cocopp 2.7.4

Benchmarking framework for all types of black-box optimization algorithms, postprocessing.

COmparing Continuous Optimisers (COCO) PostProcess

The cocopp Python package implements the postprocess part of COCO: A Platform for Comparing Continuous Optimizers in a Black-Box Setting, comparing not only continuous optimisers. The package uses data generated with the COCO framework and produces output figures and tables in html format and for inclusion into LaTeX documents.

Documentation

The main documentation pages for the coco-postprocess package cocopp can be found at

• getting-started
• API documentation
• Issue tracker and bug reports
• source code page on GitHub
• COCO home page

Below are the installation instruction and some usage examples.

Installation

To install the latest release from PyPI:

    pip install -U cocopp

To install the current main branch:

    git clone https://github.com/numbbo/coco-postprocess.git
    cd coco-postprocess
    pip install .

Usage

The main interface to the cocopp package is the main method (an alias to cocopp.rungeneric.main). The main method allows for basic usage of cocopp through a shell command-line interface. The recommended use is however from an IPython/Jupyter shell or notebook:

>>> import cocopp
>>> dslist = cocopp.main('exdata/my_output another_folder yet_another_or_not')  

postprocesses data from one or several folders, for example data generated with the help from the cocoex module. Each folder should contain data of a full experiment with a single algorithm. (Within the folder the data can be distributed over subfolders). Results can be explored from the ppdata/index.html file, unless a a different output folder is specified with the -o option. Comparative data from over 200 full experiments are archived online and can be listed, filtered, and retrieved from archives which are attributes of cocopp.archives and processed alone or together with local data. For example

>>> cocopp.archives.bbob('bfgs')  
['2009/BFGS_...

lists all data sets run on the bbob testbed containing 'bfgs' in their name. The first in the list can be postprocessed by

>>> dslist = cocopp.main('bfgs!')  

All of them can be processed like

>>> dslist = cocopp.main('bfgs*')  

Only a trailing * is accepted and any string containing the substring is matched. The postprocessing result of

>>> dslist = cocopp.main('bbob/2009/*')  

can be browsed at https://numbbo.github.io/ppdata-archive/bbob/2009. To display algorithms in the background, the cocopp.genericsettings.background variable needs to be set:

>>> cocopp.genericsettings.background = {None: cocopp.archives.bbob.get_all('bfgs')}  

where None invokes the default color (grey) and line style (solid) cocopp.genericsettings.background_default_style. Now we could compare our own data with the first 'bfgs'-matching archived algorithm where all other archived BFGS data are shown in the background with the command

>>> dslist = cocopp.main('exdata/my_output bfgs!')