The sources are available mainly in two formats; that is, M-file (for MATLAB) and MEX-C (C language). The latter is more efficient in large-scale computations. MEX-C (and not C) is used because it allows calling of compiled C binaries from MATLAB environment.

NEW!!! Access the Utility Toolboxes Websites

Gaussian Process Toolbox

Contains basic GP functions.

Gaussian Process Toeplitz Toolbox

Fast algorithms using modified Durbin-Levinson's algorithm. This algorithm was initially developed by Y. Zhang, but has been ported by me into C language for better computational efficiency.

Gaussian Process Schur Toolbox

Another fast algorithm based on the Generalised Schur algorithm. This is algorithm is much more robust than the modified Durbin-Levinson's algorithm as it is capable of handling time-series datasets with missing gaps, allows user-supplied Hessian for optimisation and computation of the standard deviation for the posterior.

Gaussian Process SSTS Toolbox

SSTS stands for State-Space Time-Series, a novel model using compound covariance matrix to incorporate time-series and state-space information of a data, to allow analysis of large-scale state-space datasets.