search

yuki-koyama / mathtoolbox

Mathematical tools (interpolation, dimensionality reduction, optimization, etc.) written in C++11 with Eigen
https://yuki-koyama.github.io/mathtoolbox/
MIT License
257 stars 35 forks source link

The GPR demo fails when the number of data points is large #47

Closed yuki-koyama closed 5 years ago

yuki-koyama commented 5 years ago

I've just found that the GPR demo fails to optimize its hyperparameters when there are many data points (e.g., more than 100). This needs to be examined.

yuki-koyama commented 5 years ago

I've found that, when GPR fails with relatively many data points, it is constructing a covariant matrix with a determinant of zero. This results in inf in its maximum likelihood calculation. I am looking for a solution to this issue.

yuki-koyama commented 5 years ago

This seems useful: https://blogs.sas.com/content/iml/2012/10/31/compute-the-log-determinant-of-a-matrix.html