epyc 0.13.1

Python computational experiment management

Overview

epyc is a Python module for controlling a long-running series of computational experiments, as is often found when writing simulations of complex networks and other such domains. There is often a need to perform a computation across a multi-dimensional parameter space, varying the parameters, performing and aggregating multiple repetitions, and wrangling results for analysis and presentation. Often the experiments being performed are on such a scale as to require the use of a computing cluster to perform multiple experiments simultaneously.

Managing all these tasks is complicated, so epyc tries to automate it. It provides a way to define a “laboratory” performing a collection of “experiments” whose parameters and results are recorded in a “lab notebook” for later retrieval. Laboratories can be sequential (for a single machine) or parallel (to use a multicore or cluster of machines); lab notebooks can be persistent to allow experiments to be fired-off and their results retrieved later – handy if you use a laptop. Notebooks store all the data and metadata in a portable format to improve the reproducibility of computational experiments.

epyc also includes a small number of “experiment combinators” that separate the logic of a single experiment from the logic of performing multiple repetitions and other structuring tasks. This means that any experiment can be repeated and statistically summarised, for example.

Installation

You can install epyc directly from PyPi using pip:

pip install epyc

The master distribution of epyc is hosted on GitHub. To obtain a copy, just clone the repo:

git clone [e-mail]:simoninireland/epyc.git
cd epyc
python setup.py install

Documentation

API documentation for epyc can be found on ReadTheDocs <https://epyc.readthedocs.io/en/latest/>. You can also read a Jupyter notebook describing several epyc use cases online at <https://github.com/simoninireland/epyc/blob/master/doc/epyc.ipynb>.

Author and license

Copyright (c) 2016-2018, Simon Dobson <[e-mail]>

Licensed under the GNU General Public Licence v.2.0 <https://www.gnu.org/licenses/old-licenses/gpl-2.0.en.html>.