Refactor Strategies to Recommenders

CHANGELOG.md

@@ -23,13 +23,24 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Changed
 - `Recommender`s now share their core logic via their base class
 - Remove progress bars in examples
+- Strategies are now called `MetaRecommender`'s and part of the `recommenders.meta`
+  module
+- `Recommender`'s are now called `PureRecommender`'s and part of the `recommenders.pure`
+  module
+- `strategy` keyword of `Campaign` renamed to `recommender`
+- `NaiveHybridRecommender` renamed to `NaiveHybridSpaceRecommender`
 
 ### Fixed
 - Unhandled exception in telemetry when username could not be inferred on Windows
 - Metadata is now correctly updated for hybrid spaces
 - Unintended deactivation of telemetry due to import problem
 - Line wrapping in examples
 
+### Deprecations
+- `TwoPhaseStrategy`
+- `SequentialStrategy`
+- `StreamingSequentialStrategy`
+
 ## [0.7.3] - 2024-02-09
 ### Added
 - Copy button for code blocks in documentation

README.md

@@ -151,10 +151,13 @@ details, and their configuration settings, see the
 of the user guide.
 
 ```python
-from baybe.strategies import TwoPhaseStrategy
-from baybe.recommenders import SequentialGreedyRecommender, FPSRecommender
+from baybe.recommenders import (
+    SequentialGreedyRecommender,
+    FPSRecommender,
+    TwoPhaseMetaRecommender,
+)
 
-strategy = TwoPhaseStrategy(
+strategy = TwoPhaseMetaRecommender(
     initial_recommender=FPSRecommender(),  # farthest point sampling
     recommender=SequentialGreedyRecommender(),  # Bayesian model-based optimization
 )

baybe/acquisition.py

@@ -111,7 +111,7 @@ class PartialAcquisitionFunction:
     pinned_part: Tensor
     """The values that will be attached whenever evaluating the acquisition function."""
 
-    pin_discrete: Tensor
+    pin_discrete: bool
     """A flag for denoting whether ``pinned_part`` corresponds to the discrete
     subspace."""
 

baybe/campaign.py

@@ -14,12 +14,12 @@
 from baybe.objective import Objective
 from baybe.parameters.base import Parameter
 from baybe.recommenders.base import RecommenderProtocol
+from baybe.recommenders.meta.sequential import TwoPhaseMetaRecommender
 from baybe.searchspace.core import (
     SearchSpace,
     validate_searchspace_from_config,
 )
 from baybe.serialization import SerialMixin, converter
-from baybe.strategies import TwoPhaseStrategy
 from baybe.targets.base import Target
 from baybe.telemetry import (
     TELEM_LABELS,
@@ -45,7 +45,7 @@ class Campaign(SerialMixin):
     In particular, a campaign:
         * Defines the objective of an experimentation process.
         * Defines the search space over which the experimental parameter may vary.
-        * Defines a strategy for traversing the search space.
+        * Defines a recommender for exploring the search space.
         * Records the measurement data collected during the process.
         * Records metadata about the progress of the experimentation process.
     """
@@ -57,8 +57,8 @@ class Campaign(SerialMixin):
     objective: Objective = field()
     """The optimization objective."""
 
-    strategy: RecommenderProtocol = field(factory=TwoPhaseStrategy)
-    """The employed strategy"""
+    recommender: RecommenderProtocol = field(factory=TwoPhaseMetaRecommender)
+    """The employed recommender"""
 
     # Metadata
     n_batches_done: int = field(default=0, init=False)
@@ -82,6 +82,9 @@ class Campaign(SerialMixin):
     numerical_measurements_must_be_within_tolerance: bool = field(default=None)
     """Deprecated! Raises an error when used."""
 
+    strategy: RecommenderProtocol = field(default=None)
+    """Deprecated! Raises an error when used."""
+
     @numerical_measurements_must_be_within_tolerance.validator
     def _validate_tolerance_flag(self, _, value) -> None:
         """Raise a DeprecationError if the tolerance flag is used."""
@@ -92,6 +95,15 @@ def _validate_tolerance_flag(self, _, value) -> None:
                 f"{self.__class__.__name__}.{Campaign.add_measurements.__name__}."
             )
 
+    @strategy.validator
+    def _validate_strategy(self, _, value) -> None:
+        """Raise a DeprecationError if the strategy attribute is used."""
+        if value is not None:
+            raise DeprecationError(
+                "Passing 'strategy' to the constructor is deprecated. The attribute "
+                "has been renamed to 'recommender'."
+            )
+
     @property
     def measurements(self) -> pd.DataFrame:
         """The experimental data added to the Campaign."""
@@ -275,7 +287,7 @@ def recommend(
             self._measurements_exp["FitNr"].fillna(self.n_fits_done, inplace=True)
 
         # Get the recommended search space entries
-        rec = self.strategy.recommend(
+        rec = self.recommender.recommend(
             self.searchspace,
             batch_size,
             self._measurements_parameters_comp,
@@ -306,6 +318,7 @@ def _add_version(dict_: dict) -> dict:
     _cattrs_include_init_false=True,
     # TODO: Remove once deprecation got expired:
     numerical_measurements_must_be_within_tolerance=cattrs.override(omit=True),
+    strategy=cattrs.override(omit=True),
 )
 structure_hook = cattrs.gen.make_dict_structure_fn(
     Campaign, converter, _cattrs_include_init_false=True

baybe/exceptions.py

@@ -37,7 +37,9 @@ class ModelParamsNotSupportedError(Exception):
 
 
 class NoRecommendersLeftError(Exception):
-    """A recommender is requested by a strategy but there are no recommenders left."""
+    """A recommender is requested by a meta recommender but there are no recommenders
+    left.
+    """
 
 
 class NumericalUnderflowError(Exception):

baybe/parameters/base.py

@@ -61,7 +61,7 @@ class DiscreteParameter(Parameter, ABC):
 
     # class variables
     encoding: Optional[ParameterEncoding] = field(init=False, default=None)
-    """An optional encoding strategy for the parameter."""
+    """An optional encoding for the parameter."""
 
     @property
     @abstractmethod

baybe/recommenders/__init__.py

@@ -1,20 +1,33 @@
 """BayBE recommenders."""
 
-from baybe.recommenders.bayesian.sequential_greedy import SequentialGreedyRecommender
-from baybe.recommenders.naive import NaiveHybridRecommender
-from baybe.recommenders.nonpredictive.clustering import (
+from baybe.recommenders.meta.sequential import (
+    SequentialMetaRecommender,
+    StreamingSequentialMetaRecommender,
+    TwoPhaseMetaRecommender,
+)
+from baybe.recommenders.naive import NaiveHybridSpaceRecommender
+from baybe.recommenders.pure.bayesian.sequential_greedy import (
+    SequentialGreedyRecommender,
+)
+from baybe.recommenders.pure.nonpredictive.clustering import (
     GaussianMixtureClusteringRecommender,
     KMeansClusteringRecommender,
     PAMClusteringRecommender,
 )
-from baybe.recommenders.nonpredictive.sampling import FPSRecommender, RandomRecommender
+from baybe.recommenders.pure.nonpredictive.sampling import (
+    FPSRecommender,
+    RandomRecommender,
+)
 
 __all__ = [
     "FPSRecommender",
     "GaussianMixtureClusteringRecommender",
     "KMeansClusteringRecommender",
     "PAMClusteringRecommender",
-    "NaiveHybridRecommender",
+    "NaiveHybridSpaceRecommender",
     "RandomRecommender",
+    "TwoPhaseMetaRecommender",
     "SequentialGreedyRecommender",
+    "SequentialMetaRecommender",
+    "StreamingSequentialMetaRecommender",
 ]