optzer is a python package for parameter optimization of any external program and to any target property that can be computed by the external program. Of course, the optzer is named after “optimizer”. This can be called hyper-parameter optimization and there are several python packages doing similar tasks for machine learning (python) programs. The main difference from those packages is that the optzer is focusing on optimizing parameters in external programs written in non-python langauges.

The optzer is hosted at https://github.com/ryokbys/optzer and the current stable version can be found at https://pypi.org/project/optzer/

The optzer performs non-gradient and global-minimum search with numerous trials at a time (population) using multi-threading. Optimization methods currently available are:

• Cuckoo search (with adaptive update of search space)
• Weighted Parzen Estimator (similar to Tree-structured Parzen Estimator)

You can install optzer by using pip as,

pip install optzer

Or you can clone the github repository and install as a develop mode as,

git clone https://github.com/ryo.kbys/optzer.git ./optzer
cd optzer
python setup.py sdist
pip install -e .