CHANGELOG

17.3.2011 (OS)
* Removed all code from __init__.py and instead used gp_base ,covar_base etc. as module
* Moved linear algebra into seprate module
* Added working gplvm demo
* Moved the demo code in gpr to main, outside the extra function
  - gradX has changed (TODO: Max. Change the time shift code to reflect this).
* Added bounds to optimization
* Exponetiation of hyperparameters is now out of the code.
  The only placed that deside upon exp. are the covariance and the priors.
* Changed lnpriors adding Exp prefix that indicates that parameters are in log space
* Speedup for cholesky
* Added Neils code for cholesky with Jitter
* Changes to the covaraince interface
* Renamed lml -> LML
* Renamed dlml -> LMLgrad
** logtheta -> theta
Parmaeters are now jus parameters of the covariance that decides internally how to handle them.
** renamed functions
  K(theta,x,[x2])
  Kdiag (x)
  Kgrad_theta(x,i)
  Kgrad_x(x,[x2],d)
  Kgrad_xdiag(x,d)
** Removed _pointwise_ditance from Covariance 
Max: why do you need pointwise_distance? 

* 23.23.2011 (OS)
Added an interace for likelihood models to get the noise out of the covariance
For now both interfaces are supported. demo_gpr shows how to use the new interface.

* 25.03.2011
Corrected use of interval_indices. 
Interval_indices are the indices on which the gp will predict.