trieste.objectives.utils#

This module contains functions convenient for creating Observer objects that return data from objective functions, appropriately formatted for usage with the toolbox.

Module Contents#

mk_observer(objective: collections.abc.Callable[[trieste.types.TensorType], trieste.types.TensorType]) → trieste.observer.SingleObserver[source]#
mk_observer(objective: collections.abc.Callable[[trieste.types.TensorType], trieste.types.TensorType], key: trieste.types.Tag) → trieste.observer.MultiObserver
Parameters:

  • objective – An objective function designed to be used with a single data set and model.

  • key – An optional key to use to access the data from the observer result.

Returns:

An observer returning the data from objective.

mk_multi_observer(**kwargs: collections.abc.Callable[[trieste.types.TensorType], trieste.types.TensorType]) → trieste.observer.MultiObserver[source]#
Parameters:

kwargs – Observation functions.

Returns:

An multi-observer returning the data from kwargs.

mk_batch_observer(objective_or_observer: collections.abc.Callable[[trieste.types.TensorType], trieste.types.TensorType] | trieste.observer.Observer, default_key: trieste.types.Tag = OBJECTIVE) → trieste.observer.MultiObserver[source]#

Create an observer that returns the data from objective or an existing observer separately for each query point in a batch.

Parameters:

  • objective_or_observer – An objective or an existing observer.

  • default_key – The default key to use if objective_or_observer is an objective or does not return a mapping.

Returns:

A multi-observer across the batch dimension of query points, returning the data from objective_or_observer.