fedetask / hyperparameter-optimization

Hyperparameter optimization python library
4 stars 1 forks source link

Negative predicted variance #8

Open fedetask opened 4 years ago

fedetask commented 4 years ago

The predicted variance can become negative when lots of points are added. I believe this is due to numerical approximation in inverting the kernel matrix. The issue seems to be very similar to the one discussed here

fedetask commented 4 years ago

I'll come up with example code to reproduce the error soon