Package for different inducing points selection methods
Author JuliaGaussianProcesses
4 Stars
Updated Last
2 Years Ago
Started In
July 2020


Stable Dev BuildStatus Coverage

A package for selecting inducing points for sparse GPs

This package provide a collection of inducing point location selection algorithms, both offline and online.

For example

using InducingPoints
Z = KmeansIP(X, 10, obsdim = 1)

will return 10 inducing points selected as clusters by the k-means algorithm

For online algorithms, one need to first create an empty object. For example

Z = OIPS(200) # Method from unpublished paper (for now!)

The first time data is met init(Z, X, args...) should be called (args depend of the algorithm). init will return a new object correctly parametrized.

In the subsequent calls update!(Z, X, args...) to add/remove/move points.