Skip to content

AndOslandsbotn/Regression-with-StreaMRAK

Repository files navigation

Regression-with-StreaMRAK

This repository contains an implementation of the kernel method StreaMRAK [1], which is built on the kernel solver FALKON [2] as a base solver.

Install

To install the code, clone the repository and install the packages in the requirements.txt file using pip install -r requirements.txt from the root of your repository

Run

To run the kernel methods on a toy example run the following two scripts

  • main_train_streamrak.py
  • main_pred_streamrak.py
  • main_train_falkon.py
  • main_pred_falkon.py

References

[1] A. Oslandsbotn, Z. Kereta, V. Naumova, Y. Freund and A. Cloninger (2022). StreaMRAK a streaming multi-resolution adaptive kernel algorithm Applied Mathematics and Computation, vol 426, p. 127112 https://doi.org/10.1016/j.amc.2022.127112

[2] A. Rudi, L. Carratino and L. Rosasco (2017). FALKON: An Optimal Large Scale Kernel Method. Advances in Neural Information Processing Systems, vol 30 https://proceedings.neurips.cc/paper/2017/file/05546b0e38ab9175cd905eebcc6ebb76-Paper.pdf

About

This repository contains an implementation of StreaMRAK

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages