Estimator (Core)

alts.core.estimator

class Estimator

Bases: ExperimentModule, ExpModSubscriber, ResultDataSubscriber

Description

The Estimator is the training model, trying to find patterns in given data and extrapolate them to new data.

abstract estimate(self, exp_mods) → results

Description

Tries to estimate the data source given a certain experiment environment based on its current training.

Parameters:

exp_mods (ExperimentModules) – The experiment modules that configure the environment of the estimator

Returns:

The estimated data structure

Return type:

data_points

Raises:

NotImplementedError

experiment_update(self, subscription) → None

Description

Updates the subscription and tries to estimate in the new environment.

Parameters:

subscription (Subscribable) – Subscription to be updated

Returns:

The estimated data structure

Return type:

data_points

abstract query(self, queries) → results

Description

Tries to predict the results to given queries.

Parameters:

queries (queries) – Queries to predict results to

Returns:

Predicted results to queries

Return type:

results

Raises:

NotImplementedError

result_update(self, subscription) → None

Description

Updates the subscription and trains itself on the new data.

Parameters:

subscription (Subscribable) – A subscription the Estimator depends on

train(self, result_pool) → None

Description

Trains the Estimator on a set of data_points.

Parameters:

result_pool (ResultDataPool) – A set of data_points