Skip to content

q-dim bugfix in ScalarizedPosteriorMean #3191

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
hvarfner wants to merge 1 commit into from
Closed

q-dim bugfix in ScalarizedPosteriorMean #3191

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 13 additions & 5 deletions botorch/acquisition/analytic.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1078,12 +1078,20 @@ def forward(self, X: Tensor) -> Tensor:
t-batches of ``d``-dim design points each.

Returns:
A ``(b1 x ... x bk)``-dim Tensor of Posterior Mean values at the given
design points ``X``.
A ``(b1 x ... x bk)``-dim Tensor of scalarized Posterior Mean values
at the given design points ``X``.
"""
# (b1 x ... x bk) x q x 1
mean, _ = self._mean_and_sigma(X, compute_sigma=False)
return mean.squeeze(-1) @ self.weights
# ScalarizedPosteriorMean cannot use self._mean_and_sigma, since that squeezes
# the q-dim.
self.to(X) # Sync weights buffer to X's device/dtype
posterior = self.model.posterior(
X=X, posterior_transform=self.posterior_transform
)
# posterior.mean has shape (b1 x ... x bk) x q x m
# squeeze(-1) removes m (should be 1), giving (b1 x ... x bk) x q
mean = posterior.mean.squeeze(-1)
# @ self.weights: (b1 x ... x bk) x q @ q -> (b1 x ... x bk)
return mean @ self.weights


class PosteriorStandardDeviation(AnalyticAcquisitionFunction):
Expand Down