Technical DescriptionPrognosticator TK (formerly BayesEst) is a machine learning toolkit with support for generalised linear regression, Gaussian processes, state-space models, covariance-stationary stochastic processes (version 0.2+), and various compound models. It also features custom model generation algorithms. It is primarily designed with financial applications in mind, but its architecture and feature set do not limit it to any specifc area of application. The library is intended to be relatively easy to use and extend, and sports an intuitive strategy and factory (versions 0.X)/builder (versions 1.X) design pattern-based API and an untemplated external interface. Private implementations based on templated classes are employed to allow for reuse of code between, e.g., double precision and single precision implementation of function bases, etc. Data container classes are used to allow for interoperability between modules internally using different numerical precisions/between CPU and GPU code (version 1.X+). OpenMP is used for CPU parallelisation.
Documentation 0.2.4 (full)
Documentation 0.2.4 (no implementation classes)