Active Machine Learning Action Set: Syntax
Provides an action set for performing active learning which interactively query the user in order to minimize the labeling effort
iterate Action
Performs active learning iteratively and assesses the model performance..
CASL Syntax
Summary: Input and Output Tables
If a row includes a subparameter, you can specify the name, caslib, and so on in the subparameter. Otherwise, you can specify the name, caslib, and so on in the parameter.
|
Parameter |
Subparameter |
Description |
|---|---|---|
|
required parameterannotatedTable |
— |
specifies the table that contains the labels to start the model training. |
|
required parametergroundTruthTable |
— |
specifies the table that contains all the ground truth labels. |
|
required parametertable |
— |
specifies the table that contains all unlabeled data. |
|
required parametertestTable |
— |
specifies the table to use for model evaluation. |
|
Parameter |
Subparameter |
Description |
|---|---|---|
|
required parameteroutIterationHistory |
— |
creates a data set that contains the model assessment across iterations. |
|
required parameteroutQueryHistory |
— |
creates a data set that contains the query history across iterations. |
Parameter Descriptions
* annotatedTable={castable}
specifies the table that contains the labels to start the model training.
For more information about specifying the annotatedTable parameter, see the common castable parameter (Appendix A: Common Parameters).
* event="string"
specifies the prediction event level for model assessment. If the target has a format, then you must specify the formatted value.
* groundTruthTable={castable}
specifies the table that contains all the ground truth labels.
For more information about specifying the groundTruthTable parameter, see the common castable parameter (Appendix A: Common Parameters).
* id="variable-name"
specifies the ID variable to use for merging the labels from the table specified in the annotatedTable parameter into the input data set. The value of this parameter must be unique and nonmissing.
* inputs={{casinvardesc-1} <, {casinvardesc-2}, ...>}
specifies variables to use for analysis.
For more information about specifying the inputs parameter, see the common casinvardesc parameter (Appendix A: Common Parameters).
| Alias | input |
|---|
* modelProgram="string"
specifies user-defined modeling.
nBins=integer
specifies the number of bins to use for model assessment.
| Default | 20 |
|---|---|
| Range | 2–MACINT |
nIterations=integer
specifies the number of iterations.
| Default | 20 |
|---|---|
| Range | 0–MACINT |
nominals={{casinvardesc-1} <, {casinvardesc-2}, ...>}
specifies nominal variables to use for analysis.
For more information about specifying the nominals parameter, see the common casinvardesc parameter (Appendix A: Common Parameters).
| Alias | nominal |
|---|
* outIterationHistory={casouttable}
creates a data set that contains the model assessment across iterations.
For more information about specifying the outIterationHistory parameter, see the common casouttable parameter (Appendix A: Common Parameters).
* outQueryHistory={casouttable}
creates a data set that contains the query history across iterations.
For more information about specifying the outQueryHistory parameter, see the common casouttable parameter (Appendix A: Common Parameters).
* selectQuery={method="RANDOM" | "RELEVANCE" | "UNCERTAINTY", method-specific-parameters}
specifies the query strategy to use. The strategy can be either random sampling, relevance sampling, or uncertainty sampling. By default, the strategy is uncertainty sampling.
The value that you specify for the method parameter determines the other parameters that apply.
* table={castable}
specifies the table that contains all unlabeled data.
For more information about specifying the table parameter, see the common castable parameter (Appendix A: Common Parameters).
* target="variable-name"
specifies the binary target variable to use for analysis. Only the binary target is currently supported.
* testTable={castable}
specifies the table to use for model evaluation.
For more information about specifying the testTable parameter, see the common castable parameter (Appendix A: Common Parameters).
topK=64-bit-integer
specifies the top K observations to query. When you specify a value of K greater than the number of observations available for query, the number of observations available is used instead.
| Default | 50 |
|---|---|
| Minimum value | 1 |
Parameters for method="RANDOM"
minNObs=64-bit-integer
specifies the maximum number of labeled observations.
| Alias | cutoff |
|---|---|
| Default | 1000 |
| Minimum value | 1 |
seed=double
specifies the seed for the random number generator that is used for random sampling.
| Default | 0 |
|---|
Parameters for method="RELEVANCE"
gamma=double
specifies the number of the inverse of the variance for Gaussian kernels in order to calculate the relevance between labeled and unlabeled observations.
| Default | 20 |
|---|---|
| Minimum value (exclusive) | 0 |
minNObs=64-bit-integer
specifies the maximum number of labeled observations.
| Alias | cutoff |
|---|---|
| Default | 1000 |
| Minimum value | 1 |
Parameters for method="UNCERTAINTY"
includeAllData=TRUE | FALSE
when set to True, indicates that observations with and without labels can be used to query. Otherwise, only unlabeled observations can be used to query.
| Default | FALSE |
|---|
metric="ENTROPY" | "LEASTCONFIDENCE" | "RATIOOFCONFIDENCE"
specifies the metric to use for calculating uncertainty.
| Default | ENTROPY |
|---|
* probVar="variable-name"
specifies the probability variable to use for calculating uncertainty.
| Alias | predEventProb |
|---|
iterate Action
Performs active learning iteratively and assesses the model performance..
Lua Syntax
Summary: Input and Output Tables
If a row includes a subparameter, you can specify the name, caslib, and so on in the subparameter. Otherwise, you can specify the name, caslib, and so on in the parameter.
|
Parameter |
Subparameter |
Description |
|---|---|---|
|
required parameterannotatedTable |
— |
specifies the table that contains the labels to start the model training. |
|
required parametergroundTruthTable |
— |
specifies the table that contains all the ground truth labels. |
|
required parametertable |
— |
specifies the table that contains all unlabeled data. |
|
required parametertestTable |
— |
specifies the table to use for model evaluation. |
|
Parameter |
Subparameter |
Description |
|---|---|---|
|
required parameteroutIterationHistory |
— |
creates a data set that contains the model assessment across iterations. |
|
required parameteroutQueryHistory |
— |
creates a data set that contains the query history across iterations. |
Parameter Descriptions
* annotatedTable={castable}
specifies the table that contains the labels to start the model training.
For more information about specifying the annotatedTable parameter, see the common castable parameter (Appendix A: Common Parameters).
* event="string"
specifies the prediction event level for model assessment. If the target has a format, then you must specify the formatted value.
* groundTruthTable={castable}
specifies the table that contains all the ground truth labels.
For more information about specifying the groundTruthTable parameter, see the common castable parameter (Appendix A: Common Parameters).
* id="variable-name"
specifies the ID variable to use for merging the labels from the table specified in the annotatedTable parameter into the input data set. The value of this parameter must be unique and nonmissing.
* inputs={{casinvardesc-1} <, {casinvardesc-2}, ...>}
specifies variables to use for analysis.
For more information about specifying the inputs parameter, see the common casinvardesc parameter (Appendix A: Common Parameters).
| Alias | input |
|---|
* modelProgram="string"
specifies user-defined modeling.
nBins=integer
specifies the number of bins to use for model assessment.
| Default | 20 |
|---|---|
| Range | 2–MACINT |
nIterations=integer
specifies the number of iterations.
| Default | 20 |
|---|---|
| Range | 0–MACINT |
nominals={{casinvardesc-1} <, {casinvardesc-2}, ...>}
specifies nominal variables to use for analysis.
For more information about specifying the nominals parameter, see the common casinvardesc parameter (Appendix A: Common Parameters).
| Alias | nominal |
|---|
* outIterationHistory={casouttable}
creates a data set that contains the model assessment across iterations.
For more information about specifying the outIterationHistory parameter, see the common casouttable parameter (Appendix A: Common Parameters).
* outQueryHistory={casouttable}
creates a data set that contains the query history across iterations.
For more information about specifying the outQueryHistory parameter, see the common casouttable parameter (Appendix A: Common Parameters).
* selectQuery={method="RANDOM" | "RELEVANCE" | "UNCERTAINTY", method-specific-parameters}
specifies the query strategy to use. The strategy can be either random sampling, relevance sampling, or uncertainty sampling. By default, the strategy is uncertainty sampling.
The value that you specify for the method parameter determines the other parameters that apply.
* table={castable}
specifies the table that contains all unlabeled data.
For more information about specifying the table parameter, see the common castable parameter (Appendix A: Common Parameters).
* target="variable-name"
specifies the binary target variable to use for analysis. Only the binary target is currently supported.
* testTable={castable}
specifies the table to use for model evaluation.
For more information about specifying the testTable parameter, see the common castable parameter (Appendix A: Common Parameters).
topK=64-bit-integer
specifies the top K observations to query. When you specify a value of K greater than the number of observations available for query, the number of observations available is used instead.
| Default | 50 |
|---|---|
| Minimum value | 1 |
Parameters for method="RANDOM"
minNObs=64-bit-integer
specifies the maximum number of labeled observations.
| Alias | cutoff |
|---|---|
| Default | 1000 |
| Minimum value | 1 |
seed=double
specifies the seed for the random number generator that is used for random sampling.
| Default | 0 |
|---|
Parameters for method="RELEVANCE"
gamma=double
specifies the number of the inverse of the variance for Gaussian kernels in order to calculate the relevance between labeled and unlabeled observations.
| Default | 20 |
|---|---|
| Minimum value (exclusive) | 0 |
minNObs=64-bit-integer
specifies the maximum number of labeled observations.
| Alias | cutoff |
|---|---|
| Default | 1000 |
| Minimum value | 1 |
Parameters for method="UNCERTAINTY"
includeAllData=true | false
when set to True, indicates that observations with and without labels can be used to query. Otherwise, only unlabeled observations can be used to query.
| Default | false |
|---|
metric="ENTROPY" | "LEASTCONFIDENCE" | "RATIOOFCONFIDENCE"
specifies the metric to use for calculating uncertainty.
| Default | ENTROPY |
|---|
* probVar="variable-name"
specifies the probability variable to use for calculating uncertainty.
| Alias | predEventProb |
|---|
iterate Action
Performs active learning iteratively and assesses the model performance..
Python Syntax
Summary: Input and Output Tables
If a row includes a subparameter, you can specify the name, caslib, and so on in the subparameter. Otherwise, you can specify the name, caslib, and so on in the parameter.
|
Parameter |
Subparameter |
Description |
|---|---|---|
|
required parameterannotatedTable |
— |
specifies the table that contains the labels to start the model training. |
|
required parametergroundTruthTable |
— |
specifies the table that contains all the ground truth labels. |
|
required parametertable |
— |
specifies the table that contains all unlabeled data. |
|
required parametertestTable |
— |
specifies the table to use for model evaluation. |
|
Parameter |
Subparameter |
Description |
|---|---|---|
|
required parameteroutIterationHistory |
— |
creates a data set that contains the model assessment across iterations. |
|
required parameteroutQueryHistory |
— |
creates a data set that contains the query history across iterations. |
Parameter Descriptions
* annotatedTable={castable}
specifies the table that contains the labels to start the model training.
For more information about specifying the annotatedTable parameter, see the common castable parameter (Appendix A: Common Parameters).
* event="string"
specifies the prediction event level for model assessment. If the target has a format, then you must specify the formatted value.
* groundTruthTable={castable}
specifies the table that contains all the ground truth labels.
For more information about specifying the groundTruthTable parameter, see the common castable parameter (Appendix A: Common Parameters).
* id="variable-name"
specifies the ID variable to use for merging the labels from the table specified in the annotatedTable parameter into the input data set. The value of this parameter must be unique and nonmissing.
* inputs=[{casinvardesc-1} <, {casinvardesc-2}, ...>]
specifies variables to use for analysis.
For more information about specifying the inputs parameter, see the common casinvardesc parameter (Appendix A: Common Parameters).
| Alias | input |
|---|
* modelProgram="string"
specifies user-defined modeling.
nBins=integer
specifies the number of bins to use for model assessment.
| Default | 20 |
|---|---|
| Range | 2–MACINT |
nIterations=integer
specifies the number of iterations.
| Default | 20 |
|---|---|
| Range | 0–MACINT |
nominals=[{casinvardesc-1} <, {casinvardesc-2}, ...>]
specifies nominal variables to use for analysis.
For more information about specifying the nominals parameter, see the common casinvardesc parameter (Appendix A: Common Parameters).
| Alias | nominal |
|---|
* outIterationHistory={casouttable}
creates a data set that contains the model assessment across iterations.
For more information about specifying the outIterationHistory parameter, see the common casouttable parameter (Appendix A: Common Parameters).
* outQueryHistory={casouttable}
creates a data set that contains the query history across iterations.
For more information about specifying the outQueryHistory parameter, see the common casouttable parameter (Appendix A: Common Parameters).
* selectQuery={"method":"RANDOM" | "RELEVANCE" | "UNCERTAINTY", method-specific-parameters}
specifies the query strategy to use. The strategy can be either random sampling, relevance sampling, or uncertainty sampling. By default, the strategy is uncertainty sampling.
The value that you specify for the method parameter determines the other parameters that apply.
* table={castable}
specifies the table that contains all unlabeled data.
For more information about specifying the table parameter, see the common castable parameter (Appendix A: Common Parameters).
* target="variable-name"
specifies the binary target variable to use for analysis. Only the binary target is currently supported.
* testTable={castable}
specifies the table to use for model evaluation.
For more information about specifying the testTable parameter, see the common castable parameter (Appendix A: Common Parameters).
topK=64-bit-integer
specifies the top K observations to query. When you specify a value of K greater than the number of observations available for query, the number of observations available is used instead.
| Default | 50 |
|---|---|
| Minimum value | 1 |
Parameters for method="RANDOM"
"minNObs":64-bit-integer
specifies the maximum number of labeled observations.
| Alias | cutoff |
|---|---|
| Default | 1000 |
| Minimum value | 1 |
"seed":double
specifies the seed for the random number generator that is used for random sampling.
| Default | 0 |
|---|
Parameters for method="RELEVANCE"
"gamma":double
specifies the number of the inverse of the variance for Gaussian kernels in order to calculate the relevance between labeled and unlabeled observations.
| Default | 20 |
|---|---|
| Minimum value (exclusive) | 0 |
"minNObs":64-bit-integer
specifies the maximum number of labeled observations.
| Alias | cutoff |
|---|---|
| Default | 1000 |
| Minimum value | 1 |
Parameters for method="UNCERTAINTY"
"includeAllData":True | False
when set to True, indicates that observations with and without labels can be used to query. Otherwise, only unlabeled observations can be used to query.
| Default | False |
|---|
"metric":"ENTROPY" | "LEASTCONFIDENCE" | "RATIOOFCONFIDENCE"
specifies the metric to use for calculating uncertainty.
| Default | ENTROPY |
|---|
* "probVar":"variable-name"
specifies the probability variable to use for calculating uncertainty.
| Alias | predEventProb |
|---|
iterate Action
Performs active learning iteratively and assesses the model performance..
R Syntax
Summary: Input and Output Tables
If a row includes a subparameter, you can specify the name, caslib, and so on in the subparameter. Otherwise, you can specify the name, caslib, and so on in the parameter.
|
Parameter |
Subparameter |
Description |
|---|---|---|
|
required parameterannotatedTable |
— |
specifies the table that contains the labels to start the model training. |
|
required parametergroundTruthTable |
— |
specifies the table that contains all the ground truth labels. |
|
required parametertable |
— |
specifies the table that contains all unlabeled data. |
|
required parametertestTable |
— |
specifies the table to use for model evaluation. |
|
Parameter |
Subparameter |
Description |
|---|---|---|
|
required parameteroutIterationHistory |
— |
creates a data set that contains the model assessment across iterations. |
|
required parameteroutQueryHistory |
— |
creates a data set that contains the query history across iterations. |
Parameter Descriptions
* annotatedTable=list(castable)
specifies the table that contains the labels to start the model training.
For more information about specifying the annotatedTable parameter, see the common castable parameter (Appendix A: Common Parameters).
* event="string"
specifies the prediction event level for model assessment. If the target has a format, then you must specify the formatted value.
* groundTruthTable=list(castable)
specifies the table that contains all the ground truth labels.
For more information about specifying the groundTruthTable parameter, see the common castable parameter (Appendix A: Common Parameters).
* id="variable-name"
specifies the ID variable to use for merging the labels from the table specified in the annotatedTable parameter into the input data set. The value of this parameter must be unique and nonmissing.
* inputs=list( list(casinvardesc-1) <, list(casinvardesc-2), ...>)
specifies variables to use for analysis.
For more information about specifying the inputs parameter, see the common casinvardesc parameter (Appendix A: Common Parameters).
| Alias | input |
|---|
* modelProgram="string"
specifies user-defined modeling.
nBins=integer
specifies the number of bins to use for model assessment.
| Default | 20 |
|---|---|
| Range | 2–MACINT |
nIterations=integer
specifies the number of iterations.
| Default | 20 |
|---|---|
| Range | 0–MACINT |
nominals=list( list(casinvardesc-1) <, list(casinvardesc-2), ...>)
specifies nominal variables to use for analysis.
For more information about specifying the nominals parameter, see the common casinvardesc parameter (Appendix A: Common Parameters).
| Alias | nominal |
|---|
* outIterationHistory=list(casouttable)
creates a data set that contains the model assessment across iterations.
For more information about specifying the outIterationHistory parameter, see the common casouttable parameter (Appendix A: Common Parameters).
* outQueryHistory=list(casouttable)
creates a data set that contains the query history across iterations.
For more information about specifying the outQueryHistory parameter, see the common casouttable parameter (Appendix A: Common Parameters).
* selectQuery=list(method="RANDOM" | "RELEVANCE" | "UNCERTAINTY", method-specific-parameters)
specifies the query strategy to use. The strategy can be either random sampling, relevance sampling, or uncertainty sampling. By default, the strategy is uncertainty sampling.
The value that you specify for the method parameter determines the other parameters that apply.
* table=list(castable)
specifies the table that contains all unlabeled data.
For more information about specifying the table parameter, see the common castable parameter (Appendix A: Common Parameters).
* target="variable-name"
specifies the binary target variable to use for analysis. Only the binary target is currently supported.
* testTable=list(castable)
specifies the table to use for model evaluation.
For more information about specifying the testTable parameter, see the common castable parameter (Appendix A: Common Parameters).
topK=64-bit-integer
specifies the top K observations to query. When you specify a value of K greater than the number of observations available for query, the number of observations available is used instead.
| Default | 50 |
|---|---|
| Minimum value | 1 |
Parameters for method="RANDOM"
minNObs=64-bit-integer
specifies the maximum number of labeled observations.
| Alias | cutoff |
|---|---|
| Default | 1000 |
| Minimum value | 1 |
seed=double
specifies the seed for the random number generator that is used for random sampling.
| Default | 0 |
|---|
Parameters for method="RELEVANCE"
gamma=double
specifies the number of the inverse of the variance for Gaussian kernels in order to calculate the relevance between labeled and unlabeled observations.
| Default | 20 |
|---|---|
| Minimum value (exclusive) | 0 |
minNObs=64-bit-integer
specifies the maximum number of labeled observations.
| Alias | cutoff |
|---|---|
| Default | 1000 |
| Minimum value | 1 |
Parameters for method="UNCERTAINTY"
includeAllData=TRUE | FALSE
when set to True, indicates that observations with and without labels can be used to query. Otherwise, only unlabeled observations can be used to query.
| Default | FALSE |
|---|
metric="ENTROPY" | "LEASTCONFIDENCE" | "RATIOOFCONFIDENCE"
specifies the metric to use for calculating uncertainty.
| Default | ENTROPY |
|---|
* probVar="variable-name"
specifies the probability variable to use for calculating uncertainty.
| Alias | predEventProb |
|---|