Pure NumPy / SciPy implementation of samplers based on Hamiltonian mechanics. Includes implementations of the following methods:
- Hamiltonian Monte Carlo from Neal, R. (2011), Handbook of Markov Chain Monte Carlo. Chapman and Hall/CRC.
- Riemannian Manifold Hamiltonian Monte Carlo from Girolami, M. and Calderhead, B. (2011), Riemann manifold Langevin and Hamiltonian Monte Carlo methods. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 73: 123-214.
- Lagrangian Monte Carlo from Lan S., Stathopoulos V., Shahbaba B., and Girolami M. (2015), Markov Chain Monte Carlo from Lagrangian Dynamics. J Comput Graph Stat: 24(2):357-378.
In order to install this software, you should execute pip install -e . from the directory containing this README. Additionally, you will require the Odyssey.
This code contains significant components that are identical to, or adapted from, a previous implementation of this software released at Evaluating the Implicit Midpoint Integrator and Thresholds in Hamiltonian Monte Carlo, which were provided under an MIT License.