otkerneldesign

This Python module generates designs of experiments based on kernel methods such as Kernel Herding and Support Points with the classes:

• KernelHerding
• KernelHerdingTensorized
• GreedySupportPoints

Additionally, optimal weights for quadrature and validation designs are provided by the classes:

• BayesianQuadratureWeighting
• TestSetWeighting

Installation

~$ pip install otkerneldesign

Documentation & references

• Package documentation
• Usage in Machine Learning model validation
• Support points
• Kernel herding

Example

>>> import openturns as ot
>>> import otkerneldesign as otkd
>>> # Distribution definition
>>> distribution = ot.ComposedDistribution([ot.Normal(0.5, 0.1)] * 2)
>>> dimension = distribution.getDimension()
>>> # Kernel definition
>>> ker_list = [ot.MaternModel([0.1], [1.0], 2.5)] * dimension
>>> kernel = ot.ProductCovarianceModel(ker_list)
>>> # Kernel herding design
>>> kh = otkd.KernelHerding(kernel=kernel, distribution=distribution)
>>> kh_design, kh_indices = kh.select_design(size=20)

Authors

• Elias Fekhari
• Joseph Muré