RFDiag Class Reference

The RFDiag class represents a receptive field with a diagonal matrix. More...

#include <rf_diag.hpp>

Inheritance diagram for RFDiag:

Inheritance graph
[legend]
Collaboration diagram for RFDiag:

Collaboration graph
[legend]
List of all members.

Public Member Functions

 RFDiag ()
 Default constructor for the RF.
 RFDiag (Types::Input c)
 A constructor taking only a center point for the RF.
 RFDiag (Types::Input c, const boost::program_options::variables_map &vm)
 A constructor taking a center point and a variables_map.
 RFDiag (const RFDiag &, Types::Input c, const boost::program_options::variables_map &vm)
 a copy constructor, sort of.
virtual Types::RValue getActivation (Types::Input x) const
 Return the activation for a test point.
virtual Types::RValue learn (Types::Input x, Types::Output y, Types::RValueP w)
 Update the receptive field and have local model update itself.
Types::Input getAlpha () const
 stupid get functions for copy constructor.
Types::Input getD () const
Types::Input getM () const
Types::Input getCenter () const

Private Attributes

Types::vector dwdM
Types::vector dJ2dM
Types::vector dJdM
Types::vector dJ2_2dM_2
Types::vector aux
Types::vector dw_2dM_2
Types::vector dJ_2dM_2
Meta Learning Sufficient Statistics
This group of variables control the behavior of the algorithm.

All paramters are defined in the paper (see PAPER).

Types::vector alpha
 Learning rate for the distance matrix updates.
Types::vector meta_h
 Meta Update Params.
Types::vector b
RF Descriptors
These variables describe the receptive field.

They include the center, distance matrix, and local regression model

Types::vector D
 the distance matrix of the RF
Types::vector M
 The Cholesky decomposition of D.

Friends

std::ostream & operator<< (std::ostream &out, const RFDiag &rf)
 print the center and distance matrix on a line

Detailed Description

The RFDiag class represents a receptive field with a diagonal matrix.

This class handles updates of the distance matrix for the receptive field and hold the local model associated with the field.

Definition at line 40 of file rf_diag.hpp.


Constructor & Destructor Documentation

RFDiag::RFDiag ( Types::Input  c  ) 

A constructor taking only a center point for the RF.

Parameters:
c An input vector corresponding to the center of the RF.

Definition at line 38 of file rf_diag.cpp.

References alpha, RF::alpha_init, b, D, RF::d_def, M, and meta_h.

RFDiag::RFDiag ( const RFDiag ,
Types::Input  c,
const boost::program_options::variables_map &  vm 
)

a copy constructor, sort of.

Just copy vm, alpha, D, M.

Definition at line 83 of file rf_diag.cpp.

References alpha, RF::alpha_init, b, D, getAlpha(), getD(), getM(), M, and meta_h.

Here is the call graph for this function:


Member Function Documentation

virtual Types::RValue RFDiag::getActivation ( Types::Input  x  )  const [inline, virtual]

Return the activation for a test point.

This currently uses ublas vectro expressions, but should be changed to use blas calls once everything is working correctly.

Parameters:
x the input vector

Implements RF.

Definition at line 116 of file rf_diag.hpp.

References RF::center, and M.

Types::RValue RFDiag::learn ( Types::Input  x,
Types::Output  y,
Types::RValueP  w 
) [virtual]

Update the receptive field and have local model update itself.

Update the distance matrix as in Equation 3.6 and Table 4 $ M^{n+1} = M^n - \alpha \frac{\partial J}{\partial M} $

Parameters:
x the input vector
y the actual output

Implements RF.

Definition at line 105 of file rf_diag.cpp.

References RF::a_E, RF::a_G, RF::a_H, RF::a_pk, alpha, aux, b, RF::center, Model::check_derivatives(), D, dJ2_2dM_2, dJ2dM, dJ_2dM_2, dJdM, dw_2dM_2, dwdM, RF::e_2, RF::gamma, Model::get_a_zz(), Model::get_e(), Model::get_e_cv(), Model::get_z(), RF::H_temp, RF::lambda, Model::learn(), RF::localModel, M, RF::meta, RF::meta_alpha, meta_h, RF::q, RF::temp2, Model::updateNumProjections(), RF::W, and RF::z.

Here is the call graph for this function:


Member Data Documentation

Types::vector RFDiag::M [private]

The Cholesky decomposition of D.

This is used for updating the distance matrix.

Definition at line 77 of file rf_diag.hpp.

Referenced by getActivation(), getM(), learn(), and RFDiag().


The documentation for this class was generated from the following files:
Generated on Fri Jul 27 00:24:02 2007 for LWPR by  doxygen 1.5.1