#!/usr/bin/env python3
# Copyright (c) Facebook, Inc. and its affiliates.
# All rights reserved.
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
from typing import Any, Dict, Optional, Sequence
import torch
from aepsych.acquisition.monotonic_rejection import MonotonicMCAcquisition
from aepsych.config import Config
from aepsych.generators.base import AEPsychGenerator
from aepsych.models.monotonic_rejection_gp import MonotonicRejectionGP
from botorch.logging import logger
from botorch.optim.initializers import gen_batch_initial_conditions
from botorch.optim.utils import columnwise_clamp, fix_features
[docs]def default_loss_constraint_fun(
loss: torch.Tensor, candidates: torch.Tensor
) -> torch.Tensor:
"""Identity transform for constrained optimization.
This simply returns loss as-is. Write your own versions of this
for constrained optimization by e.g. interior point method.
Args:
loss (torch.Tensor): Value of loss at candidate points.
candidates (torch.Tensor): Location of candidate points.
Returns:
torch.Tensor: New loss (unchanged)
"""
return loss
[docs]class MonotonicRejectionGenerator(AEPsychGenerator[MonotonicRejectionGP]):
"""Generator specifically to be used with MonotonicRejectionGP, which generates new points to sample by minimizing
an acquisition function through stochastic gradient descent."""
def __init__(
self,
acqf: MonotonicMCAcquisition,
acqf_kwargs: Optional[Dict[str, Any]] = None,
model_gen_options: Optional[Dict[str, Any]] = None,
explore_features: Optional[Sequence[int]] = None,
) -> None:
"""Initialize MonotonicRejectionGenerator.
Args:
acqf (AcquisitionFunction): Acquisition function to use.
acqf_kwargs (Dict[str, object], optional): Extra arguments to
pass to acquisition function. Defaults to no arguments.
model_gen_options: Dictionary with options for generating candidate, such as
SGD parameters. See code for all options and their defaults.
explore_features: List of features that will be selected randomly and then
fixed for acquisition fn optimization.
"""
if acqf_kwargs is None:
acqf_kwargs = {}
self.acqf = acqf
self.acqf_kwargs = acqf_kwargs
self.model_gen_options = model_gen_options
self.explore_features = explore_features
def _instantiate_acquisition_fn(self, model: MonotonicRejectionGP):
return self.acqf(
model=model,
deriv_constraint_points=model._get_deriv_constraint_points(),
**self.acqf_kwargs,
)
[docs] def gen(
self,
num_points: int, # Current implementation only generates 1 point at a time
model: MonotonicRejectionGP,
):
"""Query next point(s) to run by optimizing the acquisition function.
Args:
num_points (int, optional): Number of points to query.
model (AEPsychMixin): Fitted model of the data.
Returns:
np.ndarray: Next set of point(s) to evaluate, [num_points x dim].
"""
options = self.model_gen_options or {}
num_restarts = options.get("num_restarts", 10)
raw_samples = options.get("raw_samples", 1000)
verbosity_freq = options.get("verbosity_freq", -1)
lr = options.get("lr", 0.01)
momentum = options.get("momentum", 0.9)
nesterov = options.get("nesterov", True)
epochs = options.get("epochs", 50)
milestones = options.get("milestones", [25, 40])
gamma = options.get("gamma", 0.1)
loss_constraint_fun = options.get(
"loss_constraint_fun", default_loss_constraint_fun
)
# Augment bounds with deriv indicator
bounds = torch.cat((model.bounds_, torch.zeros(2, 1)), dim=1)
# Fix deriv indicator to 0 during optimization
fixed_features = {(bounds.shape[1] - 1): 0.0}
# Fix explore features to random values
if self.explore_features is not None:
for idx in self.explore_features:
val = (
bounds[0, idx]
+ torch.rand(1, dtype=bounds.dtype)
* (bounds[1, idx] - bounds[0, idx])
).item()
fixed_features[idx] = val
bounds[0, idx] = val
bounds[1, idx] = val
acqf = self._instantiate_acquisition_fn(model)
# Initialize
batch_initial_conditions = gen_batch_initial_conditions(
acq_function=acqf,
bounds=bounds,
q=1,
num_restarts=num_restarts,
raw_samples=raw_samples,
)
clamped_candidates = columnwise_clamp(
X=batch_initial_conditions, lower=bounds[0], upper=bounds[1]
).requires_grad_(True)
candidates = fix_features(clamped_candidates, fixed_features)
optimizer = torch.optim.SGD(
params=[clamped_candidates], lr=lr, momentum=momentum, nesterov=nesterov
)
lr_scheduler = torch.optim.lr_scheduler.MultiStepLR(
optimizer, milestones=milestones, gamma=gamma
)
# Optimize
for epoch in range(epochs):
loss = -acqf(candidates).sum()
# adjust loss based on constraints on candidates
loss = loss_constraint_fun(loss, candidates)
if verbosity_freq > 0 and epoch % verbosity_freq == 0:
logger.info("Iter: {} - Value: {:.3f}".format(epoch, -(loss.item())))
def closure():
optimizer.zero_grad()
loss.backward(
retain_graph=True
) # Variational model requires retain_graph
return loss
optimizer.step(closure)
clamped_candidates.data = columnwise_clamp(
X=clamped_candidates, lower=bounds[0], upper=bounds[1]
)
candidates = fix_features(clamped_candidates, fixed_features)
lr_scheduler.step()
# Extract best point
with torch.no_grad():
batch_acquisition = acqf(candidates)
best = torch.argmax(batch_acquisition.view(-1), dim=0)
Xopt = candidates[best][:, :-1].detach()
return Xopt
[docs] @classmethod
def from_config(cls, config: Config):
classname = cls.__name__
acqf = config.getobj("common", "acqf", fallback=None)
extra_acqf_args = cls._get_acqf_options(acqf, config)
options = {}
options["num_restarts"] = config.getint(classname, "restarts", fallback=10)
options["raw_samples"] = config.getint(classname, "samps", fallback=1000)
options["verbosity_freq"] = config.getint(
classname, "verbosity_freq", fallback=-1
)
options["lr"] = config.getfloat(classname, "lr", fallback=0.01) # type: ignore
options["momentum"] = config.getfloat(classname, "momentum", fallback=0.9) # type: ignore
options["nesterov"] = config.getboolean(classname, "nesterov", fallback=True)
options["epochs"] = config.getint(classname, "epochs", fallback=50)
options["milestones"] = config.getlist(
classname, "milestones", fallback=[25, 40] # type: ignore
)
options["gamma"] = config.getfloat(classname, "gamma", fallback=0.1) # type: ignore
options["loss_constraint_fun"] = config.getobj(
classname, "loss_constraint_fun", fallback=default_loss_constraint_fun
)
explore_features = config.getlist(classname, "explore_idxs", fallback=None) # type: ignore
return cls(
acqf=acqf,
acqf_kwargs=extra_acqf_args,
model_gen_options=options,
explore_features=explore_features,
)
