Help on GPR with multi-dimensional input (single dimension output)

[rumi1359]

Hello dev team. I'm having trouble constructing a GP regression model where train_x is multi-dimensional. Basically I have a bunch of training samples: "x" where each x has N>1 features (i.e is a 1xN tensor). The target "y" is still one dimensional for each x. The "GPyTorch Regression Tutorial" assumes both x and y to be 1 dimensional. How can I modify this for my use case? Thank you.

[gpleiss]

Your x tensor should be shaped num_data x num_dimensions, and your y tensor should be shaped num_data.

[rumi1359]

And then the same structures? :
likelihood = gpytorch.likelihoods.GaussianLikelihood()
model = ExactGPModel(train_X, train_Y, likelihood)

[gpleiss]

Yes, same structures