Skip to main content

ACBICI

License Last Commit Documentation Status

ACBICI (A Configurable BayesIan Calibration and Inference package) is a Python package designed to calibrate models including uncertainty estimates.

It has the following functionality:

  • Bayesian calibration
  • provide distribution for the model parameters
  • estimate aleatoric and epistemic uncertainty
  • Gaussian process surrogate models for faster predictions
  • visualization of results

Installation

Before installing ACBICI, you need to install make, git and one of the tested Python versions (3.9, 3.12, 3.13, 3.14).

Then, if using Python for other purposes than ACBICI, it is recommended to use a virtual environment, you can create one for example via:
- python3 -m venv acbici_env
or:
- virtualenv acbici_env
Then activate the virtual environment by:
- source acbici_env/bin/activate

Upgrade packaging tools (recommended):
pip install --upgrade pip setuptools wheel

or with Anaconda (Windows) in the Anaconda prompt to create a virtual environment:
- conda create -n myenv python=3.14
Then to activate it:
- conda activate myenv

Next you can install ACBICI in this virtual environment:
Either:
Go to the folder where you want to install it via: cd "installation_directory"
Clone and install the repo into that folder in the following way:
- git clone https://gitlab.com/schenkch/ACBICI.git
- cd ACBICI
- pip install -e .
Or:
Install as an already built pip package via:
- git clone https://gitlab.com/schenkch/ACBICI.git 
- pip install ACBICI

Examples and Tutorials

The example problems are in the examples subfolder at https://gitlab.com/schenkch/acbici/tree/master/examples.

To run all the examples and test their correctness, open a terminal and change into the `examples` directory of the downloaded repository:
cd examples

Execute the following command:
make

This will trigger the build process defined in the Makefile and run all the example problems within the folder. This may take some time but if everything is set up correctly and all dependencies are satisfied, you should see output indicating whether each example has passed or failed. In addition, each example subfolder contains a make file for automatic clean up of old results.

Automatic MCMC continuation and restart

All Type A, B, C, and D calibrators can continue one MCMC chain until it converges or a user-selected maximum is reached. The chain is considered converged when, for every parameter, the post-burn chain is at least 50 times its autocorrelation time and the split-Rhat of every parameter is below 1.01.

Automatic continuation is opt-in: existing calls that specify only nsteps still run exactly that many steps and stop. Supply next_steps, max_steps, or restart=True to enable continuation, for example:

calibrator.calibrate(method="mcmc", nsteps=5000, next_steps=10000, max_steps=100000)

If a saved chain reaches its cumulative maximum before convergence, continue it with a larger cumulative limit:

calibrator.calibrate(method="mcmc", max_steps=200000, restart=True)

The chain is saved to disk (this requires h5py) only when next_steps, max_steps, or restart=True is given. A completed run without continuation, for example one that did not converge, can still be continued with restart=True: ACBICI then starts from the last walker positions in samples.npy, but cannot check that the model, data and settings are the same as in the original run. For long runs, enabling continuation is recommended: the chain is saved after every step and can be restarted even after a crash. For a single batch, set max_steps equal to nsteps, e.g. calibrate(nsteps=20000, max_steps=20000).

Final results combine all post-burn batches and restarts, burn-in remains fixed from the initial run, and ACBICI prints important progress to both the terminal and the same acbici.log file. The scripts typeA_restart.py and typeD_restart.py in examples/example0 show a first run and a restart (python typeA_restart.py restart).

SMC runs (method="smc") save their state after every iteration and can also be restarted, either to continue a finished run to a larger nsamples or after a crash:

calibrator.calibrate(method="smc", nsamples=8000, restart=True)

The script typeB_restart_smc.py in examples/example0 shows this for SMC.

Quiet mode

By default, ACBICI prints progress and warnings to the terminal. To run a calibration without any terminal output, for example in batch jobs, call

calibrator.setVerbose(False)

before calibrate. Then calibrate, plot and identifiabilityAnalysis print nothing, while acbici.log is written as usual and errors still stop the program.

License

BSD-3

Authors

- Christina Schenk - IMDEA Materials Institute
- Ignacio Romero - UPM, IMDEA Materials Institute

Acknowledgements

- Yufei Liu - IMDEA Materials Institute
- Ross Cockcroft - Queens University Belfast

Please cite


- https://gitlab.com/schenkch/ACBICI.git
- Schenk, C., Romero, I. A Framework for the Bayesian Calibration of Complex and Data-Scarce Models in Applied Sciences. Arch Computat Methods Eng (2026). https://doi.org/10.1007/s11831-026-10665-w
- Preprint here: C. Schenk, I. Romero (2026): A Framework for the Bayesian Calibration of Complex and Data-Scarce Models in Applied Sciences, arXiv, https://arxiv.org/abs/2601.22890

Metadata

Release files for ACBICI 2.3.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for ACBICI 2.3.0
File Size Uploaded Provenance Actions
acbici-2.3.0.tar.gz 64.8 kB Without provenance Details for acbici-2.3.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for ACBICI 2.3.0
File Interpreter ABI Platform Actions
acbici-2.3.0-py3-none-any.whl Python 3 none any Details for acbici-2.3.0-py3-none-any.whl

Total release size: 135.4 kB

Release files / acbici-2.3.0.tar.gz

Download URL acbici-2.3.0.tar.gz
Size 64.8 kB
Tags Source
SHA-256 checksum
How to use checksums
01d68f74f73ac7fe70ade7536fc420a94f6baf5883e2ac4773536ce0ef514807
BLAKE2b-256 checksum
How to use checksums
36cc79fc371dbd4433f38c59adae5b278ae728550336e9fc2f2a95ae14ff0798
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.14.5

Release files / acbici-2.3.0-py3-none-any.whl

Download URL acbici-2.3.0-py3-none-any.whl
Size 70.6 kB
Tags Python 3
SHA-256 checksum
How to use checksums
1bf7546e7f8cf65365a0580aa94499e6fb8aa360b9e10e2b84c37bdbedb0e136
BLAKE2b-256 checksum
How to use checksums
327949c8df8f6bf744b6fe832df8cf23711bf5471d591308e6457793ef13da95
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.14.5

Security for ACBICI 2.3.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page