I just wonder if there's a easy way to use GPflow for multiple trials?
Basically, I now have K trials and each trial has p outputs. Each trial is an instance (i.e. a sample) of a MOGP, but all these trials are described by1 MOGP.
I guess there should be a very easy way to handel this, rather than writing my own loss function and optimize it...
I found the question here (https://github.com/GPflow/GPflow/issues/650) is almost the same as mine, but the answeres are quite unclear to me...