dfm / george

Fast and flexible Gaussian Process regression in Python

Home Page:http://george.readthedocs.io

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Kernel with dynamical number of parameters

aaronkl opened this issue · comments

Hi,
I would like to implement a new kernel as described here. In a nutshell, given an integer N, this kernel maintains the Cholesky of an NxN dimensional matrix, where the entries of the Cholesky are pre-defined Hyperparameters. I was wondering what would be the way to implement this? Ideally, this kernel gets N as input and generates an M (number of dimensions of the Cholesky) dimensional parameter vector.

@dfm Could you kindly given some hints how to implement a new kernel with multiple parameters, and the number of parameters is specified by users? Thank you so much for your help!