Siivola Eero
8ad774cb4e
Fixed two small lbugs
2017-07-04 13:18:06 +03:00
Akash Kumar Dhaka
8377d95fbe
putting exact_inference_gradients again and calling in instead from ep_gradients which is the genreic method for calculating expected gradients used in ep
2017-07-03 18:33:21 +03:00
Akash Kumar Dhaka
ed3f6c56c2
commiting code for some changes to api for calculating ep_gradients, also fixing some issues with gaussian hermite quadrature, no we have both avaialable ...
2017-07-03 18:24:02 +03:00
Akash Kumar Dhaka
51dcd85412
adding file for gaussian-kronrod integration, test cases and calculating gradients of log marginal wrt theta-likelihood params
2017-07-03 13:30:39 +03:00
Moreno
0c248e7520
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
2017-03-22 13:40:06 +00:00
mzwiessele
6cd13ac2b3
fix: Fixed numpy 1.12 indexing and shape preservation
2017-02-23 14:45:18 +00:00
Max Zwiessele
589c5d9eac
Merge branch 'devel' into binomial_laplace
2016-09-01 14:21:27 +01:00
Aki Vehtari
c3963928f1
Python2->Python3
2016-06-09 14:50:53 +03:00
Alan Saul
378ac0ad06
Added binomial derivative and test
2016-04-01 21:21:37 +01:00
Max Zwiessele
ac878f278f
Merge branch 'jameshensman-master' into gpy_one_fixes
2016-04-01 11:56:41 +01:00
Alan Saul
ebf98328fe
Merge branch 'devel' into fixing_likelihoods
2015-11-12 13:36:23 +00:00
Alan Saul
51ccaed020
Started looking at quadrature code for moments
2015-10-22 16:23:08 +01:00
Alan Saul
e7c8da2cdf
Added Y_metadata to moments_match_ep, and removed log-concave student-t test, and added EP test for bernoulli
2015-10-22 16:20:16 +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
mzwiessele
dce82847a7
[paramz] fully integrated all tests running
2015-10-15 14:59:57 +01:00
James Hensman
ff9ab62db2
ENH improved the stability of variational_expectations in the Binomial likelihood
2015-09-25 11:16:12 +01:00
Max Zwiessele
8132084de6
[ep] now calling exact inference instead of copying code
2015-09-11 15:47:07 +01:00
Max Zwiessele
1d354f5cce
[classification] sparse gp classification and dtc update
2015-09-11 15:08:30 +01:00
Alan Saul
ef778f75d4
Updated sampling and plots to be correct shape, and changed plotting of sampling to be posterior samples p(f*|f), like it used to be, and samples_y to plot samples of p(y*|y)
2015-09-10 11:08:28 +01:00
Mike Croucher
d3321251ef
Switched to scipy.special.log1p@
2015-09-07 16:34:53 +01:00
Mike Croucher
54ea44b9f9
Merge branch 'devel' of https://github.com/SheffieldML/GPy into devel
2015-09-07 16:23:19 +01:00
Mike Croucher
fb444d8933
Used scipy.log1p since it gives more consistent results cross-platform
2015-09-07 16:23:10 +01:00
Mike Croucher
c47c63b408
Python 3 fixes
2015-09-07 14:07:09 +01:00
Ricardo
2745904c01
Shape of heteroscedastic variance corrected
2015-09-03 17:28:31 +01:00
Max Zwiessele
26bdcfa82e
[hetnoise] import correction
2015-09-02 15:45:55 +01:00
Max Zwiessele
74a7c6398b
[merge]
2015-09-02 14:34:23 +01:00
Ricardo
f780693577
New likelihood: HeteroscedasticGaussian
2015-09-02 13:29:33 +01:00
Max Zwiessele
70a9a26d7e
[core] updating system, security branching
2015-09-02 09:06:17 +01:00
Max Zwiessele
938cc49aed
Merge branch 'devel' of github.com:SheffieldML/GPy into devel
2015-08-28 16:28:25 +01:00
Max Zwiessele
4bd99c674f
[heteroscedastic gauss] Implemented Heteroscedastic Guassian Lik with @ric70x7
2015-08-28 16:28:01 +01:00
Max Zwiessele
0d32652c88
[heteroscedastic gauss] Implemented Heteroscedastic Guassian Lik with @ric70x7
2015-08-28 16:26:55 +01:00
James Hensman
60b5f071b4
sqrt(pi) term fix in Bernoulli
2015-08-18 17:11:54 +01:00
Alan Saul
da5de6b814
Added fallback for non-identity link function for Gaussian variational expectations
2015-08-18 16:29:44 +01:00
James Hensman
9780fb48de
fixing qualtile code for some likelhoods
2015-08-13 12:07:01 +01:00
Mike Croucher
dd059502ac
Python 3 fixes
2015-08-12 08:37:13 +01:00
Alan Saul
926b53bfce
Changed quantile computation via sampling and added fallback for predictive mean and variance if conditional mean and variance are not implemented yet
2015-07-22 18:33:11 +01:00
Alan Saul
a1c9168de4
Passing on y_metadata
2015-06-25 15:18:08 +01:00
James Hensman
24e759199e
sqrt pi term was missing in variational expectations
2015-06-12 16:58:29 +01:00
Alan Saul
56d50e0469
Reshaped log predictive density to have D outputs
2015-05-22 21:16:38 +01:00
Alan Saul
1acbcd2593
Fixed deg free gradient
2015-04-28 10:51:13 +01:00
Alan Saul
75ebe4bf10
Added log predictive density, student t degrees of freedom gradients and plotting functionality
2015-04-27 18:56:20 +01:00
Alan Saul
7abeed51de
Quadrature appeared to be out by a factor of 1/sqrt(pi)
2015-04-23 17:00:14 +01:00
James Hensman
e88b8a88d1
manual merging with AS
2015-04-16 12:45:04 +01:00
Alan Saul
361f0a5274
Fixed log predictive density, added option for LOO to provide some intemediate variables
2015-04-16 09:25:18 +01:00
Alan Saul
ab5f359103
Changed LOO implementation for Eq 30 instead of 37
2015-04-14 16:27:58 +01:00
Alan Saul
b6761c21d7
Added LOO for laplace and exact inference on training data, Gaussian logpdf appeared to be wrong, now fixed
2015-04-14 14:17:08 +01:00
Alan Saul
5c9587404d
Added some numerical stability to link functions with tests for link functions
2015-04-10 14:58:02 +01:00
James Hensman
ea787fd376
manual merging
2015-04-09 15:46:40 +01:00
James Hensman
1e30ffd730
speed ups for normal cdf
2015-04-09 15:42:02 +01:00
Mike Croucher
82a9bb9d93
Merge remote-tracking branch 'upstream/devel' into devel
2015-04-09 14:21:15 +01:00
