Skip to content

Latest commit

 

History

History
28 lines (15 loc) · 925 Bytes

File metadata and controls

28 lines (15 loc) · 925 Bytes

CKS - Compositional Kernel Search

This repository contains an implementation of the Compositional Kernel Search algorithm for automatic structure discovery in time series and regression data using Gaussian Processes.

Reference

This is an implementation of the algorithms described in:

"Structure Discovery in Nonparametric Regression through Compositional Kernel Search"

David Duvenaud, James Robert Lloyd, Roger Grosse, Joshua B. Tenenbaum, Zoubin Ghahramani

Proceedings of the 30th International Conference on Machine Learning (ICML), 2013

Paper: https://arxiv.org/abs/1302.4922

Contents

  • compositional_kernel_search.py - Main implementation of the CKS algorithm
  • examples.py - Example usage and demonstrations

Examples

Airline Example

Decomposition Example

Kernel Summary