James T. Wilson

> @j-wilson I just noticed this PR is still not merged despite @vdutor having approved it: it seems you just need to get the tests to pass (and merge latest...

Good catch regarding transposing `other`, I've updated the earlier `__rmatmul__` method accordingly. From what you've said, it sounds like this simple solution may suffice for the time being. If so,...

@jacobrgardner Hi Jake. Fully on board with you here. As Max mentioned, I've put together a draft for this as well. At a glance, it looks pretty similar to your...

@jacobrgardner Good questions. I hadn't actually considered a solution like `fit_gpytorch_mll(mll, data_loader)`. I really like this API, but fear it may be too heavy-duty for simple use cases. My thought...

@jacobrgardner Up for discussion. A naive implementation would probably see `data_loader` as a `_fit_approximategp_stochastic`-specific keyword argument that gets ignored by other `fit_gpytorch_mll` subroutines.

Hi folks. I've put together a PR (#1439) that implements the above. This ended up being a larger change than I had originally anticipated, but hopefully people will agree that...

@Ruan-Yixiang I recommend you try using [MVNXPB](https://github.com/pytorch/botorch/blob/main/botorch/utils/probability/mvnxpb.py), which is a state-of-the-art approximator for Gaussian probabilities of the form `P(a < f(X) < b)`. In the case of qPI, this would...

Hi Adrian, 1. Can you elaborate on what you mean by "GP solution"? If you mean a GP posterior given (noisy) observations of a function, then the answer is "yes"....

> By GP solution, I meant a GP posterior with noisy observations. Our idea is to apply it to Learning from Demonstration, where we have learned the structure of knowledge...