CCL Library  1.0
Data Fields
SOLVE_NONLIN_WS Struct Reference

This structure defines the workspace variables for solving the non-linear LM optimization. More...

#include <ccl_learn_ncl.h>

Data Fields

int dim_x
 The dimensionality of the state variable.
 
int dim_n
 The number of data points.
 
int dim_b
 The number of basis functions.
 
int dim_y
 The dimensionality of the action space.
 
int r_ok
 check if the objective function is belowed the tolerence
 
int d_ok
 check if the model parameters are belowed the tolerence
 
double * xc
 A copy of the initial model parameters.
 
double * x
 The flattened and updated model parameters.
 
double * xf
 The finalised model parameters.
 
double * epsx
 The tolerence of the model paramters.
 
double epsf
 The tolerence fo the objective functions.
 
double * r
 The returned value of the residual error.
 
double * J
 The jacobian at x.
 
double S
 The sqaure root if x is taken.
 
double * A
 The dot product of J_T and J.
 
double * v
 The dot product of J_T and r.
 
double * D
 Automatic scaling.
 
double Rlo
 The lower bound of R.
 
double Rhi
 The upper bound of R.
 
double l
 The adaptive learning rate.
 
double lc
 Handling situation when learning rate happends to be 0.
 
double * d
 The parameter improvement gradient.
 
int iter
 The iteration number.
 
double * xd
 The next x.
 
double * rd
 The residual error at xd.
 
double Sd
 The squared error if xd is taken.
 
double dS
 The denomitor of the sqaured error is xd is taken.
 
double R
 The reduction if xd is taken.
 
double nu
 The coefficient of changing learning rate.
 
double * d_T
 The matrix transpose of d.
 
double * J_T
 The matrix transpose of J.
 
double * tmp
 The temporal variable.
 
double * rd_T
 The matrix transpose of rd.
 
gsl_matrix * D_pinv
 The peudo-inverse of D.
 
gsl_vector * A_d
 The A matrix at xd.
 
gsl_matrix * A_inv
 The peudo-inverse of A.
 
gsl_vector * A_inv_diag
 The vector of diagonal elements of A matrix.
 
gsl_vector * r_T
 The vector transpose of r.
 

Detailed Description

This structure defines the workspace variables for solving the non-linear LM optimization.

This structure containts the memory of the workspace variables for sovling the non-linear LM optimization problem. It should always initialise with ccl_learn_model_ws_alloc and destroy with ccl_learn_model_ws_free.


The documentation for this struct was generated from the following file: