Skip to content

Fix reaction-diffusion example and update GP predictive #639

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

Merged
sgreenbury merged 8 commits into from
Jul 22, 2025
Merged

Fix reaction-diffusion example and update GP predictive #639

Changes from 1 commit
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
3e8aeb2
Add notebook for v0 version and update hyperparams
sgreenbury Jul 14, 2025
3fb62c0
Revise to current GP default params, update comparison
sgreenbury Jul 21, 2025
6ade997
Add notebook comparison with observation noise
sgreenbury Jul 21, 2025
561e8ad
Fix lint
sgreenbury Jul 21, 2025
5b5b091
Add posterior_predictive to GP API and predict
sgreenbury Jul 21, 2025
5f11066
Update example
sgreenbury Jul 21, 2025
be3bb5e
Remove obsolete notebooks
sgreenbury Jul 21, 2025
877776b
Remove obsolete notebook, update revised notebook
sgreenbury Jul 21, 2025
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
Prev Previous commit
Next Next commit
Add posterior_predictive to GP API and predict
  • Loading branch information
@sgreenbury
sgreenbury committed Jul 21, 2025
commit 5b5b091e932ca16a4bc0843e5fa06a13a8a918ce
26 changes: 21 additions & 5 deletions autoemulate/experimental/emulators/gaussian_process/exact.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,3 @@
# import logging

import gpytorch
import torch
from gpytorch import ExactMarginalLogLikelihood
Expand Down Expand Up @@ -59,7 +57,8 @@ def __init__( # noqa: PLR0913 allow too many arguments since all currently requ
y: TensorLike,
likelihood_cls: type[MultitaskGaussianLikelihood] = MultitaskGaussianLikelihood,
mean_module_fn: MeanModuleFn = constant_mean,
covar_module_fn: CovarModuleFn = rbf,
covar_module_fn: CovarModuleFn = rbf_plus_constant,
posterior_predictive: bool = False,
epochs: int = 50,
activation: type[nn.Module] = nn.ReLU,
lr: float = 2e-1,
Expand All @@ -84,6 +83,10 @@ def __init__( # noqa: PLR0913 allow too many arguments since all currently requ
Function to create the mean module.
covar_module_fn: CovarModuleFn, default=rbf
Function to create the covariance module.
posterior_predictive: bool, default=False
If True, the model will return the posterior predictive distribution that
by default includes observation noise (both global and task-specific). If
False, it will return the posterior distribution over the modelled function.
epochs: int, default=50
Number of training epochs.
activation: type[nn.Module], default=nn.ReLU
Expand Down Expand Up @@ -135,6 +138,7 @@ def __init__( # noqa: PLR0913 allow too many arguments since all currently requ
self.optimizer = self.optimizer_cls(self.parameters(), lr=self.lr) # type: ignore[call-arg] since all optimizers include lr
self.scheduler_setup(kwargs)
self.early_stopping = early_stopping
self.posterior_predictive = posterior_predictive
self.to(self.device)

@staticmethod
Expand Down Expand Up @@ -199,8 +203,14 @@ def _fit(self, x: TensorLike, y: TensorLike):
def _predict(self, x: TensorLike, with_grad: bool) -> GaussianProcessLike:
with torch.set_grad_enabled(with_grad):
self.eval()
self.likelihood.eval()
x = x.to(self.device)
return self(x)
output = self(x)
if not self.posterior_predictive:
return output
output = self.likelihood(output)
assert isinstance(output, GaussianProcessLike)
return output

@staticmethod
def get_tune_config():
Expand Down Expand Up @@ -249,7 +259,8 @@ def __init__( # noqa: PLR0913 allow too many arguments since all currently requ
y: TensorLike,
likelihood_cls: type[MultitaskGaussianLikelihood] = MultitaskGaussianLikelihood,
mean_module_fn: MeanModuleFn = constant_mean,
covar_module_fn: CovarModuleFn = rbf,
covar_module_fn: CovarModuleFn = rbf_plus_constant,
posterior_predictive: bool = False,
epochs: int = 50,
activation: type[nn.Module] = nn.ReLU,
lr: float = 2e-1,
Expand All @@ -275,6 +286,10 @@ def __init__( # noqa: PLR0913 allow too many arguments since all currently requ
Function to create the mean module.
covar_module_fn: CovarModuleFn, default=rbf
Function to create the covariance module.
posterior_predictive: bool, default=False
If True, the model will return the posterior predictive distribution that
by default includes observation noise (both global and task-specific). If
False, it will return the posterior distribution over the modelled function.
epochs: int, default=50
Number of training epochs.
activation: type[nn.Module], default=nn.ReLU
Expand Down Expand Up @@ -335,6 +350,7 @@ def __init__( # noqa: PLR0913 allow too many arguments since all currently requ
self.optimizer = self.optimizer_cls(self.parameters(), lr=self.lr) # type: ignore[call-arg] since all optimizers include lr
self.scheduler_setup(kwargs)
self.early_stopping = early_stopping
self.posterior_predictive = posterior_predictive
self.to(self.device)

def forward(self, x):
Expand Down