#!/usr/bin/env python3
# Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
r"""
Cross-validation utilities using batch evaluation mode.
"""
from __future__ import annotations
from typing import Any, Dict, NamedTuple, Optional, Type
import torch
from botorch.fit import fit_gpytorch_model
from botorch.models.gpytorch import GPyTorchModel
from botorch.optim.utils import _filter_kwargs
from botorch.posteriors.gpytorch import GPyTorchPosterior
from gpytorch.mlls.marginal_log_likelihood import MarginalLogLikelihood
from torch import Tensor
[docs]class CVFolds(NamedTuple):
train_X: Tensor
test_X: Tensor
train_Y: Tensor
test_Y: Tensor
train_Yvar: Optional[Tensor] = None
test_Yvar: Optional[Tensor] = None
[docs]class CVResults(NamedTuple):
model: GPyTorchModel
posterior: GPyTorchPosterior
observed_Y: Tensor
observed_Yvar: Optional[Tensor] = None
[docs]def gen_loo_cv_folds(
train_X: Tensor, train_Y: Tensor, train_Yvar: Optional[Tensor] = None
) -> CVFolds:
r"""Generate LOO CV folds w.r.t. to `n`.
Args:
train_X: A `n x d` or `batch_shape x n x d` (batch mode) tensor of training
features.
train_Y: A `n x (m)` or `batch_shape x n x (m)` (batch mode) tensor of
training observations.
train_Yvar: A `batch_shape x n x (m)` or `batch_shape x n x (m)`
(batch mode) tensor of observed measurement noise.
Returns:
CVFolds tuple with the following fields
- train_X: A `n x (n-1) x d` or `batch_shape x n x (n-1) x d` tensor of
training features.
- test_X: A `n x 1 x d` or `batch_shape x n x 1 x d` tensor of test features.
- train_Y: A `n x (n-1) x m` or `batch_shape x n x (n-1) x m` tensor of
training observations.
- test_Y: A `n x 1 x m` or `batch_shape x n x 1 x m` tensor of test
observations.
- train_Yvar: A `n x (n-1) x m` or `batch_shape x n x (n-1) x m` tensor
of observed measurement noise.
- test_Yvar: A `n x 1 x m` or `batch_shape x n x 1 x m` tensor of observed
measurement noise.
Example:
>>> train_X = torch.rand(10, 1)
>>> train_Y = torch.sin(6 * train_X) + 0.2 * torch.rand_like(train_X)
>>> cv_folds = gen_loo_cv_folds(train_X, train_Y)
"""
masks = torch.eye(train_X.shape[-2], dtype=torch.uint8, device=train_X.device)
masks = masks.to(dtype=torch.bool)
if train_Y.dim() < train_X.dim():
# add output dimension
train_Y = train_Y.unsqueeze(-1)
if train_Yvar is not None:
train_Yvar = train_Yvar.unsqueeze(-1)
train_X_cv = torch.cat(
[train_X[..., ~m, :].unsqueeze(dim=-3) for m in masks], dim=-3
)
test_X_cv = torch.cat([train_X[..., m, :].unsqueeze(dim=-3) for m in masks], dim=-3)
train_Y_cv = torch.cat(
[train_Y[..., ~m, :].unsqueeze(dim=-3) for m in masks], dim=-3
)
test_Y_cv = torch.cat([train_Y[..., m, :].unsqueeze(dim=-3) for m in masks], dim=-3)
if train_Yvar is None:
train_Yvar_cv = None
test_Yvar_cv = None
else:
train_Yvar_cv = torch.cat(
[train_Yvar[..., ~m, :].unsqueeze(dim=-3) for m in masks], dim=-3
)
test_Yvar_cv = torch.cat(
[train_Yvar[..., m, :].unsqueeze(dim=-3) for m in masks], dim=-3
)
return CVFolds(
train_X=train_X_cv,
test_X=test_X_cv,
train_Y=train_Y_cv,
test_Y=test_Y_cv,
train_Yvar=train_Yvar_cv,
test_Yvar=test_Yvar_cv,
)
[docs]def batch_cross_validation(
model_cls: Type[GPyTorchModel],
mll_cls: Type[MarginalLogLikelihood],
cv_folds: CVFolds,
fit_args: Optional[Dict[str, Any]] = None,
observation_noise: bool = False,
) -> CVResults:
r"""Perform cross validation by using gpytorch batch mode.
Args:
model_cls: A GPyTorchModel class. This class must initialize the likelihood
internally. Note: Multi-task GPs are not currently supported.
mll_cls: A MarginalLogLikelihood class.
cv_folds: A CVFolds tuple.
fit_args: Arguments passed along to fit_gpytorch_model
Returns:
A CVResults tuple with the following fields
- model: GPyTorchModel for batched cross validation
- posterior: GPyTorchPosterior where the mean has shape `n x 1 x m` or
`batch_shape x n x 1 x m`
- observed_Y: A `n x 1 x m` or `batch_shape x n x 1 x m` tensor of observations.
- observed_Yvar: A `n x 1 x m` or `batch_shape x n x 1 x m` tensor of observed
measurement noise.
Example:
>>> train_X = torch.rand(10, 1)
>>> train_Y = torch.sin(6 * train_X) + 0.2 * torch.rand_like(train_X)
>>> cv_folds = gen_loo_cv_folds(train_X, train_Y)
>>> cv_results = batch_cross_validation(
>>> SingleTaskGP,
>>> ExactMarginalLogLikelihood,
>>> cv_folds,
>>> )
WARNING: This function is currently very memory inefficient, use it only
for problems of small size.
"""
fit_args = fit_args or {}
kwargs = {
"train_X": cv_folds.train_X,
"train_Y": cv_folds.train_Y,
"train_Yvar": cv_folds.train_Yvar,
}
model_cv = model_cls(**_filter_kwargs(model_cls, **kwargs))
mll_cv = mll_cls(model_cv.likelihood, model_cv)
mll_cv.to(cv_folds.train_X)
mll_cv = fit_gpytorch_model(mll_cv, **fit_args)
# Evaluate on the hold-out set in batch mode
with torch.no_grad():
posterior = model_cv.posterior(
cv_folds.test_X, observation_noise=observation_noise
)
return CVResults(
model=model_cv,
posterior=posterior,
observed_Y=cv_folds.test_Y,
observed_Yvar=cv_folds.test_Yvar,
)
