James Hensman
8fd8288fb8
ensure_default_constraints is on by default
2013-06-17 15:57:19 +01:00
James Hensman
2929c69fce
Merge branch 'psi1_transpose_fix' into devel
2013-06-06 14:39:12 +01:00
James Hensman
5e4c24d652
psi1 is now the right way around
2013-06-06 14:38:48 +01:00
Nicolas
adf2594677
Merge branch 'devel' of github.com:SheffieldML/GPy into devel
2013-06-06 11:24:06 +01:00
Nicolas
1238252a19
removed visualize dependencies and added dKdiag_dX for linear
2013-06-06 11:24:00 +01:00
James Hensman
a5208f474e
modified lengthscale gradients - demo works now.
2013-06-06 10:59:34 +01:00
Max Zwiessele
3231e94947
kern constructors now have input_dim instead of D
2013-06-05 17:40:43 +01:00
Max Zwiessele
bfd99c3607
Merge kern conflicts in examples
2013-06-05 16:16:46 +01:00
Max Zwiessele
0490861099
kern params adapted: Nparams > num_params and fixes of input_dim
2013-06-05 16:14:30 +01:00
Alan Saul
39c242a4d5
Fixed merge conflicts, M now num_inducing
2013-06-05 15:41:48 +01:00
Alan Saul
3475b52b6c
Changed all M's for num_inducing
2013-06-05 15:29:45 +01:00
Max Zwiessele
db78b233b8
Nparams > num_params and Nparam_tranformed > num_params_transformed
2013-06-05 15:29:18 +01:00
Max Zwiessele
35c2a8b521
Refactoring: self.D > self.input_dim in kernels
2013-06-05 15:21:57 +01:00
Max Zwiessele
2e5e8ac026
REFACTORING: model names, lowercase, classes uppercase
2013-06-05 13:02:03 +01:00
Alan Saul
97f3357b6d
Replaced Q by input_dim
2013-06-05 11:17:15 +01:00
Max Zwiessele
cadf822292
plotting behaviour adapted for kern and mrd
2013-06-04 18:19:14 +01:00
Max Zwiessele
4eebf99bfb
REVERT a53690ab7f, flapack back substitued in
2013-05-31 12:09:11 +01:00
Max Zwiessele
43cd5ad50b
Merge branch 'devel' of github.com:SheffieldML/GPy into devel
2013-05-30 10:37:20 +01:00
Max Zwiessele
a53690ab7f
DeprecationWarning: Substituded all (\!) flapack occ. with lapack (scipy said so)
2013-05-30 10:37:03 +01:00
James Hensman
dc33aa1b8c
used scipy.weave to improve the speed of rbf grads
...
for large number of input dimensions with ARD, wqe have approx tenfold
speedup.
2013-05-30 09:29:26 +01:00
Nicolo Fusi
19407293dc
cross-terms
2013-05-17 16:29:32 +01:00
Nicolo Fusi
ddd3ece3ce
test
2013-05-17 12:29:13 +01:00
Max Zwiessele
ba4bd50924
stability enhancing clipping in logexp_clipped and reverse of stability clipping of parameters
2013-05-16 15:17:54 +01:00
Max Zwiessele
835632e985
WARNING: added parameter clipping for catching infinity problems [p in (-1e300, 1e300)]
2013-05-16 13:29:56 +01:00
James Hensman
542759353b
minor changes to the symmetrify function
2013-05-13 17:23:49 +01:00
James Hensman
529d7534ca
fixed the bug in stick
2013-05-13 10:26:54 +01:00
Nicolas
652b3ce2c6
unified framework for addition and product of kernels, with a tensor flag (boolean) instead of and
2013-05-10 17:48:11 +01:00
James Hensman
e60eb4e236
small changes to Brownian
2013-05-08 12:06:34 +01:00
James Hensman
ce2884f0a7
weaved linear kern
2013-05-07 18:02:10 +01:00
Max Zwiessele
ecf0dc0680
linear psi2 statistics done, all gradients working
2013-05-03 13:36:04 +01:00
Max Zwiessele
40c9790529
Mu to go
2013-05-03 10:20:29 +01:00
Max Zwiessele
7529eee5ca
Mu to go
2013-05-02 17:21:43 +01:00
Max Zwiessele
83bde47d55
merge devel into mrd > transformations added
2013-05-02 16:40:39 +01:00
Max Zwiessele
5051a2fc89
correcting linearCF, mu to go
2013-05-02 16:37:47 +01:00
James Hensman
350c69799d
fixed a tie-bug for ND
2013-05-02 10:59:19 +01:00
Max Zwiessele
42474f0044
LinearCF Psi Stat not working yet, strange bug in psi computations
2013-05-01 17:09:38 +01:00
James Hensman
83510e7515
first attemot at the new constraint framework
2013-04-30 11:25:57 +01:00
James Hensman
6dced9ac39
reimplemented caching in prod_orthogonal...
2013-04-28 22:32:37 +01:00
James Hensman
7d9352c733
weaved coregionalise. much performance gained
2013-04-28 21:37:36 +01:00
James Hensman
52ba8e4ba3
remo0ved slices from models
...
slices are now handles by special indexing kern parts, such as
coregionalisation, independent_outputs. The old slicing functionality
has been removed simply to clean up the code a little.
Now that input_slices still exist (and will continue to be useful) in
kern.py. They do need a little work though, for the psi-statistics
2013-04-28 17:22:04 +01:00
James Hensman
9fb090a508
James and Nicolos massive Yak shaving session
2013-04-26 19:32:33 +01:00
Max Zwiessele
2096d062cb
merge devel into mrd
2013-04-24 10:11:14 +01:00
James Hensman
29115b0ec1
more re-enstating of some preiov commits.
2013-04-23 17:13:43 +01:00
James Hensman
e6165e6b35
re-added indepenent_output kern
2013-04-23 17:09:52 +01:00
Max Zwiessele
d789e55481
stupid kern stash merge
2013-04-23 16:40:09 +01:00
Max Zwiessele
2b3b7350cd
kern conflict
2013-04-23 16:37:13 +01:00
Max Zwiessele
68dd416640
Merge branch 'devel' into mrd
2013-04-23 16:21:57 +01:00
Max Zwiessele
389a04d2b5
bugfix: cross term psi1 bias + linear
2013-04-23 16:21:41 +01:00
Max Zwiessele
f01be172be
moved *2. of psi2 statistics into kern and corrected bias+linear cross term
2013-04-23 15:22:30 +01:00
Max Zwiessele
0c8b83454f
Revert "merge devel mrd"
...
This reverts commit 3f625a9347 , reversing
changes made to dc6faeb303 .
2013-04-23 14:02:15 +01:00