Skip to content

JuliaGaussianProcesses/InducingPoints.jl

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

92 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

InducingPoints

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.

Offline algorithms

Offline algorithms are meant to be run once over the data before training begins. Here is an example where we use the k-means algorithm

using InducingPoints
X = [rand(5) for _ in 1:100]
alg = KMeansAlg(10) # Create the kmeans algorithm
Z = inducingpoints(alg, X) # Returns a vector of vector of size 10 

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

Note that it is possible to pass data as a matrix as well following the convention of KernelFunctions.jl

X = rand(5 , 1000)
alg = KMeansAlg(10, Euclidean()) # We can also use different metrics
Z = inducingpoints(alg, X) # This still returns a vector of vector of size 10 

Online algorithms

Online algorithms needs two API, a first one to create the initial vector of inducing points and another one to update it with new data. For example following this work

alg = OIPS(200) # We expect 200 inducing points
kernel = SqExponential()
X = [rand(5) for _ in 1:100] # We have some initial data
Z = inducingpoints(alg, X; kernel=kernel) # We create an initial vector
X_new = [rand(5) for _ in 1:50] # We get some new data
updateZ!(Z, alg, X_new; kernel=kernel) # Points will be acordingly added (or removed!)

Note that Z is directly changed in place.

Notes

Make sure to check each algorithm docs independently, they will give you more details on what arguments they need and what they do!

About

Package for different inducing points selection methods

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •  

Languages