SMCPy is an open-source package for performing uncertainty quantification using a parallelized sequential Monte Carlo sampler.
- Alternative to Markov chain Monte Carlo for Bayesian inference problems
- Unbiased estimation of marginal likelihood for Bayesian model selection
- Parallelization through either numpy vectorization or mpi4py
To install SMCPy, use pip.
pip install smcpyTo operate the code, the user supplies a computational model built in Python 3.6+, defines prior distributions for each of the model parameters to be estimated, and provides data to be used for probabilistic model calibration. SMC sampling of the parameter posterior distribution can then be conducted with ease through instantiation of a sampler class and a call to the sample() method.
The two primary sampling algorithms implemented in this package are MPI-enabled versions of those presented in the following articles, respectively:
Nguyen, Thi Le Thu, et al. "Efficient sequential Monte-Carlo samplers for Bayesian inference." IEEE Transactions on Signal Processing 64.5 (2015): 1305-1319. Link to Article | BibTeX Reference
Buchholz, Alexander, Nicolas Chopin, and Pierre E. Jacob. "Adaptive tuning of hamiltonian monte carlo within sequential monte carlo." Bayesian Analysis 1.1 (2021): 1-27. Link to Article | BibTeX Reference
The first is a simple likelihood tempering approach in which the tempering sequence is fixed and user-specified (FixedSampler). The second is an adaptive approach that chooses the tempering steps based on a target effective sample size (AdaptiveSampler).
This software was funded by and developed under the High Performance Computing Incubator (HPCI) at NASA Langley Research Center.
import numpy as np
from scipy.stats import uniform
from spring_mass_model import SpringMassModel
from smcpy.utils.plotter import plot_pairwise
from smcpy import AdaptiveSampler, VectorMCMC, VectorMCMCKernel
# Load data
std_dev = 0.5
displacement_data = np.genfromtxt('noisy_data.txt')
# Define prior distributions & MCMC kernel
priors = [uniform(0, 10), uniform(0, 10)]
vector_mcmc = VectorMCMC(model.evaluate, displacement_data, priors, std_dev)
mcmc_kernel = VectorMCMCKernel(vector_mcmc, param_order=('K', 'g'))
# SMC sampling
smc = AdaptiveSampler(mcmc_kernel)
step_list, mll_list = smc.sample(num_particles=500, num_mcmc_samples=5, target_ess=0.8)
# Display results
print(f'parameter means = {step_list[-1].compute_mean()}')
plot_pairwise(step_list[-1].params, step_list[-1].weights, save=True,
param_labels=['K', 'g'])The above code produces probabilistic estimates of K, the spring stiffness divided by mass, and g, the gravitational constant on some unknown planet. These estimates are in the form of weighted particles and can be visualized by plotting the pairwise weights as shown below. The mean of each parameter is marked by the dashed red line. The true values for this example were K = 1.67 and g = 4.62. More details can be found in the spring mass example. To run this model in parallel using MPI, the MCMC kernel just needs to be built with the ParallelMCMC class in place of VectorMCMC. More details can be found in the MPI example.
To run this model in parallel using MPI, the MCMC kernel just needs to be built with the ParallelMCMC class in place of VectorMCMC. More details can be found in the MPI example (smcpy/examples/mpi_example/).
Clone the repo and move into the package directory:
git clone https://github.com/nasa/SMCPy.git
cd SMCPyInstall requirements necessary to use SMCPy:
pip install -r requirements.txtOptionally, if you'd like to use the MPI-enabled parallel sampler, install the associated requirements:
pip install -r requirements_optional.txtAdd SMCPy to your Python path. For example:
export PYTHONPATH="$PYTHONPATH:/path/to/smcpy"Run the tests to ensure proper installation:
pytest tests- Fork (https://github.com/nasa/SMCPy/fork)
- Create your feature branch (
git checkout -b feature/fooBar) - Commit your changes (
git commit -am 'Add some fooBar') - Push to the branch (
git push origin feature/fooBar) - Create a Pull Request
NASA Langley Research Center
Hampton, Virginia
This software was funded by and developed under the High Performance Computing Incubator (HPCI) at NASA Langley Research Center.
- Patrick Leser
- Michael Wang
Notices: Copyright 2018 United States Government as represented by the Administrator of the National Aeronautics and Space Administration. No copyright is claimed in the United States under Title 17, U.S. Code. All Other Rights Reserved.
Disclaimers No Warranty: THE SUBJECT SOFTWARE IS PROVIDED "AS IS" WITHOUT ANY WARRANTY OF ANY KIND, EITHER EXPRESSED, IMPLIED, OR STATUTORY, INCLUDING, BUT NOT LIMITED TO, ANY WARRANTY THAT THE SUBJECT SOFTWARE WILL CONFORM TO SPECIFICATIONS, ANY IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, OR FREEDOM FROM INFRINGEMENT, ANY WARRANTY THAT THE SUBJECT SOFTWARE WILL BE ERROR FREE, OR ANY WARRANTY THAT DOCUMENTATION, IF PROVIDED, WILL CONFORM TO THE SUBJECT SOFTWARE. THIS AGREEMENT DOES NOT, IN ANY MANNER, CONSTITUTE AN ENDORSEMENT BY GOVERNMENT AGENCY OR ANY PRIOR RECIPIENT OF ANY RESULTS, RESULTING DESIGNS, HARDWARE, SOFTWARE PRODUCTS OR ANY OTHER APPLICATIONS RESULTING FROM USE OF THE SUBJECT SOFTWARE. FURTHER, GOVERNMENT AGENCY DISCLAIMS ALL WARRANTIES AND LIABILITIES REGARDING THIRD-PARTY SOFTWARE, IF PRESENT IN THE ORIGINAL SOFTWARE, AND DISTRIBUTES IT "AS IS."
Waiver and Indemnity: RECIPIENT AGREES TO WAIVE ANY AND ALL CLAIMS AGAINST THE UNITED STATES GOVERNMENT, ITS CONTRACTORS AND SUBCONTRACTORS, AS WELL AS ANY PRIOR RECIPIENT. IF RECIPIENT'S USE OF THE SUBJECT SOFTWARE RESULTS IN ANY LIABILITIES, DEMANDS, DAMAGES, EXPENSES OR LOSSES ARISING FROM SUCH USE, INCLUDING ANY DAMAGES FROM PRODUCTS BASED ON, OR RESULTING FROM, RECIPIENT'S USE OF THE SUBJECT SOFTWARE, RECIPIENT SHALL INDEMNIFY AND HOLD HARMLESS THE UNITED STATES GOVERNMENT, ITS CONTRACTORS AND SUBCONTRACTORS, AS WELL AS ANY PRIOR RECIPIENT, TO THE EXTENT PERMITTED BY LAW. RECIPIENT'S SOLE REMEDY FOR ANY SUCH MATTER SHALL BE THE IMMEDIATE, UNILATERAL TERMINATION OF THIS AGREEMENT.