Repository files navigation

Scikit-Optimize, orskopt, is a simple and efficient library tominimize (very) expensive and noisy black-box functions. It implementsseveral methods for sequential model-based optimization.skopt aimsto be accessible and easy to use in many contexts.

The library is built on top of NumPy, SciPy and Scikit-Learn.

We do not perform gradient-based optimization. For gradient-basedoptimization algorithms look atscipy.optimizehere.

Approximated objective function after 50 iterations ofgp_minimize.Plot made usingskopt.plots.plot_objective.

• Static documentation -Staticdocumentation
• Example notebooks - can be found inexamples.
• Issue tracker -https://github.com/scikit-optimize/scikit-optimize/issues
• Releases -https://pypi.python.org/pypi/scikit-optimize

scikit-optimize requires

• Python >= 3.6
• NumPy (>= 1.13.3)
• SciPy (>= 0.19.1)
• joblib (>= 0.11)
• scikit-learn >= 0.20
• matplotlib >= 2.0.0

You can install the latest release with:

pip install scikit-optimize

This installs an essential version of scikit-optimize. To install scikit-optimizewith plotting functionality, you can instead do:

pip install 'scikit-optimize[plots]'

This will install matplotlib along with scikit-optimize.

In addition there is aconda-forge packageof scikit-optimize:

conda install -c conda-forge scikit-optimize

Using conda-forge is probably the easiest way to install scikit-optimize onWindows.

Find the minimum of the noisy functionf(x) over the range-2 < x < 2 withskopt:

importnumpyasnpfromskoptimportgp_minimizedeff(x):return (np.sin(5*x[0])* (1-np.tanh(x[0]**2))+np.random.randn()*0.1)res=gp_minimize(f, [(-2.0,2.0)])

For more control over the optimization loop you can use theskopt.Optimizerclass:

fromskoptimportOptimizeropt=Optimizer([(-2.0,2.0)])foriinrange(20):suggested=opt.ask()y=f(suggested)opt.tell(suggested,y)print('iteration:',i,suggested,y)

Read ourintroduction to bayesianoptimizationand the otherexamples.

The library is still experimental and under heavy development. Checkoutthenextmilestonefor the plans for the next release or look at someeasyissuesto get started contributing.

The development version can be installed through:

git clone https://github.com/scikit-optimize/scikit-optimize.gitcd scikit-optimizepip install -e.

Run all tests by executingpytest in the top level directory.

To only run the subset of tests with short run time, you can usepytest -m 'fast_test' (pytest -m 'slow_test' is also possible). To exclude all slow running tests trypytest -m 'not slow_test'.

This is implemented using pytestattributes. If a tests runs longer than 1 second, it is marked as slow, else as fast.

All contributors are welcome!

The release procedure is almost completely automated. By tagging a new releasetravis will build all required packages and push them to PyPI. To make a releasecreate a new issue and work through the following checklist:

• update the version tag in__init__.py
• update the version tag mentioned in the README
• check if the dependencies insetup.py are valid or need unpinning
• check that thedoc/whats_new/v0.X.rst is up to date
• did the last build of master succeed?
• create anew release
• pingconda-forge

Before making a release we usually create a release candidate. If the nextrelease is v0.X then the release candidate should be tagged v0.Xrc1 in__init__.py. Mark a release candidate as a "pre-release"on GitHub when you tag it.

Feel free toget in touch if you need commercialsupport or would like to sponsor development. Resources go towards payingfor additional work by seasoned engineers and researchers.

The scikit-optimize project was made possible with the support of

If your employer allows you to work on scikit-optimize during the day and would likerecognition, feel free to add them to the "Made possible by" list.