GP.conditional Input dimension mis-match

Questions
1

Hi, I’m new to PyMc3 and programming/gaussian processes in general. I’m trying to do a simple GP process on some 1D data (to start) using the following code and I’m getting an error that I just can’t fix and would really appreciate any ideas anyone might have.

It seems I can’t use the conditional method on a marginal gp model because the Xnew data is a different size to the X_train data, but in the examples in the documentation it works fine?
I don’t know how to enter the code and error in a nice way so I’m sorry for the screen shots.
FYI it does not work with the # out ‘zero’ mean function either.

Thanks.

image
image1272×750 143 KB

2

Here are the error images

image
image2137×812 321 KB

3

image
image2119×1012 266 KB

4

Ok, so I found the solution I think from another question by Ally-Lee. I think it’s because I made my y variable into column vectors as well as the X vectors like - y[:, None] when I should have just left it with dims (50, ). I still don’t understand why it was a problem but it seems to be working now so thanks to Ally and anyone who might want to offer an explanation.

3 Likes