COBYQA, an acronym for Constrained Optimization BY Quadratic Approximations, is designed to supersede COBYLA as a general derivative-free optimization solver. It can handle unconstrained, bound-constrained, linearly constrained, and nonlinearly constrained problems. It uses only function values of the objective and constraint functions, if any. No derivative information is needed.
Documentation: https://www.cobyqa.com.
COBYQA can be installed for Python 3.8 or above.
The following Python packages are required by COBYQA:
If you install COBYQA using pip or conda (see below), these dependencies will be installed automatically.
The easiest way to install COBYQA is using pip or conda.
To install it using pip, run in a terminal or command window
pip install cobyqaIf you are using conda, you can install COBYQA from the conda-forge channel by running
conda install conda-forge::cobyqaTo check your installation, you can execute
python -c "import cobyqa; cobyqa.show_versions()"If your python launcher is not python, you can replace it with the appropriate command (similarly for pip and conda).
For example, you may need to use python3 instead of python and pip3 instead of pip.
To execute the test suite of COBYQA, you first need to install pytest.
You can then run the test suite by executing
pytest --pyargs cobyqaThe test suite takes several minutes to run. It is unnecessary to run the test suite if you installed COBYQA using the recommended method described above.
The folder examples contains a few examples of how to use COBYQA.
These files contain headers explaining what problems they solve.
To report a bug or request a new feature, please open a new issue using the issue tracker.