ai.libs.jaicore.ml.core.optimizing.graddesc

Interface GradientDescentOptimizerConfig

All Superinterfaces:

org.aeonbits.owner.Config, org.aeonbits.owner.Mutable, java.io.Serializable

@Config.Sources(value="file:conf/opt/graddesc.properties")
public interface GradientDescentOptimizerConfig
extends org.aeonbits.owner.Mutable

Nested Class Summary

Nested classes/interfaces inherited from interface org.aeonbits.owner.Config

org.aeonbits.owner.Config.ConverterClass, org.aeonbits.owner.Config.DecryptorClass, org.aeonbits.owner.Config.DefaultValue, org.aeonbits.owner.Config.DisableableFeature, org.aeonbits.owner.Config.DisableFeature, org.aeonbits.owner.Config.EncryptedValue, org.aeonbits.owner.Config.HotReload, org.aeonbits.owner.Config.HotReloadType, org.aeonbits.owner.Config.Key, org.aeonbits.owner.Config.LoadPolicy, org.aeonbits.owner.Config.LoadType, org.aeonbits.owner.Config.PreprocessorClasses, org.aeonbits.owner.Config.Separator, org.aeonbits.owner.Config.Sources, org.aeonbits.owner.Config.TokenizerClass

Field Summary

Fields

Modifier and Type	Field and Description
static java.lang.String	GRAD_DESC_GRADIENT_THRESHOLD Specifies a threshold for the gradient (i.e. if the gradient is below this value no update will be done; if all gradients are below this value, the algorithm will terminate)
static java.lang.String	GRAD_DESC_LEARNING_RATE The learning rate in the update step (i.e. how much of the gradient should be added to the parameter)
static java.lang.String	GRAD_DESC_MAX_ITERATIONS Specifies the maximum of gradient update steps.

Method Summary

Modifier and Type	Method and Description
double	gradientThreshold()
double	learningRate()
int	maxIterations()

Methods inherited from interface org.aeonbits.owner.Mutable

addPropertyChangeListener, addPropertyChangeListener, clear, load, load, removeProperty, removePropertyChangeListener, setProperty

Field Detail

GRAD_DESC_MAX_ITERATIONS

static final java.lang.String GRAD_DESC_MAX_ITERATIONS

Specifies the maximum of gradient update steps. Can be set to -1 to specify no threshold (the algorithm may not terminate in this case).

See Also:

Constant Field Values

GRAD_DESC_LEARNING_RATE

static final java.lang.String GRAD_DESC_LEARNING_RATE

The learning rate in the update step (i.e. how much of the gradient should be added to the parameter)

See Also:

Constant Field Values

GRAD_DESC_GRADIENT_THRESHOLD

static final java.lang.String GRAD_DESC_GRADIENT_THRESHOLD

Specifies a threshold for the gradient (i.e. if the gradient is below this value no update will be done; if all gradients are below this value, the algorithm will terminate)

See Also:

Constant Field Values

Method Detail

maxIterations

@Config.Key(value="graddesc.max_iterations")
 @Config.DefaultValue(value="20")
int maxIterations()

learningRate

@Config.Key(value="gradedesc.learning_rate")
 @Config.DefaultValue(value="0.01")
double learningRate()

gradientThreshold

@Config.Key(value="graddesc.gradient_threshold")
 @Config.DefaultValue(value="0.001")
double gradientThreshold()