Implementation of ensemble-based Hamiltonian Monte Carlo for multiple architectures.
This repository contains simple implementations of Hamiltonian Monte Carlo and its ensemble-based extensions in Python. The implementations shall include physical constants as they are intended to enable the application of physical intuition to the problem of determining the parameters of stochastic models using Bayesian inference.
The code utilises (NumPyro)[https://num.pyro.ai/en/stable/] to allow a user to define a
probabilistic model in the STAN probabilistic programming language ( a derivative of C++ )
that will then be fit using the implemented methods. The intent is to widen the range
of architectures that will be usable by an existing model base.
Note that some of the methods implemented here are already present in the above framework, albeit after undergoing 'mathematical mutilation'.
This repository consists of 3 main branches
main - Containing this readme as well as miscellaneous information. The "release" branch.dev - The development branch.Any other branches may be created but will be considered transients subject to removal at a later date.
The dev branch is to be used for code development and shall contain code approved by at least one
reviewer. The main branch will consist of code that represents major milestones of development
and has passed review and testing.
This repository contains the following directories:
references: directory with reference papers, notes and a work plan.src: the main source directorysamples: contains sample code from previous projects that can be taken as referencemanuals: contains installation manuals and miscellaneous documentationFormatting guidelines and code contribution procedures are to be found in RulesAndProcedures.md
Please peruse manuals/SetUp.md.