simexpal is a Python based tool to setup, manage, launch, monitor, and evaluate algorithmic experiments. If you can not wait to try it out and dive into it, you may find our quick start guide handy.
simexpal is mainly developed by the MACSy (Modelling and Analysis of Complex Systems) group where graph data is at the core of most projects. Therefore, the tool provides some special functionality for graph algorithm experiments. However, this does not limit simexpal to be only used for graph algorithm experiments, but is a tool that can be used to setup and manage almost all algorithmic experiments.
Since algorithmic experiments require several tasks in a specific order, the goal of this tool is to clearly describe and automatize these ordered, and consecutively executed tasks. Firstly, simexpal allows to make quick changes to the experiment setup. Secondly, it allows to describe experiments that are easy to rerun. When describing the necessary steps, even users with no knowledge of simexpal should be able to quickly setup and execute the described experiments.
simexpal consists of both a command line interface (CLI) utility and a Python package. While the CLI can be used to perform many common tasks with minimal configuration, the Python package can be employed when more extensive customization is necessary.
The following is a non-exhaustive list of the functionality provided by SimexPal:
- resolve external resources such as Git repositories
- describe and setup build processes
- declaratively describe experiments using the YAML language
- define parameters and parameter permutations
- manage data sets and specifically graph data
- run and monitor experiments on multiple computing platforms
- evaluate experiment results
Whether you're a student, an algorithm enthusiast, or a researcher: developing experiments that can easily be reproduced will support the validity of your findings. Self-written scripts often bind experiments to a narrow set of soft- and hardware. If not well documented, such scripts may be a source of error, or the experiments may not be reproduced as intended.
Even though simexpal may not yet be the perfect algorithm experiment management software, our goal was to abstract algorithmic experiments from their hard- and software requirements, as well as to provide a clearly structured and well documented experiment setup. simexpal will allow you to share your experiments with third parties, and it allows the third parties to validate your experimental results and findings with only a minimal description on how to build and run the experiments on your side.
Ultimately, simexpal represents an algorithmic experiment management tool with which one can clearly define their experiment setup and execution, and therefore majorly contributing to the goal of reproducible experiments.
Please see readthedocs for documentation of the simexpal features.
Our documentation also includes a quick start guide for a quick and easy step-by-step guide to install and use simexpal.
Whether you're using simexpal for you open source project, as part of your research, or any other purpose, please use the following article as reference from MDPI:
Angriman, E.; van der Grinten, A.; von Looz, M.; Meyerhenke, H.; Nöllenburg, M.; Predari, M.; Tzovas, C. Guidelines for Experimental Algorithmics: A Case Study in Network Analysis. Algorithms 2019, 12, 127. https://doi.org/10.3390/a12070127
BibTex format (w/o abstract):
@article{SimexPal2019,
author = {Angriman, Eugenio and van der Grinten, Alexander and von Looz, Moritz and Meyerhenke, Henning and Nöllenburg, Martin and Predari, Maria and Tzovas, Charilaos},
title = {Guidelines for Experimental Algorithmics: A Case Study in Network Analysis},
journal = {Algorithms},
volume = {12},
year = {2019},
number = {7},
article-number = {127},
url = {https://www.mdpi.com/1999-4893/12/7/127},
issn = {1999-4893},
doi = {10.3390/a12070127}
}