Gmr Versions Save

Final version of JOSS paper

• Fixes numerical issues in condition and expectation step
• Add oracle approximating shrinkage to handle singular covariances
• Add sklearn-compatible GaussianMixtureRegressor
• Faster batch prediction of means
• Accept lists where previously only numpy arrays were accepted

Installation works now without any dependencies.

Installation

• Changes requires to install_requires in setup.py
• Adds optional requirements for examples, tests, and documentation

Documentation

• Links new API documentation in readme
• Documentation in example plot_iris_from_sklearn.py explains why the fit is bad. The purpose of this example is to show that we can initialize from a GMM of sklearn.
• Documentation of how to contribute to the software (in readme)

Code

• Fixes deprecation warnings for NumPy type aliases
• Adds GMM.extract_mvn

• Adds unscented transform to MVN
• Extend documentation in readme