search

cornellius-gp / gpytorch

A highly efficient implementation of Gaussian Processes in PyTorch
MIT License
3.46k stars 546 forks source link

What is intention/pattern for multi-typed $X$ inputs for kernels and everything else? #2441

Open cottrell opened 7 months ago

cottrell commented 7 months ago

I have yet to find an example for this.

It looks like there is an assumption that x1, x2 are tensors. Perhaps some parts of the framework support x1 being a list of tensors? Or a dict? It's not obvious as I'm trying things. The obvious example is int and float features which flow into different sub kernels.