scalation
API
Generated with
scalation/scalation/scalation.modeling/scalation.modeling.forecasting/scalation.modeling.forecasting.nonlinear/NARX_SR_D

NARX_SR_D

scalation.modeling.forecasting.nonlinear.NARX_SR_D
See theNARX_SR_D companion class
object NARX_SR_D

The NARX_SR_D companion object provides factory methods for the NARX_SR_D class.

Attributes

Companion
class
Graph
Supertypes
class Object
trait Matchable
class Any
Self type
NARX_SR_D.type

Members list

Value members

Concrete methods

def apply(xe: MatrixD, y: VectorD, hh: Int, fname_: Array[String] = ..., tRng: Range = ..., hparam: HyperParameter = ..., fEndo_enabled: LinkedHashSet[TransformT] = ..., fExo_enabled: Array[LinkedHashSet[TransformT]] = ..., bakcast: Boolean = ...): NARX_SR_D

Create an NARX_SR_D object by building an input matrix xy and then calling the NARX_SR_D constructor.

Create an NARX_SR_D object by building an input matrix xy and then calling the NARX_SR_D constructor.

Value parameters

bakcast

whether a backcasted value is prepended to the time series (defaults to false)

fEndo_enabled

the set of transforms to be used for the endogenous

fExo_enabled

the array containing the sets of transforms to be used for the exogenous

fname_

the feature/variable names

hh

the maximum forecasting horizon (h = 1 to hh)

hparam

the hyper-parameters

tRng

the time range, if relevant (time index may suffice)

xe

the matrix of exogenous variable values

y

the endogenous/response vector (main time series data)

Attributes

def buildMatrix(xe_bfill: MatrixD, y: VectorD, hp_: HyperParameter, fEndo: Array[Transform], fExo: Array[Array[Transform]], bakcast: Boolean): MatrixD

Build the input matrix by combining the spec + p columns for the trend and endogenous variable with the q * xe.dim2 columns for the exogenous variables. When cross = true, additional cross terms will be added. Columns produced by transformations will be added as well.

Build the input matrix by combining the spec + p columns for the trend and endogenous variable with the q * xe.dim2 columns for the exogenous variables. When cross = true, additional cross terms will be added. Columns produced by transformations will be added as well.

Value parameters

bakcast

whether a backcasted value is prepended to the time series (defaults to false)

fEndo

the transformation functions to apply on the endogenous variables

fExo

the transformation functions to apply on the exogenous variables

hp_

the hyper-parameters

xe_bfill

the matrix of exogenous variable values

y

the endogenous/response vector (main time series data)

Attributes

def formNames(n_exo: Int, hp_: HyperParameter, n_fEn: Int, n_fExArr: Array[Int]): Array[String]

Form an array of names for the features included in the model.

Form an array of names for the features included in the model.

Value parameters

hp_

the hyper-parameters

n_exo

the number of exogenous variable

n_fEn

the number of functions used to map endogenous variables

n_fExArr

the number of functions used to map each exogenous variables

Attributes

def getTransforms(w_nl: VectorD, fEndo_enabled: LinkedHashSet[TransformT], fExo_enabled: Array[LinkedHashSet[TransformT]]): (Array[Transform], Array[Array[Transform]])

Form arrays of transforms object using the vector of nonlinear parameters.

Form arrays of transforms object using the vector of nonlinear parameters.

Value parameters

fEndo_enabled

the set of transforms to be used for the endogenous

fExo_enabled

the array containing the sets of transforms to be used for the exogenous

w_nl

the vector of nonlinear parameters

Attributes

def initializeW(fEndo_enabled: LinkedHashSet[TransformT], fExo_enabled: Array[LinkedHashSet[TransformT]]): (VectorD, VectorD, VectorD)

Form vectors for the initial weights and their bounds for the transforms.

Form vectors for the initial weights and their bounds for the transforms.

Value parameters

fEndo_enabled

the set of transforms to be used for the endogenous

fExo_enabled

the array containing the sets of transforms to be used for the exogenous

Attributes

def optimize(xe_bfill: MatrixD, y: VectorD, hparam: HyperParameter = ..., fEndo_enabled: LinkedHashSet[TransformT] = ..., fExo_enabled: Array[LinkedHashSet[TransformT]] = ..., bakcast: Boolean = ...): VectorD

Fit the nonlinear + linear parameters using LBFGS_B.

Fit the nonlinear + linear parameters using LBFGS_B.

Value parameters

bakcast

whether a backcasted value is prepended to the time series (defaults to false)

fEndo

the transformation functions to apply on the endogenous variables

fExo

the transformation functions to apply on the exogenous variables

hp_

the hyper-parameters

xe_bfill

the matrix of exogenous variable values

y

the endogenous/response vector (main time series data)

Attributes

def optimize2(xe_bfill: MatrixD, y: VectorD, hparam: HyperParameter = ..., fEndo_enabled: LinkedHashSet[TransformT] = ..., fExo_enabled: Array[LinkedHashSet[TransformT]] = ..., bakcast: Boolean = ...): VectorD

Fit the nonlinear + linear parameters using LBFGS_B with VarPro.

Fit the nonlinear + linear parameters using LBFGS_B with VarPro.

Value parameters

bakcast

whether a backcasted value is prepended to the time series (defaults to false)

fEndo

the transformation functions to apply on the endogenous variables

fExo

the transformation functions to apply on the exogenous variables

hp_

the hyper-parameters

xe_bfill

the matrix of exogenous variable values

y

the endogenous/response vector (main time series data)

Attributes

def rescale(xe: MatrixD, y: VectorD, hh: Int, fname_: Array[String] = ..., tRng: Range = ..., hparam: HyperParameter = ..., fEndo_enabled: LinkedHashSet[TransformT] = ..., fExo_enabled: Array[LinkedHashSet[TransformT]] = ..., bakcast: Boolean = ..., tFormScale: (VectorD | MatrixD) => Transform = ...): NARX_SR_D

Create an NARX_SR_D object by building an input matrix xy and then calling the NARX_SR_D constructor, with rescaling of endogneous and exogenous variable values.

Create an NARX_SR_D object by building an input matrix xy and then calling the NARX_SR_D constructor, with rescaling of endogneous and exogenous variable values.

Value parameters

bakcast

whether a backcasted value is prepended to the time series (defaults to false)

fEndo_enabled

the set of transforms to be used for the endogenous

fExo_enabled

the array containing the sets of transforms to be used for the exogenous

fname_

the feature/variable names

hh

the maximum forecasting horizon (h = 1 to hh)

hparam

the hyper-parameters

tFormScale

the transform for y

tRng

the time range, if relevant (time index may suffice)

xe

the matrix of exogenous variable values

y

the endogenous/response vector (main time series data)

Attributes

Generated with