/coco-dockerized

Dockerized COCO i.e. numerical black-box optimization benchmarking framework

Primary LanguagePython

Installation

Requirements

  • Docker (version 19.03.13)
  • docker-compose (version 1.27.4)
  • Python (version 2.7)

Last requirement means that your algorithm should be executable in Python 2.7.

Clone this repository and run below command:

docker-compose build

Usage

Run benchmark

Read experiment.py documentation and modify that file according to your goals.

To run benchmark suite for your favorite algorithm execute written below command:

docker-compose up coco_executor

Application will save benchmark results to exdata directory. Don't forget to check if you have permission to read that directory.

Results processing

Modify post-processing.sh script and run below command:

docker-compose up coco_plotter

Application will create ppdata directory which contains processed benchmark(s) results.

It will also create index.html file which is a convenient way to look through obtained results.