42 commits

Author	SHA1	Message	Date
James Hensman	803c345d44	docstring for ExpQuad (thanks Mike O. )	2014-09-17 11:12:38 +01:00
James Hensman	ff63617286	for loop speedup in grdients X	2014-09-16 13:20:38 +01:00
Max Zwiessele	4543fc3480	[link|unlink_parameter] renaming add_parameter to link_parameter	2014-09-08 08:57:28 +01:00
mzwiessele	d000893878	[ard] enhanced ard handling and plotting ... Conflicts: GPy/kern/_src/linear.py GPy/models/ss_gplvm.py	2014-08-25 09:55:58 -07:00
James Hensman	fe405271e5	gradients of predictions for Trevor	2014-08-13 10:36:54 +01:00
James Hensman	3fc603f778	more bugfix	2014-08-12 12:02:16 +01:00
James Hensman	4139397dd3	bugfix merge!	2014-08-12 11:58:28 +01:00
James Hensman	7a8a622b5d	strange bug in np.einsum fixed when using the _out_ argument (thanks T. Cohn)	2014-08-12 11:27:50 +01:00
Zhenwen Dai	785c580032	merge the current devel into psi2	2014-08-11 18:01:23 +01:00
James Hensman	df8bad0cde	fizxed a numerical bug in stationary.py	2014-08-08 11:07:15 +01:00
Zhenwen Dai	59172435e2	fix the remaining problem of cache.py	2014-06-20 21:51:51 +01:00
Zhenwen Dai	6119c33078	Support non-symmetric dL_dKmm for stationary kernel	2014-06-03 11:34:32 +01:00
Max Zwiessele	76718a8fcc	[stationary] input_sensitiviy is now 1/(l**2)	2014-05-20 14:47:28 +01:00
marahman	0c2bae53f7	Working with OU kernel	2014-05-12 18:45:23 +01:00
Max Zwiessele	5fb9ce9c53	bugfix: slicing was still in stationary somehow	2014-04-16 12:19:40 +01:00
Zhenwen Dai	b945e8d01f	[GPU] psi1	2014-03-31 16:18:06 +01:00
James Hensman	01271e3868	fixing the logexp (with MZ) and some stability issue in the stationary class	2014-03-21 15:23:49 +00:00
Alan Saul	4757265b24	Added a hack fix as suggested by max, zeroing any negative values (should really be numerically negative values on diagonal)	2014-03-14 15:23:32 +00:00
Max Zwiessele	efcce6d0af	active_dims as extra parameter for kernels, it tells which input dimensions to work on	2014-03-14 09:18:08 +00:00
Max Zwiessele	db5fd17609	slicing support for kernel input dimension	2014-03-07 16:59:41 +00:00
James Hensman	8b197b79a0	stability in stationary)	2014-02-28 17:35:00 +00:00
James Hensman	d5658660a6	einsumming in stationary	2014-02-28 13:37:26 +00:00
James Hensman	ba1d6697e5	merged	2014-02-28 12:07:19 +00:00
James Hensman	7aad39e70e	non essential tidying in stationary	2014-02-28 12:06:28 +00:00
Max Zwiessele	83f5b9377a	caching in place again and working : )	2014-02-28 11:46:09 +00:00
James Hensman	7ae9d03c45	efficiencies in stationary	2014-02-28 11:01:54 +00:00
James Hensman	5cf792504a	messing with caching	2014-02-27 10:47:27 +00:00
Max Zwiessele	65fd6dd24e	observable pattern through and thorugh	2014-02-26 15:46:14 +00:00
James Hensman	26aeb5e1db	lengthscale fixes	2014-02-26 13:37:15 +00:00
James Hensman	80acca640f	messing with kernels	2014-02-25 17:15:38 +00:00
James Hensman	3a0a192362	more efficient computations in stationary	2014-02-25 15:48:35 +00:00
James Hensman	17f9764a55	kernel tests in working order (not all implemented though	2014-02-24 17:44:12 +00:00
Max Zwiessele	2f3e0611f8	fixed stationary again	2014-02-24 14:57:05 +00:00
James Hensman	b32929a8a5	fixed stationary	2014-02-24 14:53:20 +00:00
James Hensman	06dd27c634	input senitivity in stationary	2014-02-24 14:49:20 +00:00
James Hensman	1766db89fe	adding and producting in stationary is no stationary	2014-02-24 13:51:03 +00:00
James Hensman	efd262965e	some work on periodics	2014-02-24 11:23:29 +00:00
James Hensman	88c080eece	ratquad working	2014-02-24 09:55:16 +00:00
James Hensman	ff23a59d2d	unfinished work on ratinoal quadratic kern	2014-02-24 09:41:13 +00:00
James Hensman	61a6086af6	minor fixes in kerns	2014-02-24 08:22:06 +00:00
James Hensman	2da256fa93	tidying	2014-02-21 17:39:02 +00:00
James Hensman	78ceef01c3	removed materns	2014-02-21 17:06:06 +00:00