constants Module

Defines automated ML constants used in Azure Machine Learning.

Classes

API	Defines names for the Azure Machine Learning API operations that can be performed.
AcquisitionFunction	Defines names for all acquisition functions used to select the next pipeline. The default is EI (expected improvement).
AggregationFunctions	Define the aggregation functions for numeric columns.
AutoMLDefaultTimeouts	Constants to store the default timeouts
AutoMLJson	Defines constants for JSON created by automated ML.
AutoMLValidation
CheckImbalance	If the ratio of the samples in the minority class to the samples in the majority class is equal to or lower than this threshold, then Imbalance will be detected in the dataset.
ClientErrors	Defines client errors that can occur when violating user-specified cost constraints.
DatetimeDtype	Defines supported datetime datatypes. Names correspond to the output of pandas.api.types.infer_dtype().
Defaults	Defines default values for pipelines.
Dependencies
EnsembleConstants	Defines constants used for Ensemble iterations.
EnsembleMethod	Defines ensemble methods.
ExceptionFragments	Exception Fragments
FeatureSweeping	Defines constants for Feature Sweeping.
FitPipelineComponentName	Constants for the FitPipeline Component names.
HyperparameterSweepingConstants	Defines constants related with hyperparameter tunning.
IterationTimeout	Defines ways of changing the per_iteration_timeout.
LearnerColumns	Defines all columns used for learner pipeline.
LegacyModelNames	Defines names for all models supported by the Miro recommender in Automated ML. These names are still used to refer to objects in the Miro database, but are not used by any Automated ML clients.
MLFlowLiterals	Constants related to MLFlow.
MLFlowMetaLiterals	Constants related to MLFlow metdata.
MLTableLiterals
Metric	Defines all metrics supported by classification and regression.
MetricExtrasConstants	Defines internal values of Confidence Intervals
MetricObjective	Defines mappings from metrics to their objective. Objectives are maximization or minimization (regression and classification).
ModelCategories	Defines categories for models.
ModelClassNames	Defines class names for models. These are model wrapper class names in the pipeline specs.
ModelName	Defines a model name that includes customer, legacy, and class names. Init ModelName.
ModelNameMappings	Defines model name mappings.
ModelParameters	Defines parameter names specific to certain models. For example, to indicate which features in the dataset are categorical a LightGBM model accepts the 'categorical_feature' parameter while a CatBoost model accepts the 'cat_features' parameter.
NumericalDtype	Defines supported numerical datatypes. Names correspond to the output of pandas.api.types.infer_dtype().
Optimizer	Defines the categories of pipeline prediction algorithms used. "random" provides a baseline by selecting a pipeline randomly "lvm" uses latent variable models to predict probable next pipelines given performance on previous pipelines.
OptimizerObjectives	Defines nthe objectives an algorithm can have relative to a metric. Some metrics should be maximized and some should be minimized.
PipelineCost	Defines cost model modes. COST_NONE returns all predicted pipelines COST_FILTER returns only pipelines that were predicted by cost models to meet the user-specified cost conditions COST_SCALE divides the acquisition function score by the predicted time
PipelineMaskProfiles	Defines mask profiles for pipelines.
PipelineParameterConstraintCheckStatus	Defines values indicating whether pipeline is valid.
PreprocessorCategories	Defines categories for preprocessors.
RuleBasedValidation	Defines constants for the rule-based validation setting.
RunState	Defines states a run can be in.
ServerStatus	Defines server status values.
ShortSeriesHandlingValues	Define the possible values of ShortSeriesHandling config.
Status	Defines possible child run states.
SubsamplingSchedule	Defines subsampling strategies.
SubsamplingTreatment	Defines subsampling treatment in GP.
Subtasks	Defines names of the subtasks.
SupportedCategoricals	Defines supported categoricals learnersin _set_dataset_categoricals type :
SupportedInputDatatypes	Input data types supported by AutoML for different Run types.
SupportedModelNames	Defines supported models where each model has a customer name, legacy model name, and model class name.
SupportedModels	Defines customer-facing names for algorithms supported by automated ML in Azure Machine Learning.
Tasks	Defines types of machine learning tasks supported by automated ML.
TelemetryConstants	Defines telemetry constants.
TextOrCategoricalDtype	Defines supported categorical datatypes.
TimeConstraintEnforcement	Enumeration of time contraint enforcement modes.
TimeSeries	Defines parameters used for timeseries.
TimeSeriesInternal	Defines non user-facing TimeSeries constants.
TimeSeriesWebLinks	Define the web links for the time series documentation.
TrainingResultsType	Defines potential results from runners class.
TrainingType	Defines validation methods. Different experiment types will use different validation methods.
Transformers	Defines transformers used for data processing.
ValidationLimitRule	Defines validation rules. Init the rule based on the inputs.

Enums

ErrorLinks	Constants to store the link to correct the errors.
ImageTask	Available Image task types.
MLTableDataLabel	An enumeration.

Functions

get_metric_from_type

Get valid metrics for a given training type.

get_metric_from_type(t)

Parameters

Name	Description
t Required

get_status_from_type

Get valid training statuses for a given training type.

get_status_from_type(t)

Parameters

Name	Description
t Required

Sample_Weights_Unsupported

Algorithm names that we must force to run in single threaded mode.

Sample_Weights_Unsupported = {'ElasticNet', 'KNeighborsClassifier', 'KNeighborsRegressor', 'LassoLars'}

TIMEOUT_TAG

Names of algorithms that do not support sample weights.

TIMEOUT_TAG = 'timeout'