optproblems 1.3

Description

This package contains a collection of common benchmark problems for black-box optimization. Under the term “black-box problem”, we understand problems for which only little is known about their structure and properties. Such problems usually appear in practice when simulator output or some other complex system with nonlinear behavior is to be optimized.

Contained test problems:

• Binary problems OneMax, LeadingOnes, and LeadingOnesTrailingZeros and three instance generators for multimodal problems

• CEC 2005 collection of single-objective problems

• CEC 2007 collection of multiobjective problems

• Dixon-Szegö collection for global optimization

• DTLZ problems 1-7

• Multiple-Peaks Model 2

• Walking Fish Group (WFG) toolkit

• ZDT collection for multiobjective optimization

The infrastructure of this package can also be used to wrap your own (real-world) optimization problems in the problem base class. Reasons to do this may be the following features:

• Support for single-objective and multi-objective problems

• In general, no assumptions about the search space are made

• Evaluations are automatically counted

• Can use true parallelism or concurrency via multiprocessing(.dummy)

• Provides functionality for checking bound constraints and repairing violations of them in continuous optimization

• Optionally: detection and subsequently avoidance of duplicate evaluations

A few real-world problems are also included. Examples:

• Configuration of a gradient method on test problem

• Uniformity optimization of points in the unit hypercube

Dependencies

Some modules of this package have additional dependencies on third-party packages. However, these are not enforced during installation not to hinder the utilization of the base module, which will always be free of dependencies.

Documentation

The documentation is located at https://www.simonwessing.de/optproblems/doc/