trieste.observer#

Definitions and utilities for observers of objective functions.

Module Contents#

SingleObserver[source]#

Type alias for an observer of the objective function (that takes query points and returns an unlabelled dataset).

MultiObserver[source]#

Type alias for an observer of the objective function (that takes query points and returns labelled datasets).

Observer[source]#

Type alias for an observer, returning either labelled datasets or a single unlabelled dataset.

OBJECTIVE: typing_extensions.Final[trieste.types.Tag] = 'OBJECTIVE'[source]#

A tag typically used by acquisition rules to denote the data sets and models corresponding to the optimization objective.

filter_finite(query_points: trieste.types.TensorType, observations: trieste.types.TensorType) → trieste.data.Dataset[source]#
Parameters:

  • query_points – A tensor of shape (N x M).

  • observations – A tensor of shape (N x 1).

Returns:

A Dataset containing all the rows in query_points and observations where the observations are finite numbers.

Raises:

ValueError or InvalidArgumentError – If query_points or observations have invalid shapes.

map_is_finite(query_points: trieste.types.TensorType, observations: trieste.types.TensorType) → trieste.data.Dataset[source]#
Parameters:

  • query_points – A tensor.

  • observations – A tensor.

Returns:

A Dataset containing all the rows in query_points, along with the tensor result of mapping the elements of observations to: 1 if they are a finite number, else 0, with dtype tf.uint8.

Raises:

ValueError or InvalidArgumentError – If query_points and observations do not satisfy the shape constraints of Dataset.