optimizer_kwargs = {} if optimizer_kwargs is None else optimizer_kwargs.copy()
max_fit_time = kwargs.pop("max_fit_time", self.max_fit_time)
if max_fit_time is not None:
+ if "options" not in optimizer_kwargs:
+ optimizer_kwargs["options"] = {}
+
# figure out how long evaluating a single samp
starttime = time.time()
_ = mll(self(train_x), train_y)
@@ -389,7 +392,8 @@ Source code for aepsych.models.base
time.time() - starttime + 1e-6
) # add an epsilon to avoid divide by zero
n_eval = int(max_fit_time / single_eval_time)
- optimizer_kwargs["options"] = {"maxfun": n_eval}
+
+ optimizer_kwargs["options"]["maxfun"] = n_eval
logger.info(f"fit maxfun is {n_eval}")
starttime = time.time()
diff --git a/api/_modules/aepsych/models/base/index.html b/api/_modules/aepsych/models/base/index.html
index 36b1b115a..9c4550e35 100644
--- a/api/_modules/aepsych/models/base/index.html
+++ b/api/_modules/aepsych/models/base/index.html
@@ -382,6 +382,9 @@ Source code for aepsych.models.base
optimizer_kwargs = {} if optimizer_kwargs is None else optimizer_kwargs.copy()
max_fit_time = kwargs.pop("max_fit_time", self.max_fit_time)
if max_fit_time is not None:
+ if "options" not in optimizer_kwargs:
+ optimizer_kwargs["options"] = {}
+
# figure out how long evaluating a single samp
starttime = time.time()
_ = mll(self(train_x), train_y)
@@ -389,7 +392,8 @@ Source code for aepsych.models.base
time.time() - starttime + 1e-6
) # add an epsilon to avoid divide by zero
n_eval = int(max_fit_time / single_eval_time)
- optimizer_kwargs["options"] = {"maxfun": n_eval}
+
+ optimizer_kwargs["options"]["maxfun"] = n_eval
logger.info(f"fit maxfun is {n_eval}")
starttime = time.time()
diff --git a/api/_modules/aepsych/models/gp_classification.html b/api/_modules/aepsych/models/gp_classification.html
index b2db98c53..bb9511863 100644
--- a/api/_modules/aepsych/models/gp_classification.html
+++ b/api/_modules/aepsych/models/gp_classification.html
@@ -26,7 +26,7 @@ Source code for aepsych.models.gp_classification
import warnings
from copy import deepcopy
-from typing import Optional, Tuple, Union
+from typing import Any, Dict, Optional, Tuple
import gpytorch
import numpy as np
@@ -35,7 +35,7 @@ Source code for aepsych.models.gp_classification
from aepsych.factory.default import default_mean_covar_factory
from aepsych.models.base import AEPsychModelDeviceMixin
from aepsych.models.utils import select_inducing_points
-from aepsych.utils import _process_bounds, promote_0d
+from aepsych.utils import _process_bounds, get_optimizer_options, promote_0d
from aepsych.utils_logging import getLogger
from gpytorch.likelihoods import BernoulliLikelihood, BetaLikelihood, Likelihood
from gpytorch.models import ApproximateGP
@@ -75,6 +75,7 @@ Source code for aepsych.models.gp_classification
inducing_size: Optional[int] = None,
max_fit_time: Optional[float] = None,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
"""Initialize the GP Classification model
@@ -96,12 +97,17 @@ Source code for aepsych.models.gp_classification
If "pivoted_chol", selects points based on the pivoted Cholesky heuristic.
If "kmeans++", selects points by performing kmeans++ clustering on the training data.
If "auto", tries to determine the best method automatically.
+ optimizer_options (Dict[str, Any], optional): Optimizer options to pass to the SciPy optimizer during
+ fitting. Assumes we are using L-BFGS-B.
"""
lb, ub, self.dim = _process_bounds(lb, ub, dim)
-
self.max_fit_time = max_fit_time
self.inducing_size = inducing_size or 99
+ self.optimizer_options = (
+ {"options": optimizer_options} if optimizer_options else {"options": {}}
+ )
+
if self.inducing_size >= 100:
logger.warning(
(
@@ -192,6 +198,8 @@ Source code for aepsych.models.gp_classification
else:
likelihood = None # fall back to __init__ default
+ optimizer_options = get_optimizer_options(config, classname)
+
return cls(
lb=lb,
ub=ub,
@@ -202,6 +210,7 @@ Source code for aepsych.models.gp_classification
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
likelihood=likelihood,
+ optimizer_options=optimizer_options,
)
def _reset_hyperparameters(self) -> None:
@@ -269,7 +278,10 @@ Source code for aepsych.models.gp_classification
n = train_y.shape[0]
mll = gpytorch.mlls.VariationalELBO(self.likelihood, self, n)
- self._fit_mll(mll, **kwargs)
+ if "optimizer_kwargs" in kwargs:
+ self._fit_mll(mll, **kwargs)
+ else:
+ self._fit_mll(mll, optimizer_kwargs=self.optimizer_options, **kwargs)
[docs] def sample(self, x: torch.Tensor, num_samples: int) -> torch.Tensor:
"""Sample from underlying model.
@@ -353,6 +365,7 @@ Source code for aepsych.models.gp_classification
inducing_size: Optional[int] = None,
max_fit_time: Optional[float] = None,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
if likelihood is None:
likelihood = BetaLikelihood()
@@ -366,6 +379,7 @@ Source code for aepsych.models.gp_classification
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
diff --git a/api/_modules/aepsych/models/gp_classification/index.html b/api/_modules/aepsych/models/gp_classification/index.html
index b2db98c53..bb9511863 100644
--- a/api/_modules/aepsych/models/gp_classification/index.html
+++ b/api/_modules/aepsych/models/gp_classification/index.html
@@ -26,7 +26,7 @@ Source code for aepsych.models.gp_classification
import warnings
from copy import deepcopy
-from typing import Optional, Tuple, Union
+from typing import Any, Dict, Optional, Tuple
import gpytorch
import numpy as np
@@ -35,7 +35,7 @@ Source code for aepsych.models.gp_classification
from aepsych.factory.default import default_mean_covar_factory
from aepsych.models.base import AEPsychModelDeviceMixin
from aepsych.models.utils import select_inducing_points
-from aepsych.utils import _process_bounds, promote_0d
+from aepsych.utils import _process_bounds, get_optimizer_options, promote_0d
from aepsych.utils_logging import getLogger
from gpytorch.likelihoods import BernoulliLikelihood, BetaLikelihood, Likelihood
from gpytorch.models import ApproximateGP
@@ -75,6 +75,7 @@ Source code for aepsych.models.gp_classification
inducing_size: Optional[int] = None,
max_fit_time: Optional[float] = None,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
"""Initialize the GP Classification model
@@ -96,12 +97,17 @@ Source code for aepsych.models.gp_classification
If "pivoted_chol", selects points based on the pivoted Cholesky heuristic.
If "kmeans++", selects points by performing kmeans++ clustering on the training data.
If "auto", tries to determine the best method automatically.
+ optimizer_options (Dict[str, Any], optional): Optimizer options to pass to the SciPy optimizer during
+ fitting. Assumes we are using L-BFGS-B.
"""
lb, ub, self.dim = _process_bounds(lb, ub, dim)
-
self.max_fit_time = max_fit_time
self.inducing_size = inducing_size or 99
+ self.optimizer_options = (
+ {"options": optimizer_options} if optimizer_options else {"options": {}}
+ )
+
if self.inducing_size >= 100:
logger.warning(
(
@@ -192,6 +198,8 @@ Source code for aepsych.models.gp_classification
else:
likelihood = None # fall back to __init__ default
+ optimizer_options = get_optimizer_options(config, classname)
+
return cls(
lb=lb,
ub=ub,
@@ -202,6 +210,7 @@ Source code for aepsych.models.gp_classification
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
likelihood=likelihood,
+ optimizer_options=optimizer_options,
)
def _reset_hyperparameters(self) -> None:
@@ -269,7 +278,10 @@ Source code for aepsych.models.gp_classification
n = train_y.shape[0]
mll = gpytorch.mlls.VariationalELBO(self.likelihood, self, n)
- self._fit_mll(mll, **kwargs)
+ if "optimizer_kwargs" in kwargs:
+ self._fit_mll(mll, **kwargs)
+ else:
+ self._fit_mll(mll, optimizer_kwargs=self.optimizer_options, **kwargs)
[docs] def sample(self, x: torch.Tensor, num_samples: int) -> torch.Tensor:
"""Sample from underlying model.
@@ -353,6 +365,7 @@ Source code for aepsych.models.gp_classification
inducing_size: Optional[int] = None,
max_fit_time: Optional[float] = None,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
if likelihood is None:
likelihood = BetaLikelihood()
@@ -366,6 +379,7 @@ Source code for aepsych.models.gp_classification
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
diff --git a/api/_modules/aepsych/models/gp_regression.html b/api/_modules/aepsych/models/gp_regression.html
index bd316737b..c9f4c0953 100644
--- a/api/_modules/aepsych/models/gp_regression.html
+++ b/api/_modules/aepsych/models/gp_regression.html
@@ -25,7 +25,7 @@ Source code for aepsych.models.gp_regression
from __future__ import annotations
from copy import deepcopy
-from typing import Dict, Optional, Tuple, Union
+from typing import Any, Dict, Optional, Tuple, Union
import gpytorch
import numpy as np
@@ -33,7 +33,7 @@ Source code for aepsych.models.gp_regression
from aepsych.config import Config
from aepsych.factory.default import default_mean_covar_factory
from aepsych.models.base import AEPsychModelDeviceMixin
-from aepsych.utils import _process_bounds, promote_0d
+from aepsych.utils import _process_bounds, get_optimizer_options, promote_0d
from aepsych.utils_logging import getLogger
from gpytorch.likelihoods import GaussianLikelihood, Likelihood
from gpytorch.models import ExactGP
@@ -58,6 +58,7 @@ Source code for aepsych.models.gp_regression
covar_module: Optional[gpytorch.kernels.Kernel] = None,
likelihood: Optional[Likelihood] = None,
max_fit_time: Optional[float] = None,
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
"""Initialize the GP regression model
@@ -73,6 +74,8 @@ Source code for aepsych.models.gp_regression
Gaussian likelihood.
max_fit_time (float, optional): The maximum amount of time, in seconds, to spend fitting the model. If None,
there is no limit to the fitting time.
+ optimizer_options (Dict[str, Any], optional): Optimizer options to pass to the SciPy optimizer during
+ fitting. Assumes we are using L-BFGS-B.
"""
if likelihood is None:
likelihood = GaussianLikelihood()
@@ -82,6 +85,10 @@ Source code for aepsych.models.gp_regression
lb, ub, self.dim = _process_bounds(lb, ub, dim)
self.max_fit_time = max_fit_time
+ self.optimizer_options = (
+ {"options": optimizer_options} if optimizer_options else {"options": {}}
+ )
+
if mean_module is None or covar_module is None:
default_mean, default_covar = default_mean_covar_factory(
dim=self.dim, stimuli_per_trial=self.stimuli_per_trial
@@ -123,6 +130,8 @@ Source code for aepsych.models.gp_regression
max_fit_time = config.getfloat(classname, "max_fit_time", fallback=None)
+ optimizer_options = get_optimizer_options(config, classname)
+
return {
"lb": lb,
"ub": ub,
@@ -131,6 +140,7 @@ Source code for aepsych.models.gp_regression
"covar_module": covar,
"likelihood": likelihood,
"max_fit_time": max_fit_time,
+ "optimizer_options": optimizer_options,
}
[docs] @classmethod
@@ -160,7 +170,7 @@ Source code for aepsych.models.gp_regression
"""
self.set_train_data(train_x, train_y)
mll = gpytorch.mlls.ExactMarginalLogLikelihood(self.likelihood, self)
- return self._fit_mll(mll, **kwargs)
+ return self._fit_mll(mll, self.optimizer_options, **kwargs)
[docs] def sample(self, x: torch.Tensor, num_samples: int) -> torch.Tensor:
"""Sample from underlying model.
diff --git a/api/_modules/aepsych/models/gp_regression/index.html b/api/_modules/aepsych/models/gp_regression/index.html
index bd316737b..c9f4c0953 100644
--- a/api/_modules/aepsych/models/gp_regression/index.html
+++ b/api/_modules/aepsych/models/gp_regression/index.html
@@ -25,7 +25,7 @@ Source code for aepsych.models.gp_regression
from __future__ import annotations
from copy import deepcopy
-from typing import Dict, Optional, Tuple, Union
+from typing import Any, Dict, Optional, Tuple, Union
import gpytorch
import numpy as np
@@ -33,7 +33,7 @@ Source code for aepsych.models.gp_regression
from aepsych.config import Config
from aepsych.factory.default import default_mean_covar_factory
from aepsych.models.base import AEPsychModelDeviceMixin
-from aepsych.utils import _process_bounds, promote_0d
+from aepsych.utils import _process_bounds, get_optimizer_options, promote_0d
from aepsych.utils_logging import getLogger
from gpytorch.likelihoods import GaussianLikelihood, Likelihood
from gpytorch.models import ExactGP
@@ -58,6 +58,7 @@ Source code for aepsych.models.gp_regression
covar_module: Optional[gpytorch.kernels.Kernel] = None,
likelihood: Optional[Likelihood] = None,
max_fit_time: Optional[float] = None,
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
"""Initialize the GP regression model
@@ -73,6 +74,8 @@ Source code for aepsych.models.gp_regression
Gaussian likelihood.
max_fit_time (float, optional): The maximum amount of time, in seconds, to spend fitting the model. If None,
there is no limit to the fitting time.
+ optimizer_options (Dict[str, Any], optional): Optimizer options to pass to the SciPy optimizer during
+ fitting. Assumes we are using L-BFGS-B.
"""
if likelihood is None:
likelihood = GaussianLikelihood()
@@ -82,6 +85,10 @@ Source code for aepsych.models.gp_regression
lb, ub, self.dim = _process_bounds(lb, ub, dim)
self.max_fit_time = max_fit_time
+ self.optimizer_options = (
+ {"options": optimizer_options} if optimizer_options else {"options": {}}
+ )
+
if mean_module is None or covar_module is None:
default_mean, default_covar = default_mean_covar_factory(
dim=self.dim, stimuli_per_trial=self.stimuli_per_trial
@@ -123,6 +130,8 @@ Source code for aepsych.models.gp_regression
max_fit_time = config.getfloat(classname, "max_fit_time", fallback=None)
+ optimizer_options = get_optimizer_options(config, classname)
+
return {
"lb": lb,
"ub": ub,
@@ -131,6 +140,7 @@ Source code for aepsych.models.gp_regression
"covar_module": covar,
"likelihood": likelihood,
"max_fit_time": max_fit_time,
+ "optimizer_options": optimizer_options,
}
[docs] @classmethod
@@ -160,7 +170,7 @@ Source code for aepsych.models.gp_regression
"""
self.set_train_data(train_x, train_y)
mll = gpytorch.mlls.ExactMarginalLogLikelihood(self.likelihood, self)
- return self._fit_mll(mll, **kwargs)
+ return self._fit_mll(mll, self.optimizer_options, **kwargs)
[docs] def sample(self, x: torch.Tensor, num_samples: int) -> torch.Tensor:
"""Sample from underlying model.
diff --git a/api/_modules/aepsych/models/monotonic_projection_gp.html b/api/_modules/aepsych/models/monotonic_projection_gp.html
index 7fbe1636e..76c34f47f 100644
--- a/api/_modules/aepsych/models/monotonic_projection_gp.html
+++ b/api/_modules/aepsych/models/monotonic_projection_gp.html
@@ -25,7 +25,7 @@ Source code for aepsych.models.monotonic_projection_gp
from __future__ import annotations
-from typing import Any, List, Optional, Union
+from typing import Any, Dict, List, Optional, Union
import gpytorch
import numpy as np
@@ -33,6 +33,7 @@ Source code for aepsych.models.monotonic_projection_gp
from aepsych.config import Config
from aepsych.factory.default import default_mean_covar_factory
from aepsych.models.gp_classification import GPClassificationModel
+from aepsych.utils import get_optimizer_options
from botorch.posteriors.gpytorch import GPyTorchPosterior
from gpytorch.likelihoods import Likelihood
from statsmodels.stats.moment_helpers import corr2cov, cov2corr
@@ -122,6 +123,7 @@ Source code for aepsych.models.monotonic_projection_gp
inducing_size: Optional[int] = None,
max_fit_time: Optional[float] = None,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
assert len(monotonic_dims) > 0
self.monotonic_dims = [int(d) for d in monotonic_dims]
@@ -137,6 +139,7 @@ Source code for aepsych.models.monotonic_projection_gp
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
[docs] def posterior(
@@ -240,6 +243,8 @@ Source code for aepsych.models.monotonic_projection_gp
)
min_f_val = config.getfloat(classname, "min_f_val", fallback=None)
+ optimizer_options = get_optimizer_options(config, classname)
+
return cls(
lb=lb,
ub=ub,
@@ -253,6 +258,7 @@ Source code for aepsych.models.monotonic_projection_gp
monotonic_dims=monotonic_dims,
monotonic_grid_size=monotonic_grid_size,
min_f_val=min_f_val,
+ optimizer_options=optimizer_options,
)
diff --git a/api/_modules/aepsych/models/monotonic_projection_gp/index.html b/api/_modules/aepsych/models/monotonic_projection_gp/index.html
index 7fbe1636e..76c34f47f 100644
--- a/api/_modules/aepsych/models/monotonic_projection_gp/index.html
+++ b/api/_modules/aepsych/models/monotonic_projection_gp/index.html
@@ -25,7 +25,7 @@ Source code for aepsych.models.monotonic_projection_gp
from __future__ import annotations
-from typing import Any, List, Optional, Union
+from typing import Any, Dict, List, Optional, Union
import gpytorch
import numpy as np
@@ -33,6 +33,7 @@ Source code for aepsych.models.monotonic_projection_gp
from aepsych.config import Config
from aepsych.factory.default import default_mean_covar_factory
from aepsych.models.gp_classification import GPClassificationModel
+from aepsych.utils import get_optimizer_options
from botorch.posteriors.gpytorch import GPyTorchPosterior
from gpytorch.likelihoods import Likelihood
from statsmodels.stats.moment_helpers import corr2cov, cov2corr
@@ -122,6 +123,7 @@ Source code for aepsych.models.monotonic_projection_gp
inducing_size: Optional[int] = None,
max_fit_time: Optional[float] = None,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
assert len(monotonic_dims) > 0
self.monotonic_dims = [int(d) for d in monotonic_dims]
@@ -137,6 +139,7 @@ Source code for aepsych.models.monotonic_projection_gp
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
[docs] def posterior(
@@ -240,6 +243,8 @@ Source code for aepsych.models.monotonic_projection_gp
)
min_f_val = config.getfloat(classname, "min_f_val", fallback=None)
+ optimizer_options = get_optimizer_options(config, classname)
+
return cls(
lb=lb,
ub=ub,
@@ -253,6 +258,7 @@ Source code for aepsych.models.monotonic_projection_gp
monotonic_dims=monotonic_dims,
monotonic_grid_size=monotonic_grid_size,
min_f_val=min_f_val,
+ optimizer_options=optimizer_options,
)
diff --git a/api/_modules/aepsych/models/monotonic_rejection_gp.html b/api/_modules/aepsych/models/monotonic_rejection_gp.html
index c24e9424f..e7f9d7822 100644
--- a/api/_modules/aepsych/models/monotonic_rejection_gp.html
+++ b/api/_modules/aepsych/models/monotonic_rejection_gp.html
@@ -26,7 +26,7 @@ Source code for aepsych.models.monotonic_rejection_gp
from __future__ import annotations
import warnings
-from typing import Dict, List, Optional, Sequence, Tuple, Union
+from typing import Any, Dict, List, Optional, Sequence, Tuple, Union
import gpytorch
import numpy as np
@@ -38,7 +38,7 @@ Source code for aepsych.models.monotonic_rejection_gp
from aepsych.means.constant_partial_grad import ConstantMeanPartialObsGrad
from aepsych.models.base import AEPsychMixin
from aepsych.models.utils import select_inducing_points
-from aepsych.utils import _process_bounds, promote_0d
+from aepsych.utils import _process_bounds, get_optimizer_options, promote_0d
from botorch.fit import fit_gpytorch_mll
from gpytorch.kernels import Kernel
from gpytorch.likelihoods import BernoulliLikelihood, Likelihood
@@ -81,6 +81,7 @@ Source code for aepsych.models.monotonic_rejection_gp
num_samples: int = 250,
num_rejection_samples: int = 5000,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
"""Initialize MonotonicRejectionGP.
@@ -100,6 +101,8 @@ Source code for aepsych.models.monotonic_rejection_gp
acqf (MonotonicMCAcquisition, optional): Acquisition function to use for querying points. Defaults to MonotonicMCLSE.
objective (Optional[MCAcquisitionObjective], optional): Transformation of GP to apply before computing acquisition function. Defaults to identity transform for gaussian likelihood, probit transform for probit-bernoulli.
extra_acqf_args (Optional[Dict[str, object]], optional): Additional arguments to pass into the acquisition function. Defaults to None.
+ optimizer_options (Dict[str, Any], optional): Optimizer options to pass to the SciPy optimizer during
+ fitting. Assumes we are using L-BFGS-B.
"""
self.lb, self.ub, self.dim = _process_bounds(lb, ub, dim)
if likelihood is None:
@@ -163,6 +166,9 @@ Source code for aepsych.models.monotonic_rejection_gp
self.num_rejection_samples = num_rejection_samples
self.fixed_prior_mean = fixed_prior_mean
self.inducing_points = inducing_points
+ self.optimizer_options = (
+ {"options": optimizer_options} if optimizer_options else {"options": {}}
+ )
[docs] def fit(self, train_x: Tensor, train_y: Tensor, **kwargs) -> None:
"""Fit the model
@@ -201,7 +207,7 @@ Source code for aepsych.models.monotonic_rejection_gp
mll = VariationalELBO(
likelihood=self.likelihood, model=self, num_data=train_y.numel()
)
- mll = fit_gpytorch_mll(mll)
+ mll = fit_gpytorch_mll(mll, optimizer_kwargs=self.optimizer_options)
[docs] def update(self, train_x: Tensor, train_y: Tensor, warmstart: bool = True) -> None:
"""
@@ -337,6 +343,8 @@ Source code for aepsych.models.monotonic_rejection_gp
classname, "monotonic_idxs", fallback=[-1]
)
+ optimizer_options = get_optimizer_options(config, classname)
+
return cls(
monotonic_idxs=monotonic_idxs,
lb=lb,
@@ -347,6 +355,7 @@ Source code for aepsych.models.monotonic_rejection_gp
num_rejection_samples=num_rejection_samples,
mean_module=mean,
covar_module=covar,
+ optimizer_options=optimizer_options,
)
[docs] def forward(self, x: torch.Tensor) -> gpytorch.distributions.MultivariateNormal:
diff --git a/api/_modules/aepsych/models/monotonic_rejection_gp/index.html b/api/_modules/aepsych/models/monotonic_rejection_gp/index.html
index c24e9424f..e7f9d7822 100644
--- a/api/_modules/aepsych/models/monotonic_rejection_gp/index.html
+++ b/api/_modules/aepsych/models/monotonic_rejection_gp/index.html
@@ -26,7 +26,7 @@ Source code for aepsych.models.monotonic_rejection_gp
from __future__ import annotations
import warnings
-from typing import Dict, List, Optional, Sequence, Tuple, Union
+from typing import Any, Dict, List, Optional, Sequence, Tuple, Union
import gpytorch
import numpy as np
@@ -38,7 +38,7 @@ Source code for aepsych.models.monotonic_rejection_gp
from aepsych.means.constant_partial_grad import ConstantMeanPartialObsGrad
from aepsych.models.base import AEPsychMixin
from aepsych.models.utils import select_inducing_points
-from aepsych.utils import _process_bounds, promote_0d
+from aepsych.utils import _process_bounds, get_optimizer_options, promote_0d
from botorch.fit import fit_gpytorch_mll
from gpytorch.kernels import Kernel
from gpytorch.likelihoods import BernoulliLikelihood, Likelihood
@@ -81,6 +81,7 @@ Source code for aepsych.models.monotonic_rejection_gp
num_samples: int = 250,
num_rejection_samples: int = 5000,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
"""Initialize MonotonicRejectionGP.
@@ -100,6 +101,8 @@ Source code for aepsych.models.monotonic_rejection_gp
acqf (MonotonicMCAcquisition, optional): Acquisition function to use for querying points. Defaults to MonotonicMCLSE.
objective (Optional[MCAcquisitionObjective], optional): Transformation of GP to apply before computing acquisition function. Defaults to identity transform for gaussian likelihood, probit transform for probit-bernoulli.
extra_acqf_args (Optional[Dict[str, object]], optional): Additional arguments to pass into the acquisition function. Defaults to None.
+ optimizer_options (Dict[str, Any], optional): Optimizer options to pass to the SciPy optimizer during
+ fitting. Assumes we are using L-BFGS-B.
"""
self.lb, self.ub, self.dim = _process_bounds(lb, ub, dim)
if likelihood is None:
@@ -163,6 +166,9 @@ Source code for aepsych.models.monotonic_rejection_gp
self.num_rejection_samples = num_rejection_samples
self.fixed_prior_mean = fixed_prior_mean
self.inducing_points = inducing_points
+ self.optimizer_options = (
+ {"options": optimizer_options} if optimizer_options else {"options": {}}
+ )
[docs] def fit(self, train_x: Tensor, train_y: Tensor, **kwargs) -> None:
"""Fit the model
@@ -201,7 +207,7 @@ Source code for aepsych.models.monotonic_rejection_gp
mll = VariationalELBO(
likelihood=self.likelihood, model=self, num_data=train_y.numel()
)
- mll = fit_gpytorch_mll(mll)
+ mll = fit_gpytorch_mll(mll, optimizer_kwargs=self.optimizer_options)
[docs] def update(self, train_x: Tensor, train_y: Tensor, warmstart: bool = True) -> None:
"""
@@ -337,6 +343,8 @@ Source code for aepsych.models.monotonic_rejection_gp
classname, "monotonic_idxs", fallback=[-1]
)
+ optimizer_options = get_optimizer_options(config, classname)
+
return cls(
monotonic_idxs=monotonic_idxs,
lb=lb,
@@ -347,6 +355,7 @@ Source code for aepsych.models.monotonic_rejection_gp
num_rejection_samples=num_rejection_samples,
mean_module=mean,
covar_module=covar,
+ optimizer_options=optimizer_options,
)
[docs] def forward(self, x: torch.Tensor) -> gpytorch.distributions.MultivariateNormal:
diff --git a/api/_modules/aepsych/models/pairwise_probit.html b/api/_modules/aepsych/models/pairwise_probit.html
index 55fc5a483..0b3e5c081 100644
--- a/api/_modules/aepsych/models/pairwise_probit.html
+++ b/api/_modules/aepsych/models/pairwise_probit.html
@@ -23,15 +23,14 @@ Source code for aepsych.models.pairwise_probit
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
import time
-from typing import Any, Dict, Optional, Tuple, Union
+from typing import Any, Dict, Optional, Tuple
import gpytorch
-import numpy as np
import torch
from aepsych.config import Config
from aepsych.factory import default_mean_covar_factory
from aepsych.models.base import AEPsychMixin
-from aepsych.utils import _process_bounds, promote_0d
+from aepsych.utils import _process_bounds, get_optimizer_options, promote_0d
from aepsych.utils_logging import getLogger
from botorch.fit import fit_gpytorch_mll
from botorch.models import PairwiseGP, PairwiseLaplaceMarginalLogLikelihood
@@ -82,6 +81,7 @@
Source code for aepsych.models.pairwise_probit
dim: Optional[int] = None,
covar_module: Optional[gpytorch.kernels.Kernel] = None,
max_fit_time: Optional[float] = None,
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
self.lb, self.ub, dim = _process_bounds(lb, ub, dim)
@@ -111,6 +111,9 @@
Source code for aepsych.models.pairwise_probit
)
self.dim = dim # The Pairwise constructor sets self.dim = None.
+ self.optimizer_options = (
+ {"options": optimizer_options} if optimizer_options else {"options": {}}
+ )
[docs] def fit(
self,
@@ -119,6 +122,12 @@ Source code for aepsych.models.pairwise_probit
optimizer_kwargs: Optional[Dict[str, Any]] = None,
**kwargs,
) -> None:
+ if optimizer_kwargs is not None:
+ if not "optimizer_kwargs" in optimizer_kwargs:
+ optimizer_kwargs = optimizer_kwargs.copy()
+ optimizer_kwargs.update(self.optimizer_options)
+ else:
+ optimizer_kwargs = {"options": self.optimizer_options}
self.train()
mll = PairwiseLaplaceMarginalLogLikelihood(self.likelihood, self)
datapoints, comparisons = self._pairs_to_comparisons(train_x, train_y)
@@ -127,17 +136,21 @@
Source code for aepsych.models.pairwise_probit
optimizer_kwargs = {} if optimizer_kwargs is None else optimizer_kwargs.copy()
max_fit_time = kwargs.pop("max_fit_time", self.max_fit_time)
if max_fit_time is not None:
+ if "options" not in optimizer_kwargs:
+ optimizer_kwargs["options"] = {}
+
# figure out how long evaluating a single samp
starttime = time.time()
_ = mll(self(datapoints), comparisons)
single_eval_time = time.time() - starttime
n_eval = int(max_fit_time / single_eval_time)
- optimizer_kwargs["maxfun"] = n_eval
+
+ optimizer_kwargs["options"]["maxfun"] = n_eval
logger.info(f"fit maxfun is {n_eval}")
logger.info("Starting fit...")
starttime = time.time()
- fit_gpytorch_mll(mll, **kwargs, **optimizer_kwargs)
+ fit_gpytorch_mll(mll, optimizer_kwargs=optimizer_kwargs, **kwargs)
logger.info(f"Fit done, time={time.time()-starttime}")
[docs] def update(
@@ -227,7 +240,16 @@ Source code for aepsych.models.pairwise_probit
max_fit_time = config.getfloat(classname, "max_fit_time", fallback=None)
- return cls(lb=lb, ub=ub, dim=dim, covar_module=covar, max_fit_time=max_fit_time)
+ optimizer_options = get_optimizer_options(config, classname)
+
+ return cls(
+ lb=lb,
+ ub=ub,
+ dim=dim,
+ covar_module=covar,
+ max_fit_time=max_fit_time,
+ optimizer_options=optimizer_options,
+ )
diff --git a/api/_modules/aepsych/models/pairwise_probit/index.html b/api/_modules/aepsych/models/pairwise_probit/index.html
index 55fc5a483..0b3e5c081 100644
--- a/api/_modules/aepsych/models/pairwise_probit/index.html
+++ b/api/_modules/aepsych/models/pairwise_probit/index.html
@@ -23,15 +23,14 @@ Source code for aepsych.models.pairwise_probit
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
import time
-from typing import Any, Dict, Optional, Tuple, Union
+from typing import Any, Dict, Optional, Tuple
import gpytorch
-import numpy as np
import torch
from aepsych.config import Config
from aepsych.factory import default_mean_covar_factory
from aepsych.models.base import AEPsychMixin
-from aepsych.utils import _process_bounds, promote_0d
+from aepsych.utils import _process_bounds, get_optimizer_options, promote_0d
from aepsych.utils_logging import getLogger
from botorch.fit import fit_gpytorch_mll
from botorch.models import PairwiseGP, PairwiseLaplaceMarginalLogLikelihood
@@ -82,6 +81,7 @@
Source code for aepsych.models.pairwise_probit
dim: Optional[int] = None,
covar_module: Optional[gpytorch.kernels.Kernel] = None,
max_fit_time: Optional[float] = None,
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
self.lb, self.ub, dim = _process_bounds(lb, ub, dim)
@@ -111,6 +111,9 @@
Source code for aepsych.models.pairwise_probit
)
self.dim = dim # The Pairwise constructor sets self.dim = None.
+ self.optimizer_options = (
+ {"options": optimizer_options} if optimizer_options else {"options": {}}
+ )
[docs] def fit(
self,
@@ -119,6 +122,12 @@ Source code for aepsych.models.pairwise_probit
optimizer_kwargs: Optional[Dict[str, Any]] = None,
**kwargs,
) -> None:
+ if optimizer_kwargs is not None:
+ if not "optimizer_kwargs" in optimizer_kwargs:
+ optimizer_kwargs = optimizer_kwargs.copy()
+ optimizer_kwargs.update(self.optimizer_options)
+ else:
+ optimizer_kwargs = {"options": self.optimizer_options}
self.train()
mll = PairwiseLaplaceMarginalLogLikelihood(self.likelihood, self)
datapoints, comparisons = self._pairs_to_comparisons(train_x, train_y)
@@ -127,17 +136,21 @@
Source code for aepsych.models.pairwise_probit
optimizer_kwargs = {} if optimizer_kwargs is None else optimizer_kwargs.copy()
max_fit_time = kwargs.pop("max_fit_time", self.max_fit_time)
if max_fit_time is not None:
+ if "options" not in optimizer_kwargs:
+ optimizer_kwargs["options"] = {}
+
# figure out how long evaluating a single samp
starttime = time.time()
_ = mll(self(datapoints), comparisons)
single_eval_time = time.time() - starttime
n_eval = int(max_fit_time / single_eval_time)
- optimizer_kwargs["maxfun"] = n_eval
+
+ optimizer_kwargs["options"]["maxfun"] = n_eval
logger.info(f"fit maxfun is {n_eval}")
logger.info("Starting fit...")
starttime = time.time()
- fit_gpytorch_mll(mll, **kwargs, **optimizer_kwargs)
+ fit_gpytorch_mll(mll, optimizer_kwargs=optimizer_kwargs, **kwargs)
logger.info(f"Fit done, time={time.time()-starttime}")
[docs] def update(
@@ -227,7 +240,16 @@ Source code for aepsych.models.pairwise_probit
max_fit_time = config.getfloat(classname, "max_fit_time", fallback=None)
- return cls(lb=lb, ub=ub, dim=dim, covar_module=covar, max_fit_time=max_fit_time)
+ optimizer_options = get_optimizer_options(config, classname)
+
+ return cls(
+ lb=lb,
+ ub=ub,
+ dim=dim,
+ covar_module=covar,
+ max_fit_time=max_fit_time,
+ optimizer_options=optimizer_options,
+ )
diff --git a/api/_modules/aepsych/models/semi_p.html b/api/_modules/aepsych/models/semi_p.html
index d8c538778..a9d551241 100644
--- a/api/_modules/aepsych/models/semi_p.html
+++ b/api/_modules/aepsych/models/semi_p.html
@@ -26,7 +26,7 @@ Source code for aepsych.models.semi_p
from __future__ import annotations
from copy import deepcopy
-from typing import Any, Optional, Tuple, Union
+from typing import Any, Dict, Optional, Tuple, Union
import gpytorch
import numpy as np
@@ -36,7 +36,7 @@ Source code for aepsych.models.semi_p
from aepsych.config import Config
from aepsych.likelihoods import BernoulliObjectiveLikelihood, LinearBernoulliLikelihood
from aepsych.models import GPClassificationModel
-from aepsych.utils import _process_bounds, promote_0d
+from aepsych.utils import _process_bounds, get_optimizer_options, promote_0d
from aepsych.utils_logging import getLogger
from botorch.acquisition.objective import PosteriorTransform
from botorch.optim.fit import fit_gpytorch_mll_scipy
@@ -207,6 +207,7 @@ Source code for aepsych.models.semi_p
inducing_size: Optional[int] = None,
max_fit_time: Optional[float] = None,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
"""
Initialize SemiParametricGP.
@@ -230,6 +231,8 @@ Source code for aepsych.models.semi_p
If "pivoted_chol", selects points based on the pivoted Cholesky heuristic.
If "kmeans++", selects points by performing kmeans++ clustering on the training data.
If "auto", tries to determine the best method automatically.
+ optimizer_options (Dict[str, Any], optional): Optimizer options to pass to the SciPy optimizer during
+ fitting. Assumes we are using L-BFGS-B.
"""
lb, ub, dim = _process_bounds(lb, ub, dim)
@@ -270,6 +273,7 @@ Source code for aepsych.models.semi_p
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
[docs] @classmethod
@@ -312,6 +316,8 @@ Source code for aepsych.models.semi_p
slope_mean = config.getfloat(classname, "slope_mean", fallback=2)
+ optimizer_options = get_optimizer_options(config, classname)
+
return cls(
lb=lb,
ub=ub,
@@ -322,6 +328,7 @@ Source code for aepsych.models.semi_p
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
[docs] def fit(
@@ -458,6 +465,7 @@ Source code for aepsych.models.semi_p
inducing_size: Optional[int] = None,
max_fit_time: Optional[float] = None,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
"""
Initialize HadamardSemiPModel.
@@ -480,6 +488,8 @@ Source code for aepsych.models.semi_p
If "pivoted_chol", selects points based on the pivoted Cholesky heuristic.
If "kmeans++", selects points by performing kmeans++ clustering on the training data.
If "auto", tries to determine the best method automatically.
+ optimizer_options (Dict[str, Any], optional): Optimizer options to pass to the SciPy optimizer during
+ fitting. Assumes we are using L-BFGS-B.
"""
super().__init__(
lb=lb,
@@ -488,6 +498,7 @@ Source code for aepsych.models.semi_p
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
self.stim_dim = stim_dim
@@ -613,6 +624,8 @@ Source code for aepsych.models.semi_p
stim_dim = config.getint(classname, "stim_dim", fallback=0)
+ optimizer_options = get_optimizer_options(config, classname)
+
return cls(
lb=lb,
ub=ub,
@@ -627,6 +640,7 @@ Source code for aepsych.models.semi_p
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
[docs] def predict(
diff --git a/api/_modules/aepsych/models/semi_p/index.html b/api/_modules/aepsych/models/semi_p/index.html
index d8c538778..a9d551241 100644
--- a/api/_modules/aepsych/models/semi_p/index.html
+++ b/api/_modules/aepsych/models/semi_p/index.html
@@ -26,7 +26,7 @@ Source code for aepsych.models.semi_p
from __future__ import annotations
from copy import deepcopy
-from typing import Any, Optional, Tuple, Union
+from typing import Any, Dict, Optional, Tuple, Union
import gpytorch
import numpy as np
@@ -36,7 +36,7 @@ Source code for aepsych.models.semi_p
from aepsych.config import Config
from aepsych.likelihoods import BernoulliObjectiveLikelihood, LinearBernoulliLikelihood
from aepsych.models import GPClassificationModel
-from aepsych.utils import _process_bounds, promote_0d
+from aepsych.utils import _process_bounds, get_optimizer_options, promote_0d
from aepsych.utils_logging import getLogger
from botorch.acquisition.objective import PosteriorTransform
from botorch.optim.fit import fit_gpytorch_mll_scipy
@@ -207,6 +207,7 @@ Source code for aepsych.models.semi_p
inducing_size: Optional[int] = None,
max_fit_time: Optional[float] = None,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
"""
Initialize SemiParametricGP.
@@ -230,6 +231,8 @@ Source code for aepsych.models.semi_p
If "pivoted_chol", selects points based on the pivoted Cholesky heuristic.
If "kmeans++", selects points by performing kmeans++ clustering on the training data.
If "auto", tries to determine the best method automatically.
+ optimizer_options (Dict[str, Any], optional): Optimizer options to pass to the SciPy optimizer during
+ fitting. Assumes we are using L-BFGS-B.
"""
lb, ub, dim = _process_bounds(lb, ub, dim)
@@ -270,6 +273,7 @@ Source code for aepsych.models.semi_p
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
[docs] @classmethod
@@ -312,6 +316,8 @@ Source code for aepsych.models.semi_p
slope_mean = config.getfloat(classname, "slope_mean", fallback=2)
+ optimizer_options = get_optimizer_options(config, classname)
+
return cls(
lb=lb,
ub=ub,
@@ -322,6 +328,7 @@ Source code for aepsych.models.semi_p
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
[docs] def fit(
@@ -458,6 +465,7 @@ Source code for aepsych.models.semi_p
inducing_size: Optional[int] = None,
max_fit_time: Optional[float] = None,
inducing_point_method: str = "auto",
+ optimizer_options: Optional[Dict[str, Any]] = None,
) -> None:
"""
Initialize HadamardSemiPModel.
@@ -480,6 +488,8 @@ Source code for aepsych.models.semi_p
If "pivoted_chol", selects points based on the pivoted Cholesky heuristic.
If "kmeans++", selects points by performing kmeans++ clustering on the training data.
If "auto", tries to determine the best method automatically.
+ optimizer_options (Dict[str, Any], optional): Optimizer options to pass to the SciPy optimizer during
+ fitting. Assumes we are using L-BFGS-B.
"""
super().__init__(
lb=lb,
@@ -488,6 +498,7 @@ Source code for aepsych.models.semi_p
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
self.stim_dim = stim_dim
@@ -613,6 +624,8 @@ Source code for aepsych.models.semi_p
stim_dim = config.getint(classname, "stim_dim", fallback=0)
+ optimizer_options = get_optimizer_options(config, classname)
+
return cls(
lb=lb,
ub=ub,
@@ -627,6 +640,7 @@ Source code for aepsych.models.semi_p
inducing_size=inducing_size,
max_fit_time=max_fit_time,
inducing_point_method=inducing_point_method,
+ optimizer_options=optimizer_options,
)
[docs] def predict(
diff --git a/api/_modules/aepsych/utils.html b/api/_modules/aepsych/utils.html
index 82ac0c359..95f1d534c 100644
--- a/api/_modules/aepsych/utils.html
+++ b/api/_modules/aepsych/utils.html
@@ -364,6 +364,34 @@ Source code for aepsych.utils
) # Choice parameters with n_choices < 3 add n_choices - 1 dims
return dim
+
+
+[docs]def get_optimizer_options(config: Config, name: str) -> Dict[str, Any]:
+ """Return the optimizer options for the model to pass to the SciPy L-BFGS-B
+ optimizer. Only the somewhat useful ones for AEPsych are searched for: maxcor,
+ ftol, gtol, maxfun, maxiter, maxls. See docs for details:
+ https://docs.scipy.org/doc/scipy/reference/optimize.minimize-lbfgsb.html#optimize-minimize-lbfgsb
+
+ Args:
+ config (Config): Config to search for options.
+ name (str): Model name to look for options for.
+
+ Return:
+ Dict[str, Any]: Dictionary of options to pass to SciPy's minimize, assuming the
+ method is L-BFGS-B.
+ """
+ options: Dict[str, Optional[Union[float, int]]] = {}
+
+ options["maxcor"] = config.getint(name, "maxcor", fallback=None)
+ options["ftol"] = config.getfloat(name, "ftol", fallback=None)
+ options["gtol"] = config.getfloat(name, "gtol", fallback=None)
+ options["maxfun"] = config.getint(name, "maxfun", fallback=None)
+ options["maxiter"] = config.getint(name, "maxiter", fallback=None)
+ options["maxls"] = config.getint(name, "maxls", fallback=None)
+
+ # Filter all the nones out, which could just come back as an empty dict
+ options = {key: value for key, value in options.items() if value is not None}
+ return options
diff --git a/api/_modules/aepsych/utils/index.html b/api/_modules/aepsych/utils/index.html
index 82ac0c359..95f1d534c 100644
--- a/api/_modules/aepsych/utils/index.html
+++ b/api/_modules/aepsych/utils/index.html
@@ -364,6 +364,34 @@ Source code for aepsych.utils
) # Choice parameters with n_choices < 3 add n_choices - 1 dims
return dim
+
+
+[docs]def get_optimizer_options(config: Config, name: str) -> Dict[str, Any]:
+ """Return the optimizer options for the model to pass to the SciPy L-BFGS-B
+ optimizer. Only the somewhat useful ones for AEPsych are searched for: maxcor,
+ ftol, gtol, maxfun, maxiter, maxls. See docs for details:
+ https://docs.scipy.org/doc/scipy/reference/optimize.minimize-lbfgsb.html#optimize-minimize-lbfgsb
+
+ Args:
+ config (Config): Config to search for options.
+ name (str): Model name to look for options for.
+
+ Return:
+ Dict[str, Any]: Dictionary of options to pass to SciPy's minimize, assuming the
+ method is L-BFGS-B.
+ """
+ options: Dict[str, Optional[Union[float, int]]] = {}
+
+ options["maxcor"] = config.getint(name, "maxcor", fallback=None)
+ options["ftol"] = config.getfloat(name, "ftol", fallback=None)
+ options["gtol"] = config.getfloat(name, "gtol", fallback=None)
+ options["maxfun"] = config.getint(name, "maxfun", fallback=None)
+ options["maxiter"] = config.getint(name, "maxiter", fallback=None)
+ options["maxls"] = config.getint(name, "maxls", fallback=None)
+
+ # Filter all the nones out, which could just come back as an empty dict
+ options = {key: value for key, value in options.items() if value is not None}
+ return options
diff --git a/api/genindex.html b/api/genindex.html
index 67a560b26..ab43fd5e6 100644
--- a/api/genindex.html
+++ b/api/genindex.html
@@ -1021,10 +1021,10 @@ G
get_dim() (in module aepsych.utils)
-
-
- get_jnd() (aepsych.models.base.AEPsychMixin method)
- (aepsych.strategy.Strategy method)
@@ -1054,6 +1054,8 @@
G
- get_next_filename() (in module aepsych.server.server)
+- get_optimizer_options() (in module aepsych.utils)
+
- get_outcome_for() (aepsych.database.db.Database method)
- get_param_for() (aepsych.database.db.Database method)
diff --git a/api/models.html b/api/models.html
index 3bc406479..94cc86ced 100644
--- a/api/models.html
+++ b/api/models.html
@@ -542,6 +542,8 @@
SubmodulesMonotonicMCAcquisition, optional) – Acquisition function to use for querying points. Defaults to MonotonicMCLSE.
objective (Optional[MCAcquisitionObjective], optional) – Transformation of GP to apply before computing acquisition function. Defaults to identity transform for gaussian likelihood, probit transform for probit-bernoulli.
extra_acqf_args (Optional[Dict[str, object]], optional) – Additional arguments to pass into the acquisition function. Defaults to None.
optimizer_options (Dict[str, Any], optional) – Optimizer options to pass to the SciPy optimizer during
+fitting. Assumes we are using L-BFGS-B.
lb (Union[np.ndarray, torch.Tensor]) –
ub (Union[np.ndarray, torch.Tensor]) –
dim (Optional[int]) –
SubmodulesMonotonicMCAcquisition, optional) – Acquisition function to use for querying points. Defaults to MonotonicMCLSE.
objective (Optional[MCAcquisitionObjective], optional) – Transformation of GP to apply before computing acquisition function. Defaults to identity transform for gaussian likelihood, probit transform for probit-bernoulli.
extra_acqf_args (Optional[Dict[str, object]], optional) – Additional arguments to pass into the acquisition function. Defaults to None.
optimizer_options (Dict[str, Any], optional) – Optimizer options to pass to the SciPy optimizer during
+fitting. Assumes we are using L-BFGS-B.
lb (Union[np.ndarray, torch.Tensor]) –
ub (Union[np.ndarray, torch.Tensor]) –
dim (Optional[int]) –
SubmodulesMonotonicMCAcquisition, optional) – Acquisition function to use for querying points. Defaults to MonotonicMCLSE.
objective (Optional[MCAcquisitionObjective], optional) – Transformation of GP to apply before computing acquisition function. Defaults to identity transform for gaussian likelihood, probit transform for probit-bernoulli.
extra_acqf_args (Optional[Dict[str, object]], optional) – Additional arguments to pass into the acquisition function. Defaults to None.
optimizer_options (Dict[str, Any], optional) – Optimizer options to pass to the SciPy optimizer during
+fitting. Assumes we are using L-BFGS-B.
lb (Union[np.ndarray, torch.Tensor]) –
ub (Union[np.ndarray, torch.Tensor]) –
dim (Optional[int]) –
SubmodulesMonotonicMCAcquisition, optional) – Acquisition function to use for querying points. Defaults to MonotonicMCLSE.
objective (Optional[MCAcquisitionObjective], optional) – Transformation of GP to apply before computing acquisition function. Defaults to identity transform for gaussian likelihood, probit transform for probit-bernoulli.
extra_acqf_args (Optional[Dict[str, object]], optional) – Additional arguments to pass into the acquisition function. Defaults to None.
optimizer_options (Dict[str, Any], optional) – Optimizer options to pass to the SciPy optimizer during
+fitting. Assumes we are using L-BFGS-B.
lb (Union[np.ndarray, torch.Tensor]) –
ub (Union[np.ndarray, torch.Tensor]) –
dim (Optional[int]) –
Submodules
+-
+aepsych.utils.get_optimizer_options(config, name)[source]¶
+Return the optimizer options for the model to pass to the SciPy L-BFGS-B
+optimizer. Only the somewhat useful ones for AEPsych are searched for: maxcor,
+ftol, gtol, maxfun, maxiter, maxls. See docs for details:
+https://docs.scipy.org/doc/scipy/reference/optimize.minimize-lbfgsb.html#optimize-minimize-lbfgsb
+
+- Parameters:
+
+config (Config) – Config to search for options.
name (str) – Model name to look for options for.
+- Returns:
+
+- Dictionary of options to pass to SciPy’s minimize, assuming the
method is L-BFGS-B.
+
+
+- Return type:
+Dict[str, Any]
+
+aepsych.utils
+-
+aepsych.utils.get_optimizer_options(config, name)[source]¶
+Return the optimizer options for the model to pass to the SciPy L-BFGS-B
+optimizer. Only the somewhat useful ones for AEPsych are searched for: maxcor,
+ftol, gtol, maxfun, maxiter, maxls. See docs for details:
+https://docs.scipy.org/doc/scipy/reference/optimize.minimize-lbfgsb.html#optimize-minimize-lbfgsb
+
+- Parameters:
+
+config (Config) – Config to search for options.
name (str) – Model name to look for options for.
+- Returns:
+
+- Dictionary of options to pass to SciPy’s minimize, assuming the
method is L-BFGS-B.
+
+
+- Return type:
+Dict[str, Any]
+
+Particle Effect Demo¶