search

pints-team / pints

Probabilistic Inference on Noisy Time Series
http://pints.readthedocs.io
Other
229 stars 33 forks source link
bayesian-methods inverse-problems numerical-optimization parameter-estimation

readme

Unit tests on multiple python versions Unit tests on multiple operating systems codecov Change-point testing code Change-point testing results binder readthedocs

What is Pints?

PINTS (Probabilistic Inference on Noisy Time-Series) is a framework for optimisation and Bayesian inference on ODE models of noisy time-series, such as arise in electrochemistry and cardiac electrophysiology.

PINTS is described in this publication in JORS, and can be cited using the information given in our CITATION file. More information about PINTS papers can be found in the papers directory.

Using PINTS

PINTS can work with any model that implements the pints.ForwardModel interface. This has just two methods:

n_parameters() --> Returns the dimension of the parameter space.

simulate(parameters, times) --> Returns a vector of model evaluations at
                                the given times, using the given parameters

Experimental data sets in PINTS are defined simply as lists (or arrays) of times and corresponding experimental values. If you have this kind of data, and if your model (or model wrapper) implements the two methods above, then you are ready to start using PINTS to infer parameter values using optimisation or sampling.

A brief example is shown below: An example of using PINTS in an optimisation (Left) A noisy experimental time series and a computational forward model. (Right) Example code for an optimisation problem. The full code can be viewed here but a friendlier, more elaborate, introduction can be found on the examples page.

Beyond time-series models, PINTS can be used on any error function or log-likelihood that takes real-valued, continuous parameters.

A graphical overview of the methods included in PINTS can be viewed here.

Examples and documentation

PINTS comes with a number of detailed examples, hosted here on github. In addition, there is a full API documentation, hosted on readthedocs.io.

Installing PINTS

The latest release of PINTS can be installed without downloading (cloning) the git repository, by opening a console and typing

$ pip install --upgrade pip
$ pip install pints

Note that you'll need Python 3.6 or newer.

If you prefer to have the latest cutting-edge version, you can instead install from the repository, by typing

$ git clone https://github.com/pints-team/pints.git
$ cd pints
$ pip install -e .[dev,docs]

To uninstall again, type:

$ pip uninstall pints

What's new in this version of PINTS?

To see what's changed in the latest release, see the CHANGELOG.

Contributing to PINTS

There are lots of ways to contribute to PINTS development, and anyone is free to join in! For example, you can report problems or make feature requests on the issues pages.

Similarly, if you want to contribute documentation or code you can tell us your idea on this page, and then provide a pull request for review. Because PINTS is a big project, we've written extensive contribution guidelines to help standardise the code — but don't worry, this will become clear during review.

License

PINTS is fully open source. For more information about its license, see LICENSE.

Get in touch

Questions, suggestions, or bug reports? Open an issue and let us know.

Alternatively, feel free to email us at pints at maillist.ox.ac.uk.