Source

BayesOpt / src / gaussian_process.cpp

Author Commit Message Labels Comments Date
Ruben Martinez-Cantin
Error handling in kernel, mean and criterion functions
Tags
v0.5.1
Ruben Martinez-Cantin
Adding parameters to criteria. Adding Lizotte's EI.
Ruben Martinez-Cantin
Adding conditional compiling for test display
Ruben Martinez-Cantin
Solved bug in gaussian process
Ruben Martinez-Cantin
Gaussian process ml working!
Ruben Martinez-Cantin
New likelihood model for kernel parameters. Adding some forgotten files.
Ruben Martinez-Cantin
Refactoring namespaces. Improve criteria factory.
Ruben Martinez-Cantin
Refactoring namespaces. Cleaning dependencies in distributions.
Ruben Martinez-Cantin
Working on a new interface for kernels.
Ruben Martinez-Cantin
Adding new criterion. Few minor corrections.
Ruben Martinez-Cantin
Minor changes
Ruben Martinez-Cantin
Minor bugs in likelihood computation.
Ruben Martinez-Cantin
Refactoring process to separate distribution computations
Ruben Martinez-Cantin
Separate stochastic process from distribution. Minor penalty.
Ruben Martinez-Cantin
Cleanup interfaces. Added more smart pointers. Moved metacriteria to new files. Remove incremental inverse matrix for gaussian process, cholesky version is alwasy more efficient.
Ruben Martinez-Cantin
Added new logging system. Added smart pointer for surrogate function. Cleaned surrogate function interface.
Ruben Martinez-Cantin
Optimized cholesky version of GP_IGN. Restructured code to support cholesky based predictions in all surrogate functions (more efficient). Added C and Python wrappers for discrete optimization. Bug in Student t process.
Ruben Martinez-Cantin
Improvements on documentation
Ruben Martinez-Cantin
Parameters allow for multiple kernel hyperparameters (ARD). Added documentation in several files.
Ruben Martinez-Cantin
Cleaning code and refactoring the criteria component.
Ruben Martinez-Cantin
Solved bug in the computation of inverse correlation. Current implementation works but it is highly inefficient.
Ruben Martinez-Cantin
Major bug in covariance matrix update. Now it is done in using cholesky decomposition (more efficient). The new version is function for the basic gaussian process, but the rest of the surrogate functions require some modifications.
Ruben Martinez-Cantin
Refactoring with proper naming