FunctionApproximatorRBFN Class Reference

RBFN (Radial Basis Function Network) function approximator. More...

#include <FunctionApproximatorRBFN.hpp>

Inheritance diagram for FunctionApproximatorRBFN:

Collaboration diagram for FunctionApproximatorRBFN:

Public Member Functions
	FunctionApproximatorRBFN (const MetaParametersRBFN *const meta_parameters, const ModelParametersRBFN *const model_parameters=NULL)
	Initialize a function approximator with meta- and model-parameters. More...
	FunctionApproximatorRBFN (const ModelParametersRBFN *const model_parameters)
	Initialize a function approximator with model parameters. More...
FunctionApproximator *	clone (void) const
	Return a pointer to a deep copy of the FunctionApproximator object. More...
void	train (const Eigen::Ref< const Eigen::MatrixXd > &inputs, const Eigen::Ref< const Eigen::MatrixXd > &targets)
	Train the function approximator with corresponding input and target examples. More...
void	predict (const Eigen::Ref< const Eigen::MatrixXd > &inputs, Eigen::MatrixXd &output)
	Query the function approximator to make a prediction. More...
void	preallocateMemory (int n_basis_functions)
	Preallocate memory to make certain functions real-time. More...
std::string	getName (void) const
	Get the name of this function approximator. More...
bool	saveGridData (const Eigen::VectorXd &min, const Eigen::VectorXd &max, const Eigen::VectorXi &n_samples_per_dim, std::string directory, bool overwrite=false) const
	Generate a grid of inputs, and output the response of the basis functions and line segments for these inputs. More...
Public Member Functions inherited from FunctionApproximator
	FunctionApproximator (const MetaParameters *const meta_parameters, const ModelParameters *const model_parameters=NULL)
	Initialize a function approximator with meta- and optionally model-parameters. More...
	FunctionApproximator (const ModelParameters *const model_parameters)
	Initialize a function approximator with model-parameters. More...
void	train (const Eigen::Ref< const Eigen::MatrixXd > &inputs, const Eigen::Ref< const Eigen::MatrixXd > &targets, std::string save_directory, bool overwrite=false)
	Train the function approximator with corresponding input and target examples (and write results to file). More...
void	reTrain (const Eigen::Ref< const Eigen::MatrixXd > &inputs, const Eigen::Ref< const Eigen::MatrixXd > &targets)
	Re-train the function approximator with corresponding input and target examples. More...
void	reTrain (const Eigen::Ref< const Eigen::MatrixXd > &inputs, const Eigen::Ref< const Eigen::MatrixXd > &targets, std::string save_directory, bool overwrite=false)
	Re-train the function approximator with corresponding input and target examples (and write results to file). More...
virtual void	predict (const Eigen::Ref< const Eigen::MatrixXd > &inputs, Eigen::MatrixXd &outputs, Eigen::MatrixXd &variances)
	Query the function approximator to make a prediction, and also to predict its variance. More...
virtual void	predict (const Eigen::Ref< const Eigen::MatrixXd > &inputs, Eigen::MatrixXd &outputs, std::vector< Eigen::MatrixXd > &variances)
	Query the function approximator to make a prediction, and also to predict its variance. More...
virtual void	predictVariance (const Eigen::Ref< const Eigen::MatrixXd > &inputs, Eigen::MatrixXd &variances)
	Query the function approximator to get the variance of a prediction This function is not implemented by all function approximators. More...
bool	isTrained (void) const
	Determine whether the function approximator has already been trained with data or not. More...
int	getExpectedInputDim (void) const
	The expected dimensionality of the input data. More...
int	getExpectedOutputDim (void) const
	The expected dimensionality of the output data. More...
void	getSelectableParameters (std::set< std::string > &selected_values_labels) const
	Return all the names of the parameter types that can be selected. More...
void	setSelectedParameters (const std::set< std::string > &selected_values_labels)
	Determine which subset of parameters is represented in the vector returned by Parameterizable::getParameterVectorSelected. More...
void	getParameterVectorSelectedMinMax (Eigen::VectorXd &min, Eigen::VectorXd &max) const
	Get the minimum and maximum of the selected parameters in one vector. More...
int	getParameterVectorSelectedSize (void) const
	Get the size of the vector of selected parameters, as returned by getParameterVectorSelected(. More...
void	setParameterVectorSelected (const Eigen::VectorXd &values, bool normalized=false)
	Set all the values of the selected parameters with one vector. More...
void	getParameterVectorSelected (Eigen::VectorXd &values, bool normalized=false) const
	Get the values of the selected parameters in one vector. More...
void	getParameterVectorMask (const std::set< std::string > selected_values_labels, Eigen::VectorXi &selected_mask) const
	Get a mask for selecting parameters. More...
int	getParameterVectorAllSize (void) const
	Get the size of the parameter values vector when it contains all available parameter values. More...
void	getParameterVectorAll (Eigen::VectorXd &values) const
	Return a vector that returns all available parameter values. More...
void	setParameterVectorAll (const Eigen::VectorXd &values)
	Set all available parameter values with one vector. More...
UnifiedModel *	getUnifiedModel (void) const
	Return a representation of this function approximator's model as a unified model. More...
std::string	toString (void) const
	Returns a string representation of the object. More...
const MetaParameters *	getMetaParameters (void) const
	Accessor for FunctionApproximator::meta_parameters_. More...
const ModelParameters *	getModelParameters (void) const
	Accessor for FunctionApproximator::model_parameters_. More...
void	setParameterVectorModifierPrivate (std::string modifier, bool new_value)
	Turn certain modifiers on or off, see Parameterizable::setParameterVectorModifier(). More...
Public Member Functions inherited from Parameterizable
virtual	~Parameterizable (void)
	Destructor.
virtual void	getParameterVectorSelectedNormalized (Eigen::VectorXd &values) const
	Get the normalized values of the selected parameters in one vector. More...
void	getParameterVectorSelectedMinMax (Eigen::VectorXd &min, Eigen::VectorXd &max) const
	Get the minimum and maximum of the selected parameters in one vector. More...
void	getParameterVectorAllMinMax (Eigen::VectorXd &min, Eigen::VectorXd &max) const
	Get the minimum and maximum values of the current parameter vector. More...
void	getParameterVectorSelectedRanges (Eigen::VectorXd &ranges) const
	Get the ranges of the selected parameters, i.e. More...
virtual void	setParameterVectorSelectedNormalized (const Eigen::VectorXd &values)
	Set all the values of the selected parameters with one vector of normalized values. More...
void	setSelectedParametersOne (std::string selected)
	Set the parameters that are currently selected. More...
void	setParameterVectorModifier (std::string modifier, bool new_value)
	Turn certain modifiers on or off. More...
void	setVectorLengthsPerDimension (const Eigen::VectorXi &lengths_per_dimension)
	The vector (VectorXd) with parameter values can be split into different parts (as vector<VectorXd>; this function specifices the length of each sub-vector. More...
Eigen::VectorXi	getVectorLengthsPerDimension (void) const
	Get the specified length of each vector in each dimension. More...
void	getParameterVectorSelected (std::vector< Eigen::VectorXd > &values, bool normalized=false) const
	Get the values of the selected parameters in one vector. More...
void	setParameterVectorSelected (const std::vector< Eigen::VectorXd > &values, bool normalized=false)
	Set all the values of the selected parameters with a vector of vectors. More...

Friends
class	boost::serialization::access
	Give boost serialization access to private members. More...

Additional Inherited Members
Static Public Member Functions inherited from FunctionApproximator
static void	generateInputsGrid (const Eigen::VectorXd &min, const Eigen::VectorXd &max, const Eigen::VectorXi &n_samples_per_dim, Eigen::MatrixXd &inputs_grid)
	Generate a input samples that lie on a grid (much like Matlab's meshgrid) For instance, if min = [2 6], and max = [3 8], and n_samples_per_dim = [3 5] then this function first makes linearly spaces samples along each dimension, e.g. More...
Protected Member Functions inherited from FunctionApproximator
void	setModelParameters (ModelParameters *model_parameters)
	Accessor for FunctionApproximator::model_parameters_. More...
	FunctionApproximator (void)
	Default constructor. More...

Detailed Description

RBFN (Radial Basis Function Network) function approximator.

Definition at line 43 of file FunctionApproximatorRBFN.hpp.

Constructor & Destructor Documentation

FunctionApproximatorRBFN	(	const MetaParametersRBFN *const	meta_parameters,
		const ModelParametersRBFN *const	model_parameters = NULL
	)

Initialize a function approximator with meta- and model-parameters.

Parameters

[in]	meta_parameters	The training algorithm meta-parameters
[in]	model_parameters	The parameters of the trained model. If this parameter is not passed, the function approximator is initialized as untrained. In this case, you must call FunctionApproximator::train() before being able to call FunctionApproximator::predict(). Either meta_parameters XOR model-parameters can passed as NULL, but not both.

Definition at line 51 of file FunctionApproximatorRBFN.cpp.

:
  FunctionApproximator(meta_parameters,model_parameters)
{
  if (model_parameters!=NULL)
    preallocateMemory(model_parameters->getNumberOfBasisFunctions());
}

Here is the call graph for this function:

FunctionApproximatorRBFN

(

const ModelParametersRBFN *const

model_parameters

)

Initialize a function approximator with model parameters.

Parameters

[in]

model_parameters

The parameters of the (previously) trained model.

Definition at line 59 of file FunctionApproximatorRBFN.cpp.

:
  FunctionApproximator(model_parameters)
{
  preallocateMemory(model_parameters->getNumberOfBasisFunctions());
}

Here is the call graph for this function:

Member Function Documentation

FunctionApproximator * clone ( void  ) const

virtual

Return a pointer to a deep copy of the FunctionApproximator object.

Returns

Pointer to a deep copy

Implements FunctionApproximator.

Definition at line 74 of file FunctionApproximatorRBFN.cpp.

                                                                {
  // All error checking and cloning is left to the FunctionApproximator constructor.
  return new FunctionApproximatorRBFN(
    dynamic_cast<const MetaParametersRBFN*>(getMetaParameters()),
    dynamic_cast<const ModelParametersRBFN*>(getModelParameters())
    );
};

Here is the call graph for this function:

void train ( const Eigen::Ref< const Eigen::MatrixXd > &  inputs, const Eigen::Ref< const Eigen::MatrixXd > &  targets  )

virtual

Train the function approximator with corresponding input and target examples.

Parameters

[in]	inputs	Input values of the training examples
[in]	targets	Target values of the training examples

Implements FunctionApproximator.

Definition at line 82 of file FunctionApproximatorRBFN.cpp.

{
  if (isTrained())  
  {
    cerr << "WARNING: You may not call FunctionApproximatorRBFN::train more than once. Doing nothing." << endl;
    cerr << "   (if you really want to retrain, call reTrain function instead)" << endl;
    return;
  }
  
  assert(inputs.rows() == targets.rows());
  assert(inputs.cols()==getExpectedInputDim());

  const MetaParametersRBFN* meta_parameters_rbfn = 
    dynamic_cast<const MetaParametersRBFN*>(getMetaParameters());
                      
  // Determine the centers and widths of the basis functions, given the range of the input data
  VectorXd min = inputs.colwise().minCoeff();
  VectorXd max = inputs.colwise().maxCoeff();
  MatrixXd centers, widths;
  meta_parameters_rbfn->getCentersAndWidths(min,max,centers,widths);

  // Get the activations of the basis functions 
  bool normalized_basis_functions=false;  
  bool asymmetric_kernels=false;  
  int n_samples = inputs.rows();
  int n_kernels = centers.rows(); 
  MatrixXd activations(n_samples,n_kernels);
  BasisFunction::Gaussian::activations(centers,widths,inputs,activations,
    normalized_basis_functions,asymmetric_kernels);

  // Least squares, with activations as design matrix
  bool use_offset=false;
  double regularization = meta_parameters_rbfn->regularization();
  VectorXd weights = leastSquares(activations,targets,use_offset,regularization);

  setModelParameters(new ModelParametersRBFN(centers,widths,weights));
  
  preallocateMemory(n_kernels);

}

Here is the call graph for this function:

void predict ( const Eigen::Ref< const Eigen::MatrixXd > &  inputs, Eigen::MatrixXd &  outputs  )

virtual

Query the function approximator to make a prediction.

Parameters

[in]	inputs	Input values of the query
[out]	outputs	Predicted output values

Remarks

This method should be const. But third party functions which is called in this function have not always been implemented as const (Examples: LWPRObject::predict or RRRFF::predict ). Therefore, this function cannot be const.

Implements FunctionApproximator.

Definition at line 123 of file FunctionApproximatorRBFN.cpp.

{
  if (!isTrained())  
  {
    cerr << "WARNING: You may not call FunctionApproximatorLWPR::predict if you have not trained yet. Doing nothing." << endl;
    return;
  }

  const ModelParametersRBFN* model_parameters_rbfn = static_cast<const ModelParametersRBFN*>(getModelParameters());
  
  model_parameters_rbfn->weights(weights_prealloc_);
  
  int n_basis_functions = model_parameters_rbfn->getNumberOfBasisFunctions();
  
  bool only_one_sample = (inputs.rows()==1);
  if (only_one_sample)
  {
    ENTERING_REAL_TIME_CRITICAL_CODE
    
    // Get the basis function activations  
    model_parameters_rbfn->kernelActivations(inputs,activations_one_prealloc_);
      
    // Weight the basis function activations  
    for (int b=0; b<n_basis_functions; b++)
      activations_one_prealloc_.col(b).array() *= weights_prealloc_(b);
  
    // Sum over weighed basis functions
    outputs = activations_one_prealloc_.rowwise().sum();
    
    EXITING_REAL_TIME_CRITICAL_CODE
  }
  else 
  {
    int n_time_steps = inputs.rows();

    // The next two lines may not be real-time, as they may allocate memory.
    // (if the size are already correct, it will be realtime)
    activations_prealloc_.resize(n_time_steps,n_basis_functions);
    outputs.resize(n_time_steps,getExpectedOutputDim());
    
    // Get the basis function activations  
    model_parameters_rbfn->kernelActivations(inputs,activations_prealloc_);
      
    // Weight the basis function activations  
    for (int b=0; b<n_basis_functions; b++)
      activations_prealloc_.col(b).array() *= weights_prealloc_(b);
  
    // Sum over weighed basis functions
    outputs = activations_prealloc_.rowwise().sum();
  }
    
}

Here is the call graph for this function:

void preallocateMemory

(

int

n_basis_functions

)

Preallocate memory to make certain functions real-time.

Parameters

[in]

n_basis_functions

Number of basis functions in the RBFN.

Definition at line 66 of file FunctionApproximatorRBFN.cpp.

{
  weights_prealloc_ = VectorXd(n_basis_functions);
  activations_one_prealloc_ = MatrixXd(1,n_basis_functions);
  activations_prealloc_ = MatrixXd(1,n_basis_functions);
}

std::string getName ( void  ) const

inlinevirtual

Get the name of this function approximator.

Returns

Name of this function approximator

Implements FunctionApproximator.

Definition at line 73 of file FunctionApproximatorRBFN.hpp.

                                             {
    return std::string("RBFN");  
  };

Here is the call graph for this function:

bool saveGridData ( const Eigen::VectorXd &  min, const Eigen::VectorXd &  max, const Eigen::VectorXi &  n_samples_per_dim, std::string  directory, bool  overwrite = false  ) const

virtual

Generate a grid of inputs, and output the response of the basis functions and line segments for these inputs.

This function is not pure virtual, because this might not make sense for every model parameters class.

Parameters

[in]	min	Minimum values for the grid (one for each dimension)
[in]	max	Maximum values for the grid (one for each dimension)
[in]	n_samples_per_dim	Number of samples in the grid along each dimension
[in]	directory	Directory to which to save the results to.
[in]	overwrite	Whether to overwrite existing files. true=do overwrite, false=don't overwrite and give a warning.

Returns

Whether saving the data was successful.

Reimplemented from FunctionApproximator.

Definition at line 176 of file FunctionApproximatorRBFN.cpp.

{
  if (save_directory.empty())
    return true;
  
  MatrixXd inputs_grid;
  FunctionApproximator::generateInputsGrid(min, max, n_samples_per_dim, inputs_grid);
      
  const ModelParametersRBFN* model_parameters_rbfn = static_cast<const ModelParametersRBFN*>(getModelParameters());
  
  MatrixXd activations_grid;
  model_parameters_rbfn->kernelActivations(inputs_grid, activations_grid);
  
  saveMatrix(save_directory,"n_samples_per_dim.txt",n_samples_per_dim,overwrite);
  saveMatrix(save_directory,"inputs_grid.txt",inputs_grid,overwrite);
  saveMatrix(save_directory,"activations_grid.txt",activations_grid,overwrite);

  // Weight the basis function activations  
  VectorXd weights = model_parameters_rbfn->weights();
  for (int b=0; b<activations_grid.cols(); b++)
    activations_grid.col(b).array() *= weights(b);
  saveMatrix(save_directory,"activations_weighted_grid.txt",activations_grid,overwrite);
  
  // Sum over weighed basis functions
  MatrixXd predictions_grid = activations_grid.rowwise().sum();
  saveMatrix(save_directory,"predictions_grid.txt",predictions_grid,overwrite);
  
  return true;
  
}

Here is the call graph for this function:

Friends And Related Function Documentation

friend class boost::serialization::access

friend

Give boost serialization access to private members.

Definition at line 98 of file FunctionApproximatorRBFN.hpp.

---

The documentation for this class was generated from the following files:

• FunctionApproximatorRBFN.hpp
• FunctionApproximatorRBFN.cpp