Class GWRBasic

Inheritance Relationships

Base Types

Derived Type

Class Documentation

class GWRBasic : public gwm::GWRBase, public gwm::IBandwidthSelectable, public gwm::IVarialbeSelectable, public gwm::IParallelizable, public gwm::IParallelOpenmpEnabled, public gwm::IParallelCudaEnabled, public gwm::IParallelMpiEnabled

Basic implementation of geographically weighted regression. This algorithm can auto select bandwidth and variables. This algorithm can speed-up by OpenMP.

Subclassed by gwm::GWRRobust

Public Types

enum BandwidthSelectionCriterionType

Type of criterion for bandwidth selection.

Values:

enumerator AIC

AIC.

enumerator CV

CV.

typedef arma::mat (GWRBasic::* PredictCalculator)(const arma::mat&, const arma::mat&, const arma::vec&)

Predict function declaration.

typedef arma::mat (GWRBasic::* FitCalculator)()

Fit function declaration.

typedef arma::mat (GWRBasic::* FitCoreCalculator)(const arma::mat&, const arma::vec&, const SpatialWeight&)

Fit function declaration.

typedef arma::mat (GWRBasic::* FitCoreSHatCalculator)(const arma::mat&, const arma::vec&, const SpatialWeight&, arma::vec&)

Fit function declaration.

typedef arma::mat (GWRBasic::* FitCoreCVCalculator)(const arma::mat&, const arma::vec&, const SpatialWeight&)

Fit function declaration.

typedef double (GWRBasic::* BandwidthSelectionCriterionCalculator)(BandwidthWeight*)

Declaration of criterion calculator for bandwidth selection.

typedef double (GWRBasic::* IndepVarsSelectCriterionCalculator)(const std::vector<std::size_t>&)

Declaration of criterion calculator for variable selection.

Public Functions

inline GWRBasic()

Construct a new CGwmGWRBasic object.

inline GWRBasic(const arma::mat &x, const arma::vec &y, const arma::mat &coords, const SpatialWeight &spatialWeight, bool hasHatMatrix = true, bool hasIntercept = true)

Construct a new CGwmGWRBasic object.

Parameters:

  • x – Independent variables.

  • y – Dependent variable.

  • coords – Coordinate matrix.

  • spatialWeight – Spatial weighting configuration.

  • hasHatMatrix – Whether has hat-matrix.

  • hasIntercept – Whether has intercept.

inline ~GWRBasic()

Destroy the CGwmGWRBasic object.

inline bool isAutoselectBandwidth() const

Get whether auto select bandwidth.

Returns:

true if auto select bandwidth.

Returns:

false if not auto select bandwidth.

inline void setIsAutoselectBandwidth(bool isAutoSelect)

Set whether auto select bandwidth.

Parameters:

isAutoSelect – true if auto select bandwidth, otherwise false.

inline BandwidthSelectionCriterionType bandwidthSelectionCriterion() const

Get type of criterion for bandwidth selection.

Returns:

BandwidthSelectionCriterionType Type of criterion for bandwidth selection.

void setBandwidthSelectionCriterion(const BandwidthSelectionCriterionType &criterion)

Set type of criterion for bandwidth selection.

Parameters:

criterion – Type of criterion for bandwidth selection.

inline void setGoldenUpperBounds(double value)

Set the upper bounds of golden selection.

inline void setGoldenLowerBounds(double value)

Set the lower bounds of golden selection.

inline bool isAutoselectIndepVars() const

Get whether auto select variables.

Returns:

true if auto select variables.

Returns:

false if not auto select variables.

inline void setIsAutoselectIndepVars(bool isAutoSelect)

Set whether auto select variables.

Parameters:

isAutoSelect – true if auto select variables, otherwise false.

inline double indepVarSelectionThreshold() const

Get threshold for variable selection.

Returns:

double Threshold for variable selection.

inline void setIndepVarSelectionThreshold(double threshold)

Set threshold for variable selection.

Parameters:

threshold – Threshold for variable selection. This value dependends on the size of samples. Larger samples, larger threshold.

inline const VariablesCriterionList &indepVarsSelectionCriterionList() const

Get criterion list for variable selection.

Returns:

VariablesCriterionList Criterion list for variable selection.

inline const BandwidthCriterionList &bandwidthSelectionCriterionList() const

Get criterion list for bandwidth selection.

Returns:

BandwidthCriterionList Criterion list for bandwidth selection.

inline bool hasHatMatrix() const

Get whether has hat-matrix.

Returns:

true if has hat-matrix.

Returns:

false if does not have hat-matrix.

inline void setHasHatMatrix(const bool has)

Set the Has Hat-Matrix object.

Parameters:

has – true if has hat-matrix, otherwise false.

inline const arma::mat &betasSE()

Get standard errors of coefficient estimates.

Returns:

arma::mat Standard errors of coefficient estimates.

inline const arma::vec &sHat()

Get a vector of \(tr(S)\) and \(tr(SS^T)\).

Returns:

arma::vec A vector of \(tr(S)\) and \(tr(SS^T)\).

inline const arma::vec &qDiag()

Get the diagonal elements of matrix \(Q\).

Returns:

arma::vec The diagonal elements of matrix \(Q\).

inline const arma::mat &s()

Get the hat-matrix \(S\).

Returns:

arma::mat The hat-matrix \(S\).

inline const arma::cube &c()
inline bool isStoreS()

Whether to store hat-matrix \(S\).

Returns:

true if store hat-matrix.

Returns:

false if not to store hat-matrix.

inline bool isStoreC()
inline void setStoreS(bool flag)
inline void setStoreC(bool flag)
virtual bool isValid() override

Check whether the algorithm’s configuration is valid.

Returns:

true if the algorithm’s configuration is valid.

Returns:

false if the algorithm’s configuration is invalid.

virtual arma::mat predict(const arma::mat &locations) override

Predict coefficients on specified locations.

Parameters:

locations – Locations where to predict coefficients.

Returns:

mat Predicted coefficients.

virtual arma::mat fit() override

Fit coefficient estimates.

Returns:

mat Coefficient estimates

inline Status getCriterion(const std::vector<size_t> &variables, double &criterion) override
inline virtual std::vector<std::size_t> selectedVariables() override

Get selected variables.

Returns:

std::vector<std::size_t> Selected variables.

double indepVarsSelectionCriterion(const std::vector<std::size_t> &indepVars)

Get AIC value with given variables for variable optimization (serial implementation).

Parameters:

indepVars – Given variables

Returns:

double Criterion value

inline virtual Status getCriterion(BandwidthWeight *weight, double &criterion) override

Get criterion value with given bandwidth for bandwidth optimization.

Parameters:

  • weight – Given bandwidth.

  • criterion – [out] Criterion value.

Returns:

Status Algorithm status.

double bandwidthSizeCriterionCV(BandwidthWeight *bandwidthWeight)

Get CV value with given bandwidth for bandwidth optimization (serial implementation).

Parameters:

bandwidthWeight – Given bandwidth

Returns:

double Criterion value

double bandwidthSizeCriterionAIC(BandwidthWeight *bandwidthWeight)

Get AIC value with given bandwidth for bandwidth optimization (serial implementation).

Parameters:

bandwidthWeight – Given bandwidth

Returns:

double Criterion value

inline virtual int parallelAbility() const override

Return the parallel ability of this algorithm.

Returns:

Bitwise OR of aviliable parallel types of this algorithm.

inline virtual ParallelType parallelType() const override

Return the parallel type of this algorithm.

Returns:

Parallel type of this algorithm

virtual void setParallelType(const ParallelType &type) override

Set the parallel type of this algorithm.

Parameters:

type – Parallel type of this algorithm.

inline virtual void setOmpThreadNum(const int threadNum) override

Set the thread numbers while paralleling.

Parameters:

threadNum – Number of threads.

inline virtual void setGPUId(const int gpuId) override

Set ID of used GPU while paralleling.

Parameters:

gpuId – ID of used GPU. Start from 0.

inline virtual void setGroupSize(const std::size_t size) override

Set the group size while paralleling.

Parameters:

size – Group size \(g\) while paralleling. Usually this is up to memory size \(m\) (B) of GPU with ID gpuId. If there are \(n\) samples and \(k\) variables,

\[ k * n * g * 8 < m \]
For most GPU, 64 is OK.

inline virtual int workerId() override
inline virtual void setWorkerId(int id) override
inline virtual void setWorkerNum(int size) override

Public Static Attributes

static std::unordered_map<BandwidthSelectionCriterionType, std::string> BandwidthSelectionCriterionTypeNameMapper

Protected Functions

void createPredictionDistanceParameter(const arma::mat &locations)

Create distance parameters for prediction.

Parameters:

locations – Distance parameters for prediction.

Protected Attributes

bool mHasHatMatrix = true

Whether has hat-matrix.

bool mHasFTest = false

Todo:

Whether has F-test

bool mHasPredict = false

Deprecated:

Whether has variables to predict dependent variable.

bool mIsAutoselectIndepVars = false

Whether to auto select variables.

double mIndepVarSelectionThreshold = 3.0

The threshold for variable selection.

IndepVarsSelectCriterionCalculator mIndepVarsSelectionCriterionFunction = &GWRBasic::indepVarsSelectionCriterion

Criterion calculator for variable selection.

VariablesCriterionList mIndepVarsSelectionCriterionList

Criterion list of each variable combination.

std::vector<std::size_t> mSelectedIndepVars

Selected variables.

std::size_t mIndepVarSelectionProgressTotal = 0

Total number of independent variable combination.

std::size_t mIndepVarSelectionProgressCurrent = 0

Current progress of independent variable selection.

bool mIsAutoselectBandwidth = false

Whether to auto select bandwidth.

BandwidthSelectionCriterionType mBandwidthSelectionCriterion = BandwidthSelectionCriterionType::AIC

Type criterion for bandwidth selection.

BandwidthSelectionCriterionCalculator mBandwidthSelectionCriterionFunction = &GWRBasic::bandwidthSizeCriterionCV

Criterion calculator for bandwidth selection.

BandwidthCriterionList mBandwidthSelectionCriterionList

Criterion list of each bandwidth.

double mBandwidthLastCriterion = DBL_MAX

Last criterion for bandwidth selection.

std::optional<double> mGoldenUpperBounds
std::optional<double> mGoldenLowerBounds
PredictCalculator mPredictFunction = &GWRBasic::predictSerial

Implementation of predict function.

FitCalculator mFitFunction = &GWRBasic::fitBase

Implementation of fit function.

FitCoreCalculator mFitCoreFunction = &GWRBasic::fitCoreSerial

Implementation of fit function.

FitCoreSHatCalculator mFitCoreSHatFunction = &GWRBasic::fitCoreSHatSerial

Implementation of fit function.

FitCoreCVCalculator mFitCoreCVFunction = &GWRBasic::fitCoreCVSerial

Implementation of fit function.

ParallelType mParallelType = ParallelType::SerialOnly

Type of parallel method.

int mOmpThreadNum = 8

Number of threads to create.

size_t mGroupLength = 64

Size of a group computing together.

int mGpuId = 0

The ID of selected GPU.

int mWorkerId = 0
int mWorkerNum = 1
arma::uword mWorkRangeSize = 0
std::optional<std::pair<arma::uword, arma::uword>> mWorkRange
arma::mat mBetasSE

Standard errors of coefficient estimates.

arma::vec mSHat

A vector of \(tr(S)\) and \(tr(SS^T)\).

arma::vec mQDiag

The diagonal elements of matrix \(Q\).

arma::mat mS

The hat-matrix \(S\).

arma::cube mC

All \(S\) matrices.

bool mStoreS = false

Whether to save S.

bool mStoreC = false

Whether to save C.