575 commits

Author	SHA1	Message	Date
Zhenwen Dai	f11090ad8a	improve the documentation for LVMOGP	2017-11-23 09:20:05 +00:00
Zhenwen Dai	c294cf715f	the implementation of SVI-MOGP	2017-09-19 16:28:36 +01:00
Moreno	e572bfb746	Basic framework for serializing GPy models	2017-09-11 12:18:01 +01:00
Siivola Eero	a9cdfad4b6	Fixed two small lbugs	2017-08-10 11:14:40 +03:00
Akash Kumar Dhaka	8b621a409c	commiting code for some changes to api for calculating ep_gradients, also fixing some issues with gaussian hermite quadrature, no we have both avaialable ...	2017-08-10 11:14:40 +03:00
Akash Kumar Dhaka	ae27e1225d	adding file for gaussian-kronrod integration, test cases and calculating gradients of log marginal wrt theta-likelihood params	2017-08-10 11:14:40 +03:00
Akash Kumar Dhaka	69a81acff9	bug fix to prevent infinite loop because of incorrect stopping condition	2017-08-10 11:14:40 +03:00
Moreno	255d97a917	Refactor EP and EPDTC	2017-08-10 11:14:40 +03:00
esiivola	7c4e2f21f8	Added initial version of the refactored EP	2017-08-10 11:14:40 +03:00
esiivola	0d06dc3b4e	Added initial version of the refactored EP	2017-08-10 11:14:40 +03:00
esiivola	dfc5bd42dc	Added initial version of the refactored EP	2017-08-10 11:14:40 +03:00
Moreno	63751de912	Changes in EP/EPDTC to fix numerical issues and increase the flexibility of the inference. ... Changes to avoid numerical issues and improve the performance: - Keep value of the EP parameters between calls - Enforce positivity of tau_tilde - Stable computation of the EP moments for the Bernoulli likelihood - Compute marginal in the GP model without directly inverting tau_tilde Changes to improve the flexibility: - Add parameter for maximum number of iterations - Distinguish between alternated/nested mode - Distinguish between sequential/parallel updates in EP	2017-06-01 11:59:04 +03:00
mzwiessele	96cc0838eb	fix: Fixed numpy 1.12 indexing and shape preservation	2017-06-01 11:59:04 +03:00
Thang Bui	df7c7539f9	added tests	2016-10-27 14:39:32 +01:00
Thang Bui	eee1fba8ad	Added pep.py -- Sparse Gaussian processes using Power Expectation Propagation ... This allows interpolation between FITC (EP or alpha = 1), and Titsias's variational (VarDTC, VFE when alpha = 0). Reference: A Unifying Framework for Sparse Gaussian Process Approximation using Power Expectation Propagation https://arxiv.org/abs/1605.07066	2016-10-19 15:56:10 +01:00
Max Zwiessele	dc9a67ad74	[merge] into new devel	2016-09-05 15:56:53 +01:00
Max Zwiessele	896887bd94	[inference] rename wrong precision into variance	2016-09-01 13:16:50 +01:00
kcutajar	1810549b60	Making consistent with python 3	2016-05-10 01:56:59 +02:00
kcutajar	3320146eec	Fixed incorrect import	2016-05-09 23:30:36 +02:00
kcutajar	5b7566dfac	Removed erreneous lines indicating merge conflicts	2016-05-09 16:21:10 +02:00
kcutajar	debdb545dc	Fixed conflicts	2016-05-09 15:57:53 +02:00
kcutajar	bcbc58e065	Added fixes to repo + rebased	2016-05-09 15:48:17 +02:00
kcutajar	83bd94185d	Removed SSM functionality - updated Kronecker grid case	2016-05-09 15:48:17 +02:00
kcutajar	3d346cbdd6	Added kernels for GpGrid and GpSsm regression	2016-05-09 15:46:51 +02:00
kcutajar	9a2670b98e	Added core code for GpSSM and GpGrid	2016-05-09 15:45:20 +02:00
kcutajar	8e63472237	Removed SSM functionality - updated Kronecker grid case	2016-05-01 21:50:34 +02:00
mzwiessele	61dff6e62b	[fix #380] reloading ep	2016-04-27 12:41:24 +01:00
mzwiessele	e4d76a133a	[fix #380] reloading ep	2016-04-27 12:37:04 +01:00
Max Zwiessele	0116b03f3c	[plotting] still testing the testing	2016-03-31 14:06:23 +01:00
kcutajar	08fa69fdd1	Added kernels for GpGrid and GpSsm regression	2016-03-22 09:09:30 +01:00
kcutajar	ea2f814d55	Added core code for GpSSM and GpGrid	2016-03-22 08:53:39 +01:00
Zhenwen Dai	67ba9b60c6	move _raw_predict into posterior object	2016-03-17 11:29:33 +00:00
Zhenwen Dai	f2b813551a	bug fix for mcmc sampler and add test case	2016-03-10 18:37:53 +00:00
Zhenwen Dai	ba74e29aee	merge with upstream	2016-03-10 18:17:35 +00:00
mzwiessele	e4a4344334	[stochastics] update for new stochastic iptimizers in gpy	2016-03-08 09:47:24 +00:00
mzwiessele	b1e073318f	[chaching] changing all chacher limits to 3	2016-03-07 11:37:22 +00:00
Zhenwen Dai	00e4ac152a	merge GPy upstream	2016-02-23 13:52:45 +00:00
Max Zwiessele	98daaba57e	[plotting] xlim setting	2016-02-19 14:55:36 +00:00
Zhenwen Dai	8b279175c5	fix the issue of negative prediction variance of normal GP	2016-01-21 11:22:57 +00:00
Max Zwiessele	f68001a84e	[vardtc] these two lines are overridden by the next two lines...	2016-01-12 12:33:42 +00:00
Zhenwen Dai	c0e6978054	merge with current GPy devel	2015-12-09 17:27:06 +00:00
Alan Saul	3d90b1b3ff	Documentation	2015-11-20 15:52:19 +02:00
Alan Saul	02af482585	Removed old code	2015-11-20 15:50:01 +02:00
Alan Saul	efa9b920c3	Allow EP to have a auto reset option	2015-11-20 15:06:20 +02:00
Alan Saul	ebf98328fe	Merge branch 'devel' into fixing_likelihoods	2015-11-12 13:36:23 +00:00
Max Zwiessele	913d80f712	[optimization] deleted and backwardscompatible	2015-11-09 13:27:24 +00:00
Max Zwiessele	850c10beaa	[pickle] load errors bc of kernel changes, backwards compatibility issues fixed	2015-11-09 10:09:07 +00:00
Zhenwen Dai	2bf1be5500	bug fix for lbfgs with max_iters big than 15000	2015-11-02 21:18:56 +00:00
Alan Saul	3dba1c05de	Added Z_tilde contribution for EP, now log_marginal is correct, need to check for var_dtc case	2015-10-22 16:19:27 +01:00
Alan Saul	5b4abf4c34	Fixing bernoulli likelihood for Laplace, fixing Zep for EP, and starting working on quadrature limits	2015-10-19 19:29:57 +01:00