Autotune Action Set: Syntax

Provides actions to tune machine learning algorithm hyperparameters

tuneFactMac Action

Automatically adjusts factorization machine parameters to tune a model for minimum error.

CASL Syntax

autotune.tuneFactMac <result=results> <status=rc> /
scoreOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}
trainOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}
tunerOptions={
evaluationHistory=TRUE | FALSE,
logLevel=integer,
maxEvals=integer,
maxIters=integer,
maxTime=double,
maxTrainTime=64-bit-integer,
nCrossValFolds=integer,
nParallel=integer,
popSize=integer,
sampleSize=integer,
seed=integer,
shuffleGridPoints=TRUE | FALSE,
targetEvent="string",
userDefinedPartition=TRUE | FALSE,
}
tuningParameters={{
exclude=TRUE | FALSE,
initValue=integer | 64-bit-integer | double | TRUE | FALSE | "string",
lowerBound=double,
namePath="string",
upperBound=double,
valueList={any-list-or-data-type-1 <, any-list-or-data-type-2, ...>}
}, {...}}
useParameters="COMBINED" | "CUSTOM" | "STANDARD"
;

Parameter Descriptions

scoreOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}

specifies a list of parameters to be used by the score action in the aStore action set; for more information, see the parameters for that action.

trainOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}

specifies a list of parameters to be used by the factmac action in the factmac action set; for more information, see the parameters for that action.

tunerOptions={tunerOptions}

specifies a list of parameters for adjusting how the tuner behaves.

AliasoptMinerOpts
evaluationHistory=TRUE | FALSE

when set to True, creates a table of every evaluation (every model configuration) and returns those tables to the client.

DefaultTRUE
logLevel=integer

specifies the level of log messages to be written: no logs (0), initialization and completion logs (1), setup and iteration results (2), or detailed evaluation results (3).

Default2
Range0–3
maxBayesianModelSize=integer

specifies the maximum number of points in the kriging model. This parameter is ignored unless the value "BAYESIAN" is specified in the searchMethod parameter.

AliasmaxBayesianSize
Default100
Minimum value10
maxEvals=integer

specifies the maximum number of objective evaluations (model configurations) to be trained and scored.

AliasmaxEvaluations
Default50
Minimum value1
maxIters=integer

specifies the maximum number of iterations of local search optimization for model tuning. At each iteration, the number of evaluated models is less than or equal to the value of the popSize parameter.

AliasmaxIterations
Default5
Minimum value1
maxTime=double

specifies the maximum time allowed for tuning. The actual time might exceed this value by the amount of time required to train and score a model configuration.

Default36000
Minimum value1
maxTrainTime=64-bit-integer

specifies the maximum time allowed for a single model training. The model training is terminated if it exceeds this time, and the objective value is set to missing.

Minimum value0
nCrossValFolds=integer

specifies the number of folds to use for cross validation to assess model fit error as the tuning objective.

AliasnFolds
Default5
Minimum value2
nParallel=integer

specifies the number of threads to be used by the optimizer for parallel evaluation of model configurations.

Default0
nSubsessionWorkers=integer

specifies the number of workers to be used by each subsession for parallel evaluation of model configurations.

AliasnSubWorkers
Default0
Minimum value0
objective="ASE" | "AUC" | "F05" | "F1" | "GAMMA" | "GINI" | "KS" | "MAE" | "MCE" | "MCLL" | "MISC" | "MSE" | "MSLE" | "RASE" | "RMAE" | "RMSLE" | "TAU"

specifies the objective metric to use during the tuning.

ASE

specifies the average square error as the objective function.

AUC

specifies area under the curve as the objective function (nominal targets only).

F05

specifies the F0.5 coefficient as the objective function (nominal targets only).

F1

specifies the F1 coefficient as the objective function (nominal targets only).

GAMMA

specifies the gamma coefficient as the objective function (nominal targets only).

GINI

specifies the Gini coefficient as the objective function (nominal targets only).

KS

specifies the Kolmogorov-Smirnov coefficient as the objective function (nominal targets only).

MAE

specifies the mean absolute error as the objective function (interval targets only).

MCE

specifies the misclassification rate as the objective function (nominal targets only).

MCLL

specifies the multiclass log loss as the objective function (nominal targets only).

MISC

specifies the misclassification error percentage as the objective function (nominal targets only).

MSE

specifies the mean square error as the objective function (interval targets only).

MSLE

specifies the mean square logarithmic error as the objective function (interval targets only).

RASE

specifies the root average square error as the objective function.

RMAE

specifies the root mean absolute error as the objective function (interval targets only).

RMSLE

specifies the root mean square logarithmic error as the objective function (interval targets only).

TAU

specifies the tau coefficient as the objective function (nominal targets only).

popSize=integer

specifies the maximum number of model configurations to evaluate at each iteration.

AliaspopulationSize
Default10
Minimum value2
sampleSize=integer

specifies the number of model configurations to evaluate when the value of the searchMethod parameter is "LHS" or "RANDOM".

Default50
Minimum value2
searchMethod="BAYESIAN" | "GA" | "GRID" | "LHS" | "RANDOM"

specifies the search method to use during tuning.

DefaultGA
BAYESIAN

uses the Bayesian search method.

GA

uses the genetic algorithm search method.

GRID

uses the grid search method.

LHS

uses the Latin hypercube sample search method.

RANDOM

uses the random search method.

seed=integer

specifies the seed to use for selecting training data and validation data for a single validation partition or fold number for each observation for cross validation.

Default0
shuffleGridPoints=TRUE | FALSE

when set to True, shuffles the points that are generated by the grid search method before execution.

DefaultTRUE
targetEvent="string"

specifies the name of the nominal target event to use for tuning.

trainPartitionFraction=double

specifies the size of a single data partition to use for model training during tuning.

AliastrainFraction
Range0.01–0.99
userDefinedPartition=TRUE | FALSE

when set to True, includes a user-defined partition for training and scoring.

DefaultFALSE
validationPartitionFraction=double

specifies the size of a single validation partition to use in order to assess model fit error as the tuning objective.

AliasvalidateFraction
Default0.3
Range0.01–0.99

tuningParameters={{autotuneTuningParmDefinition-1} <, {autotuneTuningParmDefinition-2}, ...>}

specifies a list of custom tuning parameters and their definitions.

The autotuneTuningParmDefinition value can be one or more of the following:

exclude=TRUE | FALSE

when set to True, excludes a tuning parameter from the tuning process.

DefaultFALSE
initValue=integer | 64-bit-integer | double | TRUE | FALSE | "string"

specifies the initial (default) value of a tuning parameter.

lowerBound=double

specifies the lower bound of the range for a tuning parameter. This parameter can be used only for numeric tuning parameters, and is ignored if a value list is specified for the tuning parameter.

AliasLB
namePath="string"

specifies the name path of a tuning parameter. For a nested action parameter, this parameter specifies a dot-separated path that includes all its parent parameter names. For a top-level action parameter, this parameter is simply the name of the parameter.

Aliasname
upperBound=double

specifies the upper bound of the range for a tuning parameter. This parameter can be used only for numeric tuning parameters, and is ignored if a value list is specified for the tuning parameter.

AliasUB
valueList={any-list-or-data-type-1 <, any-list-or-data-type-2, ...>}

specifies the list of values to try for a tuning parameter. If this parameter is specified, the lower bound and upper bound values of the tuning parameter are ignored.

useParameters="COMBINED" | "CUSTOM" | "STANDARD"

specifies whether to use only standard tuning parameters, only custom tuning parameters, or a combination of both.

DefaultCOMBINED

tuneFactMac Action

Automatically adjusts factorization machine parameters to tune a model for minimum error.

Lua Syntax

results, info = s:autotune_tuneFactMac{
scoreOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>},
trainOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>},
tunerOptions={
evaluationHistory=true | false,
logLevel=integer,
maxEvals=integer,
maxIters=integer,
maxTime=double,
maxTrainTime=64-bit-integer,
nCrossValFolds=integer,
nParallel=integer,
popSize=integer,
sampleSize=integer,
seed=integer,
shuffleGridPoints=true | false,
targetEvent="string",
userDefinedPartition=true | false,
},
tuningParameters={{
exclude=true | false,
initValue=integer | 64-bit-integer | double | true | false | "string",
lowerBound=double,
namePath="string",
upperBound=double,
valueList={any-list-or-data-type-1 <, any-list-or-data-type-2, ...>}
}, {...}},
useParameters="COMBINED" | "CUSTOM" | "STANDARD"
}

Parameter Descriptions

scoreOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}

specifies a list of parameters to be used by the score action in the aStore action set; for more information, see the parameters for that action.

trainOptions={key-1=any-list-or-data-type-1 <, key-2=any-list-or-data-type-2, ...>}

specifies a list of parameters to be used by the factmac action in the factmac action set; for more information, see the parameters for that action.

tunerOptions={tunerOptions}

specifies a list of parameters for adjusting how the tuner behaves.

AliasoptMinerOpts
evaluationHistory=true | false

when set to True, creates a table of every evaluation (every model configuration) and returns those tables to the client.

Defaulttrue
logLevel=integer

specifies the level of log messages to be written: no logs (0), initialization and completion logs (1), setup and iteration results (2), or detailed evaluation results (3).

Default2
Range0–3
maxBayesianModelSize=integer

specifies the maximum number of points in the kriging model. This parameter is ignored unless the value "BAYESIAN" is specified in the searchMethod parameter.

AliasmaxBayesianSize
Default100
Minimum value10
maxEvals=integer

specifies the maximum number of objective evaluations (model configurations) to be trained and scored.

AliasmaxEvaluations
Default50
Minimum value1
maxIters=integer

specifies the maximum number of iterations of local search optimization for model tuning. At each iteration, the number of evaluated models is less than or equal to the value of the popSize parameter.

AliasmaxIterations
Default5
Minimum value1
maxTime=double

specifies the maximum time allowed for tuning. The actual time might exceed this value by the amount of time required to train and score a model configuration.

Default36000
Minimum value1
maxTrainTime=64-bit-integer

specifies the maximum time allowed for a single model training. The model training is terminated if it exceeds this time, and the objective value is set to missing.

Minimum value0
nCrossValFolds=integer

specifies the number of folds to use for cross validation to assess model fit error as the tuning objective.

AliasnFolds
Default5
Minimum value2
nParallel=integer

specifies the number of threads to be used by the optimizer for parallel evaluation of model configurations.

Default0
nSubsessionWorkers=integer

specifies the number of workers to be used by each subsession for parallel evaluation of model configurations.

AliasnSubWorkers
Default0
Minimum value0
objective="ASE" | "AUC" | "F05" | "F1" | "GAMMA" | "GINI" | "KS" | "MAE" | "MCE" | "MCLL" | "MISC" | "MSE" | "MSLE" | "RASE" | "RMAE" | "RMSLE" | "TAU"

specifies the objective metric to use during the tuning.

ASE

specifies the average square error as the objective function.

AUC

specifies area under the curve as the objective function (nominal targets only).

F05

specifies the F0.5 coefficient as the objective function (nominal targets only).

F1

specifies the F1 coefficient as the objective function (nominal targets only).

GAMMA

specifies the gamma coefficient as the objective function (nominal targets only).

GINI

specifies the Gini coefficient as the objective function (nominal targets only).

KS

specifies the Kolmogorov-Smirnov coefficient as the objective function (nominal targets only).

MAE

specifies the mean absolute error as the objective function (interval targets only).

MCE

specifies the misclassification rate as the objective function (nominal targets only).

MCLL

specifies the multiclass log loss as the objective function (nominal targets only).

MISC

specifies the misclassification error percentage as the objective function (nominal targets only).

MSE

specifies the mean square error as the objective function (interval targets only).

MSLE

specifies the mean square logarithmic error as the objective function (interval targets only).

RASE

specifies the root average square error as the objective function.

RMAE

specifies the root mean absolute error as the objective function (interval targets only).

RMSLE

specifies the root mean square logarithmic error as the objective function (interval targets only).

TAU

specifies the tau coefficient as the objective function (nominal targets only).

popSize=integer

specifies the maximum number of model configurations to evaluate at each iteration.

AliaspopulationSize
Default10
Minimum value2
sampleSize=integer

specifies the number of model configurations to evaluate when the value of the searchMethod parameter is "LHS" or "RANDOM".

Default50
Minimum value2
searchMethod="BAYESIAN" | "GA" | "GRID" | "LHS" | "RANDOM"

specifies the search method to use during tuning.

DefaultGA
BAYESIAN

uses the Bayesian search method.

GA

uses the genetic algorithm search method.

GRID

uses the grid search method.

LHS

uses the Latin hypercube sample search method.

RANDOM

uses the random search method.

seed=integer

specifies the seed to use for selecting training data and validation data for a single validation partition or fold number for each observation for cross validation.

Default0
shuffleGridPoints=true | false

when set to True, shuffles the points that are generated by the grid search method before execution.

Defaulttrue
targetEvent="string"

specifies the name of the nominal target event to use for tuning.

trainPartitionFraction=double

specifies the size of a single data partition to use for model training during tuning.

AliastrainFraction
Range0.01–0.99
userDefinedPartition=true | false

when set to True, includes a user-defined partition for training and scoring.

Defaultfalse
validationPartitionFraction=double

specifies the size of a single validation partition to use in order to assess model fit error as the tuning objective.

AliasvalidateFraction
Default0.3
Range0.01–0.99

tuningParameters={{autotuneTuningParmDefinition-1} <, {autotuneTuningParmDefinition-2}, ...>}

specifies a list of custom tuning parameters and their definitions.

The autotuneTuningParmDefinition value can be one or more of the following:

exclude=true | false

when set to True, excludes a tuning parameter from the tuning process.

Defaultfalse
initValue=integer | 64-bit-integer | double | true | false | "string"

specifies the initial (default) value of a tuning parameter.

lowerBound=double

specifies the lower bound of the range for a tuning parameter. This parameter can be used only for numeric tuning parameters, and is ignored if a value list is specified for the tuning parameter.

AliasLB
namePath="string"

specifies the name path of a tuning parameter. For a nested action parameter, this parameter specifies a dot-separated path that includes all its parent parameter names. For a top-level action parameter, this parameter is simply the name of the parameter.

Aliasname
upperBound=double

specifies the upper bound of the range for a tuning parameter. This parameter can be used only for numeric tuning parameters, and is ignored if a value list is specified for the tuning parameter.

AliasUB
valueList={any-list-or-data-type-1 <, any-list-or-data-type-2, ...>}

specifies the list of values to try for a tuning parameter. If this parameter is specified, the lower bound and upper bound values of the tuning parameter are ignored.

useParameters="COMBINED" | "CUSTOM" | "STANDARD"

specifies whether to use only standard tuning parameters, only custom tuning parameters, or a combination of both.

DefaultCOMBINED

tuneFactMac Action

Automatically adjusts factorization machine parameters to tune a model for minimum error.

Python Syntax

results= s.autotune.tuneFactMac(
scoreOptions={"key-1":{any-list-or-data-type-1} <, "key-2":{any-list-or-data-type-2}, ...>},
trainOptions={"key-1":{any-list-or-data-type-1} <, "key-2":{any-list-or-data-type-2}, ...>},
tunerOptions={
"evaluationHistory":True | False,
"logLevel":integer,
"maxEvals":integer,
"maxIters":integer,
"maxTime":double,
"maxTrainTime":64-bit-integer,
"nCrossValFolds":integer,
"nParallel":integer,
"nSubsessionWorkers":integer,
"popSize":integer,
"sampleSize":integer,
"seed":integer,
"shuffleGridPoints":True | False,
"targetEvent":"string",
"userDefinedPartition":True | False,
},
tuningParameters=[{
"exclude":True | False,
"initValue":integer | 64-bit-integer | double | True | False | "string",
"lowerBound":double,
"namePath":"string",
"upperBound":double,
"valueList":{any-list-or-data-type-1 <, any-list-or-data-type-2, ...>}
}<, {...}>],
useParameters="COMBINED" | "CUSTOM" | "STANDARD"
)

Parameter Descriptions

scoreOptions={"key-1":{any-list-or-data-type-1} <, "key-2":{any-list-or-data-type-2}, ...>}

specifies a list of parameters to be used by the score action in the aStore action set; for more information, see the parameters for that action.

trainOptions={"key-1":{any-list-or-data-type-1} <, "key-2":{any-list-or-data-type-2}, ...>}

specifies a list of parameters to be used by the factmac action in the factmac action set; for more information, see the parameters for that action.

tunerOptions={tunerOptions}

specifies a list of parameters for adjusting how the tuner behaves.

AliasoptMinerOpts
"evaluationHistory":True | False

when set to True, creates a table of every evaluation (every model configuration) and returns those tables to the client.

DefaultTrue
"logLevel":integer

specifies the level of log messages to be written: no logs (0), initialization and completion logs (1), setup and iteration results (2), or detailed evaluation results (3).

Default2
Range0–3
"maxBayesianModelSize":integer

specifies the maximum number of points in the kriging model. This parameter is ignored unless the value "BAYESIAN" is specified in the searchMethod parameter.

AliasmaxBayesianSize
Default100
Minimum value10
"maxEvals":integer

specifies the maximum number of objective evaluations (model configurations) to be trained and scored.

AliasmaxEvaluations
Default50
Minimum value1
"maxIters":integer

specifies the maximum number of iterations of local search optimization for model tuning. At each iteration, the number of evaluated models is less than or equal to the value of the popSize parameter.

AliasmaxIterations
Default5
Minimum value1
"maxTime":double

specifies the maximum time allowed for tuning. The actual time might exceed this value by the amount of time required to train and score a model configuration.

Default36000
Minimum value1
"maxTrainTime":64-bit-integer

specifies the maximum time allowed for a single model training. The model training is terminated if it exceeds this time, and the objective value is set to missing.

Minimum value0
"nCrossValFolds":integer

specifies the number of folds to use for cross validation to assess model fit error as the tuning objective.

AliasnFolds
Default5
Minimum value2
"nParallel":integer

specifies the number of threads to be used by the optimizer for parallel evaluation of model configurations.

Default0
"nSubsessionWorkers":integer

specifies the number of workers to be used by each subsession for parallel evaluation of model configurations.

AliasnSubWorkers
Default0
Minimum value0
"objective":"ASE" | "AUC" | "F05" | "F1" | "GAMMA" | "GINI" | "KS" | "MAE" | "MCE" | "MCLL" | "MISC" | "MSE" | "MSLE" | "RASE" | "RMAE" | "RMSLE" | "TAU"

specifies the objective metric to use during the tuning.

ASE

specifies the average square error as the objective function.

AUC

specifies area under the curve as the objective function (nominal targets only).

F05

specifies the F0.5 coefficient as the objective function (nominal targets only).

F1

specifies the F1 coefficient as the objective function (nominal targets only).

GAMMA

specifies the gamma coefficient as the objective function (nominal targets only).

GINI

specifies the Gini coefficient as the objective function (nominal targets only).

KS

specifies the Kolmogorov-Smirnov coefficient as the objective function (nominal targets only).

MAE

specifies the mean absolute error as the objective function (interval targets only).

MCE

specifies the misclassification rate as the objective function (nominal targets only).

MCLL

specifies the multiclass log loss as the objective function (nominal targets only).

MISC

specifies the misclassification error percentage as the objective function (nominal targets only).

MSE

specifies the mean square error as the objective function (interval targets only).

MSLE

specifies the mean square logarithmic error as the objective function (interval targets only).

RASE

specifies the root average square error as the objective function.

RMAE

specifies the root mean absolute error as the objective function (interval targets only).

RMSLE

specifies the root mean square logarithmic error as the objective function (interval targets only).

TAU

specifies the tau coefficient as the objective function (nominal targets only).

"popSize":integer

specifies the maximum number of model configurations to evaluate at each iteration.

AliaspopulationSize
Default10
Minimum value2
"sampleSize":integer

specifies the number of model configurations to evaluate when the value of the searchMethod parameter is "LHS" or "RANDOM".

Default50
Minimum value2
"searchMethod":"BAYESIAN" | "GA" | "GRID" | "LHS" | "RANDOM"

specifies the search method to use during tuning.

DefaultGA
BAYESIAN

uses the Bayesian search method.

GA

uses the genetic algorithm search method.

GRID

uses the grid search method.

LHS

uses the Latin hypercube sample search method.

RANDOM

uses the random search method.

"seed":integer

specifies the seed to use for selecting training data and validation data for a single validation partition or fold number for each observation for cross validation.

Default0
"shuffleGridPoints":True | False

when set to True, shuffles the points that are generated by the grid search method before execution.

DefaultTrue
"targetEvent":"string"

specifies the name of the nominal target event to use for tuning.

"trainPartitionFraction":double

specifies the size of a single data partition to use for model training during tuning.

AliastrainFraction
Range0.01–0.99
"userDefinedPartition":True | False

when set to True, includes a user-defined partition for training and scoring.

DefaultFalse
"validationPartitionFraction":double

specifies the size of a single validation partition to use in order to assess model fit error as the tuning objective.

AliasvalidateFraction
Default0.3
Range0.01–0.99

tuningParameters=[{autotuneTuningParmDefinition-1} <, {autotuneTuningParmDefinition-2}, ...>]

specifies a list of custom tuning parameters and their definitions.

The autotuneTuningParmDefinition value can be one or more of the following:

"exclude":True | False

when set to True, excludes a tuning parameter from the tuning process.

DefaultFalse
"initValue":integer | 64-bit-integer | double | True | False | "string"

specifies the initial (default) value of a tuning parameter.

"lowerBound":double

specifies the lower bound of the range for a tuning parameter. This parameter can be used only for numeric tuning parameters, and is ignored if a value list is specified for the tuning parameter.

AliasLB
"namePath":"string"

specifies the name path of a tuning parameter. For a nested action parameter, this parameter specifies a dot-separated path that includes all its parent parameter names. For a top-level action parameter, this parameter is simply the name of the parameter.

Aliasname
"upperBound":double

specifies the upper bound of the range for a tuning parameter. This parameter can be used only for numeric tuning parameters, and is ignored if a value list is specified for the tuning parameter.

AliasUB
"valueList":{any-list-or-data-type-1 <, any-list-or-data-type-2, ...>}

specifies the list of values to try for a tuning parameter. If this parameter is specified, the lower bound and upper bound values of the tuning parameter are ignored.

useParameters="COMBINED" | "CUSTOM" | "STANDARD"

specifies whether to use only standard tuning parameters, only custom tuning parameters, or a combination of both.

DefaultCOMBINED

tuneFactMac Action

Automatically adjusts factorization machine parameters to tune a model for minimum error.

R Syntax

results <– cas.autotune.tuneFactMac(s,
scoreOptions=list(key-1=list(any-list-or-data-type-1) <, key-2=list(any-list-or-data-type-2), ...>),
trainOptions=list(key-1=list(any-list-or-data-type-1) <, key-2=list(any-list-or-data-type-2), ...>),
tunerOptions=list(
evaluationHistory=TRUE | FALSE,
logLevel=integer,
maxEvals=integer,
maxIters=integer,
maxTime=double,
maxTrainTime=64-bit-integer,
nCrossValFolds=integer,
nParallel=integer,
popSize=integer,
sampleSize=integer,
seed=integer,
shuffleGridPoints=TRUE | FALSE,
targetEvent="string",
userDefinedPartition=TRUE | FALSE,
),
tuningParameters=list( list(
exclude=TRUE | FALSE,
initValue=integer | 64-bit-integer | double | TRUE | FALSE | "string",
lowerBound=double,
namePath="string",
upperBound=double,
valueList=list(any-list-or-data-type-1 <, any-list-or-data-type-2, ...>)
) <, list(...)>),
useParameters="COMBINED" | "CUSTOM" | "STANDARD"
)

Parameter Descriptions

scoreOptions=list(key-1=list(any-list-or-data-type-1) <, key-2=list(any-list-or-data-type-2), ...>)

specifies a list of parameters to be used by the score action in the aStore action set; for more information, see the parameters for that action.

trainOptions=list(key-1=list(any-list-or-data-type-1) <, key-2=list(any-list-or-data-type-2), ...>)

specifies a list of parameters to be used by the factmac action in the factmac action set; for more information, see the parameters for that action.

tunerOptions=list(tunerOptions)

specifies a list of parameters for adjusting how the tuner behaves.

AliasoptMinerOpts
evaluationHistory=TRUE | FALSE

when set to True, creates a table of every evaluation (every model configuration) and returns those tables to the client.

DefaultTRUE
logLevel=integer

specifies the level of log messages to be written: no logs (0), initialization and completion logs (1), setup and iteration results (2), or detailed evaluation results (3).

Default2
Range0–3
maxBayesianModelSize=integer

specifies the maximum number of points in the kriging model. This parameter is ignored unless the value "BAYESIAN" is specified in the searchMethod parameter.

AliasmaxBayesianSize
Default100
Minimum value10
maxEvals=integer

specifies the maximum number of objective evaluations (model configurations) to be trained and scored.

AliasmaxEvaluations
Default50
Minimum value1
maxIters=integer

specifies the maximum number of iterations of local search optimization for model tuning. At each iteration, the number of evaluated models is less than or equal to the value of the popSize parameter.

AliasmaxIterations
Default5
Minimum value1
maxTime=double

specifies the maximum time allowed for tuning. The actual time might exceed this value by the amount of time required to train and score a model configuration.

Default36000
Minimum value1
maxTrainTime=64-bit-integer

specifies the maximum time allowed for a single model training. The model training is terminated if it exceeds this time, and the objective value is set to missing.

Minimum value0
nCrossValFolds=integer

specifies the number of folds to use for cross validation to assess model fit error as the tuning objective.

AliasnFolds
Default5
Minimum value2
nParallel=integer

specifies the number of threads to be used by the optimizer for parallel evaluation of model configurations.

Default0
nSubsessionWorkers=integer

specifies the number of workers to be used by each subsession for parallel evaluation of model configurations.

AliasnSubWorkers
Default0
Minimum value0
objective="ASE" | "AUC" | "F05" | "F1" | "GAMMA" | "GINI" | "KS" | "MAE" | "MCE" | "MCLL" | "MISC" | "MSE" | "MSLE" | "RASE" | "RMAE" | "RMSLE" | "TAU"

specifies the objective metric to use during the tuning.

ASE

specifies the average square error as the objective function.

AUC

specifies area under the curve as the objective function (nominal targets only).

F05

specifies the F0.5 coefficient as the objective function (nominal targets only).

F1

specifies the F1 coefficient as the objective function (nominal targets only).

GAMMA

specifies the gamma coefficient as the objective function (nominal targets only).

GINI

specifies the Gini coefficient as the objective function (nominal targets only).

KS

specifies the Kolmogorov-Smirnov coefficient as the objective function (nominal targets only).

MAE

specifies the mean absolute error as the objective function (interval targets only).

MCE

specifies the misclassification rate as the objective function (nominal targets only).

MCLL

specifies the multiclass log loss as the objective function (nominal targets only).

MISC

specifies the misclassification error percentage as the objective function (nominal targets only).

MSE

specifies the mean square error as the objective function (interval targets only).

MSLE

specifies the mean square logarithmic error as the objective function (interval targets only).

RASE

specifies the root average square error as the objective function.

RMAE

specifies the root mean absolute error as the objective function (interval targets only).

RMSLE

specifies the root mean square logarithmic error as the objective function (interval targets only).

TAU

specifies the tau coefficient as the objective function (nominal targets only).

popSize=integer

specifies the maximum number of model configurations to evaluate at each iteration.

AliaspopulationSize
Default10
Minimum value2
sampleSize=integer

specifies the number of model configurations to evaluate when the value of the searchMethod parameter is "LHS" or "RANDOM".

Default50
Minimum value2
searchMethod="BAYESIAN" | "GA" | "GRID" | "LHS" | "RANDOM"

specifies the search method to use during tuning.

DefaultGA
BAYESIAN

uses the Bayesian search method.

GA

uses the genetic algorithm search method.

GRID

uses the grid search method.

LHS

uses the Latin hypercube sample search method.

RANDOM

uses the random search method.

seed=integer

specifies the seed to use for selecting training data and validation data for a single validation partition or fold number for each observation for cross validation.

Default0
shuffleGridPoints=TRUE | FALSE

when set to True, shuffles the points that are generated by the grid search method before execution.

DefaultTRUE
targetEvent="string"

specifies the name of the nominal target event to use for tuning.

trainPartitionFraction=double

specifies the size of a single data partition to use for model training during tuning.

AliastrainFraction
Range0.01–0.99
userDefinedPartition=TRUE | FALSE

when set to True, includes a user-defined partition for training and scoring.

DefaultFALSE
validationPartitionFraction=double

specifies the size of a single validation partition to use in order to assess model fit error as the tuning objective.

AliasvalidateFraction
Default0.3
Range0.01–0.99

tuningParameters=list( list(autotuneTuningParmDefinition-1) <, list(autotuneTuningParmDefinition-2), ...>)

specifies a list of custom tuning parameters and their definitions.

The autotuneTuningParmDefinition value can be one or more of the following:

exclude=TRUE | FALSE

when set to True, excludes a tuning parameter from the tuning process.

DefaultFALSE
initValue=integer | 64-bit-integer | double | TRUE | FALSE | "string"

specifies the initial (default) value of a tuning parameter.

lowerBound=double

specifies the lower bound of the range for a tuning parameter. This parameter can be used only for numeric tuning parameters, and is ignored if a value list is specified for the tuning parameter.

AliasLB
namePath="string"

specifies the name path of a tuning parameter. For a nested action parameter, this parameter specifies a dot-separated path that includes all its parent parameter names. For a top-level action parameter, this parameter is simply the name of the parameter.

Aliasname
upperBound=double

specifies the upper bound of the range for a tuning parameter. This parameter can be used only for numeric tuning parameters, and is ignored if a value list is specified for the tuning parameter.

AliasUB
valueList=list(any-list-or-data-type-1 <, any-list-or-data-type-2, ...>)

specifies the list of values to try for a tuning parameter. If this parameter is specified, the lower bound and upper bound values of the tuning parameter are ignored.

useParameters="COMBINED" | "CUSTOM" | "STANDARD"

specifies whether to use only standard tuning parameters, only custom tuning parameters, or a combination of both.

DefaultCOMBINED
Last updated: June 07, 2018