPLATE.md Small changes inline 11 April 2018, 15:23 UTC
28e043c Brushing up parts of PR template. 11 April 2018, 15:10 UTC
6d5e10d Compatibility section in README.md 11 April 2018, 14:34 UTC
ea4f07d Modify contributing.md file and add github templates. 11 April 2018, 14:19 UTC
f5d8cc4 Name attribute for Gaussian likelihood. 04 April 2018, 11:10 UTC
70f18c1 allow params_as_tensors_for() to take several objects as varargs (#672) Allow params_as_tensors_for() to take several objects as varargs 04 April 2018, 10:13 UTC
455a71f Merge pull request #701 from ialong/expectations expectations: pass num hermite points to _expectation() and minor cleaning 04 April 2018, 08:27 UTC
back to top