https://github.com/GPflow/GPflow

sort by:
Revision Author Date Message Commit Date
1076e47 Fixes #161 (#164) * Fixes #161 * pep8ified test_likelihoods 15 August 2016, 11:10 UTC
b3dab02 use batch ops in KLs and conditionals (#159) * adding a simple Uniform prior * fixed sign bug * removed python loops from KL code * removed python loop from conditionals * removing loops from conditionals.py * removed loops form vgp * removing dead code 10 August 2016, 13:07 UTC
3cc5604 adding coregion kernel (#139) * adding coregion kernel * tests for Coregino kernel * slight iprovements to Coregion docstring 10 August 2016, 09:59 UTC
57a6c69 adding a simple Uniform prior (#146) * adding a simple Uniform prior * fixed sign bug 10 August 2016, 09:38 UTC
5726dd4 Fix covariance overparameterisation (#150) GPflow often optimizes positive-definite matrices. To maintain positive-definiteness without constrained optimization, a lower-triangular matrix is optimized. Sigma + L L ^T The previous approach to optimizing L was to ignore the upper half. The mean that there were some extra variables in the optimization vector, which did nothing. This PR implements a tensorflow op which transforms back-and-forth between triangular matrix L and a 'packed' vector representation. The result is that there are no redundant parameters in the optimization vector. 09 August 2016, 18:39 UTC
9c2faf6 add Apache headers (fix #135) (#157) 09 August 2016, 09:41 UTC
87dad9f Fixed optimiser options to be consistent with scipy (#153) * Fixed optimiser options to be consistent with scipy. * Bumped version & added deprecation warnings. 09 August 2016, 09:21 UTC
3f6f063 Get tests working with Python 3. (#152) - Added relative imports (.reference) - Changed all failIf/Unless to the non-deprecated assertTrue/False. 02 August 2016, 11:06 UTC
2d98580 Merge pull request #151 from GPflow/active_dims_pickle_bug make kernels with active_dims pickleable 01 August 2016, 16:53 UTC
1da8382 Merge branch 'master' into active_dims_pickle_bug 29 July 2016, 19:38 UTC
bd0b423 make kernels with active_dims pickleable 29 July 2016, 19:27 UTC
e155d76 Track logprobs (#136) * added code for returning log-densities from HMC * added test code for log-prob tracking * adding access to logprobs from model class 29 July 2016, 09:25 UTC
6fa46bb enables pickling with fixed parameters (#141) 28 July 2016, 16:30 UTC
e0719ed some small edits to keep GPflow working with python 3 (#137) * python3.5 compat in test_likelihoods * py35 compatibility in autoflow 28 July 2016, 16:12 UTC
b4a8d1b made transforms all pickleable (#134) 27 July 2016, 14:40 UTC
9497d34 A dummy commit to trigger codecov.io 24 July 2016, 12:05 UTC
ddd0c3f Merge pull request #130 from GPflow/autokill_autoflow Autokill autoflow 13 July 2016, 08:04 UTC
985946b extending test to cover simplified autoflow killing 12 July 2016, 21:52 UTC
486775c some simplification of kill_autoflow 12 July 2016, 21:45 UTC
874957c Merge pull request #89 from GPflow/get_param_index a simple function to work out where a parameter is on the free_state 11 July 2016, 11:01 UTC
ce66d24 Merge branch 'master' into get_param_index 11 July 2016, 10:39 UTC
28fb48a Merge pull request #128 from GPflow/jitter_for_samples adding slight jitter to cholesky when computing samples 11 July 2016, 10:38 UTC
896ed83 Merge branch 'master' into get_param_index 11 July 2016, 08:36 UTC
31b48a1 Merge branch 'master' into jitter_for_samples 11 July 2016, 08:27 UTC
02286ca Merge pull request #127 from GPflow/softplus using the built-in softplus function for simplicity 11 July 2016, 08:27 UTC
30c2190 adding slight jitter to cholesky when computing samples 10 July 2016, 20:37 UTC
7172bd3 using the built-in softplu function for simplicity 10 July 2016, 07:01 UTC
691b20b minor edits to tests 09 July 2016, 13:25 UTC
3235b0d changes to regression ipynb 09 July 2016, 11:59 UTC
5ae3f60 minor edits in param.py - get_sample_dict became get_samples_df (since it retuns a datafram - set_parameter_dict no longer 'pops', resulting in an empty dict 09 July 2016, 11:57 UTC
1388d9f Merge branch 'master' into get_param_index 09 July 2016, 11:31 UTC
dc5ac86 Correction to VGP from @fujiisoup (#126) Correction to VGP from @fujiisoup. See also #125 09 July 2016, 10:54 UTC
e0b2217 Merge pull request #123 from GPflow/lognormal Lognormal 09 July 2016, 09:09 UTC
73488dc Merge branch 'master' into lognormal 09 July 2016, 08:59 UTC
9d005bc Merge pull request #122 from GPflow/svi-example Extended svi_test.ipynb to show how to use the stochastic optimizer 08 July 2016, 15:18 UTC
f959f29 improvements to svi notebook 08 July 2016, 12:00 UTC
3722e10 correction to lognormal test 08 July 2016, 11:28 UTC
4f21e67 adding a lognormal prior 08 July 2016, 11:27 UTC
5faf578 Extended svi_test.ipynb to show how to use the stochastic optimizer. 08 July 2016, 10:54 UTC
b331313 Merge branch 'master' of github.com:GPflow/GPflow into get_param_index 08 July 2016, 10:02 UTC
62972a2 Fixed tests and small bug. 07 July 2016, 11:49 UTC
57876ab Merge pull request #120 from GPflow/lower_bound_transforms improvements to transforms 07 July 2016, 06:52 UTC
3f71351 improvements to transforms 06 July 2016, 17:33 UTC
dea3e94 get_samples_dict now returns a pandas DataFrame object. 06 July 2016, 12:55 UTC
6253621 Merge branch 'master' of github.com:GPflow/GPflow into get_param_index 06 July 2016, 11:11 UTC
de38ef7 Merge pull request #108 from GPflow/Placeholder Placeholder 01 July 2016, 12:31 UTC
ab157e8 testing samples_dict with fixed param 30 June 2016, 09:27 UTC
e8d7432 a simple test for get_samples_dict 30 June 2016, 09:24 UTC
45bcb9c improvements to the regression notebook 30 June 2016, 09:13 UTC
8a58c57 a new function to make an array of samples into a dictionary of the same form as get_param_dict 30 June 2016, 08:53 UTC
5be1aa6 improved docstring 30 June 2016, 08:31 UTC
6a0dc6e trivial merge 30 June 2016, 08:27 UTC
deaab00 some improvements to update_data notebook 29 June 2016, 19:53 UTC
e425bbc removing more notebooks 29 June 2016, 19:20 UTC
d2c16d9 removing ipynb_checkpoints 29 June 2016, 19:15 UTC
ca32ff2 small improvements to test_param 29 June 2016, 17:06 UTC
e8b04bc trivial merge 29 June 2016, 16:11 UTC
804435b Merge pull request #112 from GPflow/to_dict `to_dict` 29 June 2016, 15:24 UTC
86a4ac5 Merge branch 'master' into to_dict 29 June 2016, 14:12 UTC
70ca464 Merge pull request #114 from GPflow/PabloLeon-readmeTypos Update README.md. Thanks @PabloLeon 29 June 2016, 14:11 UTC
8a32f66 Update README.md 29 June 2016, 13:55 UTC
1a5fba1 improving tests for models that should recompile 29 June 2016, 13:50 UTC
c140477 correcting no-cover statement 29 June 2016, 13:45 UTC
4505df2 removing exectuable bit for test_data_object 29 June 2016, 13:28 UTC
c050115 removing debug statements 29 June 2016, 12:51 UTC
0e1aa36 removing remnanat of calc_feed_dict 29 June 2016, 12:48 UTC
1e6d90c Merge branch 'master' into to_dict 29 June 2016, 12:27 UTC
5c1a44a Merge pull request #113 from GPflow/kern_slicing Kern slicing 29 June 2016, 12:18 UTC
4bd0764 lots of new simple tests for DAtaHolder 29 June 2016, 10:57 UTC
c5a1306 Updating RELEASE, _version, README incrementing version to 0.2.0 adding @fujiisoup to README detailing new features in RELEASE 29 June 2016, 10:37 UTC
6bb7e15 minor edits to docstrings 29 June 2016, 10:29 UTC
dfed3ca Merge remote-tracking branch 'origin' into Placeholder 29 June 2016, 10:08 UTC
8e008d7 adding a string function for the DataHolder Class2 29 June 2016, 09:45 UTC
8ed17d0 silly bug fix 29 June 2016, 09:40 UTC
7905bb6 Changing data now only recompiles if necessary An option in the dataholder class lets it choose what to do when the shape of data are changed. So now we can do X = np.random.randn(10, 1) Y = np.sin(X) m = GPflow.models.GPR(X, Y, GPflow.kernels.RBF(1)) m.optimize() X_new = np.random.randn(12, 1) Y_new = np.sin(X_new) m.X = X_new m.Y = Y_new m.optimize() # no recompile neded. 29 June 2016, 08:36 UTC
61df1f1 some pep8 changes in test_kerns.py 28 June 2016, 19:00 UTC
4d628f1 Small improvements to kernel slicing Kernels with active dimensions now select columns of data using tensorflow's gather, rather that a packed series of scalar indexes. old: tf.pack([X[:,i] for in in active_dims]) new: tf.transpose(tf.gather(tf.transpose(X), active_dims)) the latter is preferred because the graph is much smaller, especially for a large nuber of active dimensions. In addition, the kernel tests have been tidied and extended to cover more of kernels.py. 28 June 2016, 18:53 UTC
6d99d4c Fixed test. Also works with python3 now. 28 June 2016, 18:03 UTC
4148cd9 Merge branch 'master' of github.com:GPflow/GPflow into to_dict 28 June 2016, 17:33 UTC
1cb047c Merge pull request #109 from GPflow/multiclass Multiclass 28 June 2016, 12:28 UTC
abf3015 Merge branch 'multiclass' of https://github.com/GPflow/GPflow into multiclass 28 June 2016, 12:04 UTC
e615c3c Correcting comment. 28 June 2016, 12:04 UTC
18bddea fixed typos Some small typos fixed 28 June 2016, 10:51 UTC
568ff1a Building up multiclass demo. 27 June 2016, 15:11 UTC
1acef30 Building up test. 27 June 2016, 13:43 UTC
aeb4de9 docstring improvements 27 June 2016, 13:34 UTC
7f4fae9 Building up likelihood tests. 27 June 2016, 13:31 UTC
a7c6473 changed the order of the classe and improved docstrings 27 June 2016, 13:23 UTC
6516d0e Building up multiclass tests. 27 June 2016, 13:21 UTC
7a54ddf Correcting bug. 27 June 2016, 13:14 UTC
b91294e Correcting comment. 27 June 2016, 13:02 UTC
881f905 Removing partial softmax functionality because it is confusing. 27 June 2016, 12:59 UTC
194aaa9 Building up test likelihoods. 27 June 2016, 12:07 UTC
a208969 Tidying up likelihood test. 27 June 2016, 12:00 UTC
eac7334 Changes to likelihood tests. 27 June 2016, 11:17 UTC
050bf0d Correcting an added bug and reimplimenting clipping. 27 June 2016, 11:16 UTC
1eb9707 Some additional clipping. 27 June 2016, 11:03 UTC
f677790 Test added for different size. 27 June 2016, 10:20 UTC
a4d2d04 Some documentations prepared. 27 June 2016, 10:02 UTC
2c3e80e Removing clipping. 27 June 2016, 09:31 UTC
back to top