Trainable Unit

This context lets you access and manage a single trainable unit, which is a child of the Machine Learning context.

Unique Actions

[?]

Configure

This Configure action is used to edit the properties of the trainable unit.

Changing Name field during this operation will cause renaming of current context. This may lead to malfunctioning of other system components that use context name/path as a primary identifier.

Action Type:

Configure

Reset

This action is used to reset the state of the trainable unit. If the trainable unit is trained it will become untrained as a result of this action. The evaluation statistics (if any have been accumulated) will also be cleared.

Reset Evaluation

This action clears the accumulated evaluation statistics of the trainable unit.

Common Actions

[?]

Delete, Make Copy, Replicate, Edit Context Permissions, Monitor Related Events, View Status

Context States and Icons

Icon	Code	State
	0	This trainable unit is untrained.
	1	This trainable unit is trained.

Advanced Information

Context Information

Context Type: trainableUnit

Context Name: provided by user

Context Description: provided by user

Context Path: users.USER_NAME.machineLearning.TRAINABLE_UNIT_NAME

Context Mask: users.*.machineLearning.*

Context Permissions

[?]

Level	Description
None	No access allowed.
Observer	Trainable unit configuration browsing. Basic event monitoring. Status browsing.
Operator	Same as Observer.
Manager	Trainable unit removal.
Engineer	Same as Manager.
Administrator	Trainable unit configuration.

Public Variables (Properties)

[?]

This context has no public variables.

Common Variables: groupMembership (Group Membership), validity (Validity), activeAlerts (Active Alerts)

Properties

See description of the variable and its fields here.

Variable Name:	childInfo
Records:	1
Permissions:	Readable at Observer permission level, writable at Manager permission level

Record Format:

Field Name	Field Type	Notes
name	String	1 - 50 characters
description	String	1 - 50 characters
task	String
algorithm	String
hyperparameters	Data Table	Contains algorithm hyperparameters.
labelFieldName	String	1 - 50 characters
hasWeights	Boolean
weightFieldName	String	1 - 50 characters or NULL
filters	Data Table	Contains a list of internal filters for data preprocessing. Each filter is described by the three fields: Description. A textual description of the filter. It may indicate type, purpose or other characteristics. Type. The filter type, chosen from the list of available filters. Parameters. The parameters of the filter.
cvNumFolds	Integer
cvRandomSeed	Integer

Public Functions

[?]

Train

Trains the trainable unit on a given training set. Can update trainable unit for updatable algorithms (Filtered Predictor, Stochastic Gradient Descent and Multiclass Updateable Classifer) on a new training set.

Function Name:	train
Permissions:	Accessible at Observer permission level
Input Records:	1...unlimited
Input Format:	Dynamic, depends on the structure of the training set. See the Arguments of Trainable Unit Functions section for detail.
Output Records:	1
Output Format:	Name	Type	Description
Output Format:	trainedUnitInfo	String	Information about the trained unit

Operate

Predicts the value of the target variable for data instances of a given dataset.

	Name	Type	Description
Function Name:	operate
Permissions:	Accessible at Observer permission level
Input Records:	0...unlimited
Input Format:	Dynamic, depends on the structure of the dataset. Must match the format of the dataset (the training set) that was passed to the Train function. See the Arguments of Trainable Unit Functions section for detail.
Output Records:	0...unlimited
Output Format:	Comprised of the input format fields plus the following fields: For regression problems:
	predicted	Double	Predicted value of the target variable (label). Note that the predicted value is always of the Double type, while the target variable can be of a different numeric type.
	error	Double	Difference between the predicted value and the actual value of the target variable. If the actual value is not given (equals NULL) then the error value is also NULL.
	For classification problems:
	Name	Type	Description
	predicted	String	Predicted class. If the target variable is not of the String type then the String representation of the class value is given.
	probability	Double	Probability of the given data instance to belong to the predicted class.
	For anomaly detection problems:
	Name	Type	Description
	predicted	String	Prediction whether the given data instance is "Normal" or "Abnormal".

Evaluate

Evaluates the performance of the trained unit on a given dataset and returns a set of evaluation metrics specific to the task.

	Name	Type	Description
Function Name:	evaluate
Permissions:	Accessible at Observer permission level
Input Records:	1...unlimited
Input Format:	Dynamic, depends on the structure of the dataset. Must match the format of the dataset (the training set) that was passed to the Train function. See the Arguments of Trainable Unit Functions section for detail.
Output Records:	1
Output Format:	For regression problems:
	correlation	Double	Correlation coefficient.
	meanAbsoluteError	Double	Mean absolute error.
	rootMeanSquaredError	Double	Root mean squared error.
	relativeAbsoluteError	Double	Relative absolute error.
	rootRelativeSquaredError	Double	Root relative squared error.
	unclassified	Long	Number of unclassified instances.
	pctUnclassified	Double	Unclassified instances in percentage of the total number of instances.
	totalNumInstances	Long	Total number of instances.
	For classification and anomaly detection problems:
	Name	Type	Description
	correct	Long	Number of correctly classified instances.
	pctCorrect	Double	Correctly classified instances in percentage of the total number of instances.
	incorrect	Long	Number of incorrectly classified instances.
	pctIncorrect	Double	Incorrectly classified instances in percentage of the total number of instances.
	unclassified	Long	Number of unclassified instances.
	pctUnclassified	Double	Unclassified instances in percentage of the total number of instances.
	totalNumInstances	Long	Total number of instances.
	kappa	Double	Kappa statistic.
	meanAbsoluteError	Double	Mean absolute error.
	rootMeanSquaredError	Double	Root mean squared error.
	relativeAbsoluteError	Double	Relative absolute error.
	rootRelativeSquaredError	Double	Root relative squared error.
	detailedAccuracy	Data Table	Detailed accuracy metrics by class. Includes the following fields: Class True Positive Rate False Positive Rate Precision Recall F-Measure Matthews Correlation Coefficient Area under ROC (Receiver Operating Characteristic) Curve Area under Precision-Recall Curve
	confusionMatrix	Data Table	Confusion matrix.

Cross Validate

Evaluates the performance of the trainable unit on a given dataset via cross-validation and returns a set of evaluation metrics specific to the task. In contrast to the evaluate function, the trainable unit can be either trained or untrained. The function does not change the state of the trainable unit.

Function Name:	crossValidate
Permissions:	Accessible at Observer permission level
Input Records:	1...unlimited
Input Format:	Dynamic, depends on the structure of the dataset. See the Arguments of trainable unit functions section for detail.
Output Records:	1
Output Format:	The same as for the evaluate function.

Reset

This function is used to reset the state of the trainable unit. If the trainable unit is trained it will become untrained as a result of the execution of this action. This function is called by the Reset action.

Function Name:	reset
Permissions:	Accessible at Observer permission level
Input Records:	0
Input Format:	none
Output Records:	0
Output Format:	none

Reset Evaluation

This function clears the accumulated evaluation statistics of the trainable unit. This function is called by the Reset Evaluation action.

Function Name:	resetEvaluation
Permissions:	Accessible at Observer permission level
Input Records:	0
Input Format:	none
Output Records:	0
Output Format:	none

Public Events

[?]

Common Events: info (Information)

DPF_EXPORT_V5.8.8 / TIME_1743710561 / 2025-04-03 20:02:41 / STRLEN_26880 / STRLEN_HTML_26382

{VERSION}

Was this page helpful?