##change / RobustGaSP 0.5.7
 *Implement the Parallel partial Gaussian stochastic process through the ppgasp() function. Implements the ppgasp-class, predppgasp-class ppgasp.predict(), show.ppgasp().
 *Include the data set humanity_model.rda. The document of the ppgasp()  function uses it as an example. 
 *Implement the separable_multi_kernel() function to allow users to specify a different kernel function for each dimension of the input
 *For the rgasp() and ppgasp() functions, now users can specify the number of initial values of the kernel parameters to be optimized and specify the initial values to be optimizers through the arguments “num_initial_values” and “initial_values”, respectively. 
 * Imports the R package nloptr as the main optimizer in rgasp() and ppgasp() depend on the lbfgs() function in nloptr. 


## change / RobustGaSP 0.5.6
 *Implement two functions called simulate() and plot() to replace Sample() and Plot() functions, respectively.
 *Reimplemented the demo-functions, including higdon.1.data(), limetal.2.data(), dettepepel.3.data(), borehole() and friedman.5.data()
 *The license is modified to GPL-2 | GPL-3. 
 *change the slight typo of the reference prior when the emulator has zero mean 
 *Add two references. The main paper is recently published in Annals of Statistics and another paper is on arXiv.

## RobustGaSP 0.5.5
 * Update the maximum num of evaluation in the optimizer by nloptr::lbfgs(). The maximum num of evaluation should increase with the increase of the dimension of the inputs
 * Implement leave_one_out_rgasp() function for model checking
 * Allow user to choose whether we use an S3 or S4 method for prediction


## RobustGaSP 0.5.4
  * Add the choice of the zero mean GaSP
  * Change the setting of nloptr to be nl.info=F.  Delete sink()
  * Correct printing of the noise

## change / RobustGaSP 0.5.2
  * Bugs corrected in rgasp() about the sink() function when the optimization step fails
  * Typo corrected in printing out the nugget value after the optimization is done
  * Allow the argument design in rgasp() to be initialized as a sequence (previously it has to be the matrix type)
  * Change of the maintainer’s Email address


## change / RobustGaSP 0.5
  * Implement a function ‘findInertInputs’ to identify inert inputs
  * Implement a function ‘Sample’ to sample from the predictive distribution
  * Modify the printing information of the function ‘rgasp’ by removing the automatic printing message from the function of ‘nloptr::lbfgs’ and adding other printing information
  * Change the parameter ‘beta’ to ‘beta_hat’ in rgasp class
  * Fix the bug in the function ‘predict’ of the value ‘sd’

 
