search

WentaoZhan1998 / geospaNN

MIT License
5 stars 1 forks source link

Using geospaNN.make_graph() on data_test #2

Open ritviksahajpal opened 3 months ago

ritviksahajpal commented 3 months ago

In the simulation example, it says:

_Kriging prediction from the model. The first variable is supposed to be the data used for training, and the second variable a torch_geometric.data.Data object which can be composed by geospaNN.makegraph()'.

How can we pass data_test to geospaNN.make_graph() which requires both X and Y, since data_test does not have Y information. Taking a look at your code, it seems like it should be ok for the Y to be all NaNs or None from data_test, would it have any unexpected consequences if I passed all NaNs for Y in data_test?

WentaoZhan1998 commented 3 months ago

Hello, sorry about the confusing statement. Yes, geospaNN.make_graph() can accept Y being all NaNs, and the prediction actually does not rely on the Y values in data_test. Currently what you can try is to input any Y (NaNs, or 0's if NaN gives any error) of the same length as X to geospaNN.make_graph() and compose the test data. In the next release, we will fix this issue by giving options on Y. Please let us know if there's still any problem.

ritviksahajpal commented 3 months ago

Thank you for your quick response! I will try and update here.