MetaOpt is a library that optimizes black-box functions using a limited amount of time and utilizing multiple processors. The main focus of MetaOpt is the parameter tuning for machine learning and heuristic optimization.
MetaOpt has been developed by the Computational Intelligence Group at the University of Oldenburg. The authors are: Renke Grunwald, Bengt Lüers, Jendrik Poloczek, Justin Heinermann, Oliver Kramer.
For a user guide, see the documentation. The following describes basic operations with the repository.
PyPIn |
GitHub |
master | |||
develop |
MetaOpt is obtainable via archives of past releases, but you can also get the sources by cloning the repository.
To get a working copy of the MetaOpt repository:
$ git clone https://github.com/cigroup-ol/metaopt.git
MetaOpt is available on PyPI, but you can also install it from source.
To install MetaOpt from PyPI using pip:
$ sudo pip install metaopt
To install MetaOpt from a working copy:
$ cd metaopt
$ sudo python setup.py install
To verify MetaOpt was installed correctly:
$ python metaopt --version
metaopt 0.1.0.0
MetaOpt has automated online tests, but you can also run them locally.
To run MetaOpt's test suite from a working copy:
$ cd metaopt
$ sudo pip install -r requirements_test.txt
$ make tests-all
MetaOpt has automated online test coverage reports, but you can also create them locally.
To create MetaOpt's coverage reports from a working copy:
$ cd metaopt
$ sudo pip install -r requirements_coverage.txt
$ make coverage
MetaOpt comes with examples, which you can view in the docs, but you can also run them locally.
To install MetaOpt's example dependecies and run an example from a working copy:
$ cd metaopt
$ sudo pip install -r requirements_examples.txt
$ PYTHONPATH=. python examples/svm_saes_global_timeout.py
MetaOpt has automatically generated online documentation, but you can build yourself a local copy.
To build MetaOpt's html documentation from a working copy:
$ cd metaopt
$ sudo pip install -r requirements_docs.txt
$ make docs