apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-04-27 13:56:23 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions
530 commits 38 branches 52 tags 68 MiB
Commit graph

530 commits

This branch
This branch
All branches
Author SHA1 Message Date
Nicolo Fusi
11dacb378a added autodection of Rasmussen's minimize 2012-12-14 13:57:29 +00:00
Nicolo Fusi
dd7f3a2711 Merge branch 'master' into SGD 2012-12-10 17:28:19 +00:00
Nicolo Fusi
da51f69ec3 Merge branch 'master' of github.com:SheffieldML/GPy 2012-12-10 17:26:05 +00:00
Nicolo Fusi
8c4606355d fixed SGD to work with new interface 2012-12-10 17:25:54 +00:00
Nicolo Fusi
c9358c6aec fixed merge 2012-12-10 17:23:20 +00:00
Nicolo Fusi
e2770b01bb models are now pickleable 2012-12-10 17:20:45 +00:00
Nicolo Fusi
5e7ca787a5 working implementation of SGD 2012-12-10 17:15:08 +00:00
James Hensman
053ebbe2d4 t of the gradients are now working in the uncollapsed sparse GP: one 
term still to do.
 2012-12-09 14:41:36 -08:00
Neil Lawrence
493f236b90 Added notes on issues found. 2012-12-09 01:53:42 -08:00
James Hensman
ef26888b78 dL_dbeta now works in the uncollapsed sparse GP 2012-12-08 15:29:17 -08:00
James Hensman
5f2c2a04f8 chenged a little of the notation in the sparse GP 
This should allow easier implementation of het. noise
 2012-12-08 15:15:31 -08:00
James Hensman
ab56f3f9b7 small tweak to the gradients in sparse GP 2012-12-07 23:03:53 -08:00
James Hensman
e6d08348c7 minor perfomance tweak for GP regression 2012-12-07 17:18:52 -08:00
James Hensman
fe13d2c09a more skeletal work on the uncollapsed GP 
None of the gradients work, but lots more things are in place
 2012-12-07 16:35:15 -08:00
James Hensman
1c0a223329 added the raw_predict function in the uncollapsed sparse GP 2012-12-07 15:51:27 -08:00
James Hensman
4fb28dcbd4 Added get and set attributes to the mode class 
... so that we can deal with the parameters in a Neil friendly way.
 2012-12-07 15:39:47 -08:00
Nicolo Fusi
e415dcdd93 GPLVM accepts an initial value for X (in case you don't want to use the default random/PCA init) 2012-12-07 15:16:52 +00:00
Nicolo Fusi
b7021261ea now passing a reference of the model to the optimizer (used in SGD) 2012-12-07 15:15:59 +00:00
James Hensman
50029ca4a7 implemented default constraints 
via m.ensure_default_constraints()
 2012-12-06 11:46:34 -08:00
James Hensman
ac84f4f34c changed the name of GP_EP (from simple) in the unit test, added a messages option for full EP 2012-12-06 11:30:12 -08:00
James Hensman
9af637e214 changes the naming of kern components 
Kern components now only get a number if their name is duplicated
 2012-12-06 11:28:04 -08:00
James Hensman
f634350398 fixed small bugs in linalg, setup.py 2012-12-06 10:16:19 -08:00
James Hensman
c72d1b1a5b changed the behaviour of pdinv: now returns L and Li as well as the inverse 2012-12-06 09:51:13 -08:00
James Hensman
574f9f4e0a more tidying in EP, removed examples from _module_ ( and opened discussion on github 2012-12-06 09:21:37 -08:00
James Hensman
69cc506b9e trivial merge resolution 2012-12-06 08:49:46 -08:00
Nicolo Fusi
f2d6a8a171 fixed bug in rbf_ARD kernel (dK_dX) 2012-12-06 12:38:15 +00:00
Nicolo Fusi
fb8cfc8f34 fixed import error for examples and import error for uncertain inputs GP 2012-12-06 12:35:29 +00:00
Nicolo Fusi
a0f758bb69 Merge branch 'master' of github.com:SheffieldML/GPy 2012-12-06 12:32:39 +00:00
Nicolo Fusi
a0df883ec9 fixed bug in GP_regression 2012-12-06 12:32:31 +00:00
James Hensman
c8e2ca351d some tidying in the EP code 2012-12-05 22:45:08 -08:00
James Hensman
20a02cc445 removed uncertain gp regression from the model __init__, since it's now 
just a switch in the sparse GP
 2012-12-05 22:25:37 -08:00
Neil Lawrence
60fd1c55dc Merge changes for model.py and optimization.py on comments. 2012-12-05 19:22:36 -08:00
Neil Lawrence
0ee11a2077 Added datasets.py back in and minor changes. 2012-12-05 19:19:15 -08:00
James Hensman
c003b2f34d Minor edits 2012-12-05 09:10:27 -08:00
James Hensman
4dcf2b85ce models can now specify a preferred optimser (defaults to tnc) 2012-12-04 13:10:17 -08:00
James Hensman
1a9872c187 Some tidying in the uncollapsed GP 2012-12-03 10:53:37 -08:00
James Hensman
5f22bd00e6 made uncertain inputs a simple swith in the sparse GP class. This simplifies the inherritance structure 2012-12-03 18:35:05 +00:00
James Hensman
b1c2282bfc Merge branch 'master' of github.com:SheffieldML/GPy 2012-12-03 16:38:12 +00:00
Nicolas
049de98d3b rbf_ARD now in the updated format for the computation of the derivatives (included for the psi-statistics, but not tested) 2012-12-03 11:26:10 +00:00
Nicolas
908b1d2801 Merge branch 'master' of github.com:SheffieldML/GPy 2012-12-03 10:26:53 +00:00
Nicolas
3edd867ece GPy: Some rewriting for the exponential and Matern kernels. They now pass the unit test. 2012-12-03 10:26:28 +00:00
Nicolo Fusi
2db9e7bb57 Merge branch 'master' of github.com:SheffieldML/GPy 2012-12-03 10:08:25 +00:00
Nicolo Fusi
8b73dafbae added precomputation of linear kernel, changed the logic a bit 2012-12-03 10:08:12 +00:00
James Hensman
a5b90fa541 Added a skeleton of the uncollapsed sparse GP 2012-12-02 15:50:16 +00:00
James Hensman
c4761bedf5 minor commenting 2012-12-02 15:12:14 +00:00
James Hensman
1d8104bcc2 general tidying in sparse_GP_regression 
Have also made a small ampount of headway in
enabling heteroscedatic noise.
 2012-12-02 13:40:19 +00:00
James Hensman
365d069ae8 saved a little computation by exploiting the symmetry of a matrix 2012-12-02 12:35:54 +00:00
James Hensman
1c60e50fed I think the gradients bug in the sparse GP model is due to Kmm being unstable to invert. REducing M in some of the examples really helps 2012-12-02 12:32:20 +00:00
Nicolas
d71ad99db9 derivatives of the exponential kernel in the right format 2012-11-30 17:39:04 +00:00
James Hensman
fb21a3589b added sympykern as a 'kernpart' object. 
now we can add sympykerns to any other kern
 2012-11-30 17:32:32 +00:00