Max Balandat
Max Balandat
Hmm there is some complication here since some of the transforms also take a `batch_shape` arg, and it's on the user to figure out how to modify this in a...
So the other option could be to make the transform API a bit richer so as to allow to programmatically `expand` to batch dimensions. Then one could do something like...
> Could be then necessary to also be able to pass transform specific kwargs Yeah that's going to make the interface a bit clunky, you'd have to do sth like...
Hi @jduerholt, great question. Using a Probit classification model with a Bernoulli likelihood for actual classification should work in this context. You're right that since such model maps the latent...
I can throw something together in a notebook on some toy problem, kind of curious myself how this works. Might take a bit to get this cleaned up and make...
Made some initial strides, making a BoTorch model based on the simple classification GP is easy, hooking things up into the model API will require a bit of thinking since...
wondering if @bletham or @mshvartsman have any thoughts here, as they were working on similar / related things.
@jduerholt, curious if you have any intuition as to whether in your case there is reason to expect / not expect smoothness in the probability of feasibility.
Adding @dme65 and @jtwilson who have been making strides towards this recently.
Yeah we're working on implementing this with a new probabilistic classification model. I don't have a timeline for open source for this yet, but hopefully we have something preliminary ready...