bask 0.7.2

A fully Bayesian implementation of sequential model-based optimization

Bayes-skopt

A fully Bayesian implementation of sequential model-based optimization

• Free software: Apache Software License 2.0

• Documentation: https://bayes-skopt.readthedocs.io.

• Built on top of the excellent Scikit-Optimize (skopt).

Features

• A fully Bayesian variant of the GaussianProcessRegressor.

• State of the art information-theoretic acquisition functions, such as the Max-value entropy search or Predictive variance reduction search, for even faster convergence in simple regret.

• Familiar Optimizer interface known from Scikit-Optimize.

Installation

To install the latest stable release it is best to install the version on PyPI:

pip install bask

The latest development version of Bayes-skopt can be installed from Github as follows:

pip install git+https://github.com/kiudee/bayes-skopt

Another option is to clone the repository and install Bayes-skopt using:

python setup.py install

Credits

This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.

History

0.7.2 (2020-08-01)

• Fix guess_priors not correctly adding the prior for the WhiteKernel. It is now called directly in BayesGPR.sample.

0.7.1 (2020-07-28)

• Restrict length scale bounds of the default kernel to a tighter interval. This should help start the MCMC walkers in a region with higher likelihood.

0.7.0 (2020-07-26)

• Replace the default inverse gamma distribution prior for the lengthscales by the round-flat distribution.

• Fix guess_priors to correctly add kernels with multiple lengthscales.

0.6.0 (2020-05-21)

• Add Optimizer.expected_optimality_gap which estimates the expected optimality gap of the current global optimum to random optima sampled from the Gaussian process.

• Check that the list of priors has the correct length.

• Require emcee to be at least version 3.0.

0.5.0 (2020-05-21)

• Add Optimizer.probability_of_optimality which estimates the probability that the current global optimum is optimal within a certain tolerance. This can be used to make stopping rules.

0.4.1 (2020-05-19)

• Update and fix dependencies.

0.4.0 (2020-04-27)

• Add return_policy parameter to BayesSearchCV. Allows the user to choose between returning the best observed configuration (in a noise-less setting) or the best predicted configuration (for noisy targets).

0.3.3 (2020-03-16)

• Fix error occuring when an unknown argument was passed to Optimizer.

0.3.0 (2020-03-12)

• Add predictive variance reduction search criterion. This is the new default acquisition function.

• Implement BayesSearchCV for use with scikit-learn estimators and pipelines. This is an easy to use drop-in replacement for GridSearchCV or RandomSearchCV. It is implemented as a wrapper around skopt.BayesSearchCV.

• Determine default kernels and priors to use, if the user provides none.

• Add example notebooks on how to use the library.

• Add API documentation of the library.

0.2.0 (2020-03-01)

• Allow user to pass a vector of noise variances to tell, fit and sample. This can be used to warm start the optimization process.

0.1.2 (2020-02-16)

• Fix the tell method of the optimizer not updating _n_initial_points correctly, when using replace.

0.1.0 (2020-02-01)

• First release on PyPI.