Skip to content

Source code for our Paper "Learning in High-Dimensional Feature Spaces Using ANOVA-Based Matrix-Vector Multiplication"

License

Notifications You must be signed in to change notification settings

wagnertheresa/NFFT4ANOVA

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 

Repository files navigation

NFFT4ANOVA

This package uses the FastAdjacency package by Dominik Alfke to perform NFFT-based fast summation to speed up kernel-vector multiplications for the ANOVA kernel. It is targeted at large-scale kernel evaluations. We demonstrate our method's computational power by using it for kernel ridge regression, which is just one of many possible applications. For more details, see the above-mentioned paper. A huge benefit of this package is that even for very large-scale data, all codes can easily be run on a standard laptop computer in absolutely reasonable time, so that no superior hardware is required.

Installation

Usage

This package consists of the following three classes:

  • kernel_vector_multiplication compares the standard kernel-vector multiplication with kernel-vector multiplication with NFFT-based fast summation in runtime and approximation error.
  • NFFTKernelRidge performs NFFT-based kernel ridge regression.
  • GridSearch searches on candidate parameter values for one of the classifiers NFFTKernelRidge, sklearn KRRor sklearn SVC.

See test/showcase_kernel_vector_multiplication.ipynb and test/showcase_nfft_krr.ipynb for an example.

About

Source code for our Paper "Learning in High-Dimensional Feature Spaces Using ANOVA-Based Matrix-Vector Multiplication"

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published