Basic GPR regression: model.predict_f shape not the same as the example #2010
thismartian
started this conversation in
General
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi,
I have used the GPR regression model earlier without any issues, however, I recently tried to reproduce the example from https://gpflow.github.io/GPflow/2.6.3/notebooks/basics/regression.html and found that the shape of the predicted response tensor using m.predict_f for a 1D regression is not (length of x test, 1), but rather (length of x test, number of training samples). It's the same for m.predict_y.
m.predict_f_samples is of the shape (number of samples requested, length of x test, number of training samples).
Wondering if something changed recently...
Beta Was this translation helpful? Give feedback.
All reactions