Adversarial Resilience Learning --- Design of Experiments

Introduction

Reproducibility of scientific experiments is more important than ever. The Adversarial Resilience Learning -- Design of Experiments Suite arsenAI serves exactly this purpose. It also enables to test different combinations of agents and environments in palaestrAI.

Installation

ArsenAI is written in Python and available on pypi.org. A virtual environment is recommended. To install arsenAI, type:

pip install palaestrai-arsenai

To execute experiment runs, you also need palaestrAI.

Usage

In the sources of arsenAI, there is an example_experiment.yml file. You find it in the folder tests/fixtures/. Download this file and save it to your current working directory. To use arsenAI, simply type

arsenai generate ./example_experiment.yml

An output folder will be created (default: (current working directory)/_outputs)). After the arsenAI command has finished, you will find palaestrAI run files and that directory, which can be executed with

$ palaestrai start _outputs/Dummy\ Experiment_run-0.yml

You can copy the example file and modify it to your needs.

Documentation

In the future, you will find a more comprehensive documentation on docs.palaestr.ai

Copyright & Authors

All source code, except where otherwise mentioned, is Copyright (C) 2020, 2021, 2022 OFFIS e.V. Contributing authors are listed in order of their appearance in the file CONTRIBUTING.md.