Machine learning fingerprints for particle environments
Pythia is a library to generate numerical descriptions of particle systems. Most methods rely heavily on freud for efficient neighbor search and other accelerated calculations.
Pythia is available on PyPI as pythia-learn:
$ pip install pythia-learn freud-analysis
You can install pythia from source like this:
$ git clone https://bitbucket.org/glotzer/pythia.git $ # now install $ cd pythia && python setup.py install --user
If using conda or a virtualenv, the –user argument in the pip command above is unnecessary.
In addition to the citations referenced in the docstring of each function, we encourage users to cite the pythia project itself.
The documentation is available as standard sphinx documentation:
$ cd doc $ make html
Automatically-built documentation is available at https://pythia-learn.readthedocs.io .
In general, data types follow the hoomd-blue schema:
- Positions are an Nx3 array of particle coordinates, with (0, 0, 0) being the center of the box
- Boxes are specified as an object with Lx, Ly, Lz, xy, xz, and yz elements
- Orientations are specified as orientation quaternions: an Nx4 array of (r, i, j, k) elements
Example notebooks are available in the examples directory: