TMB Documentation  v1.9.10
Public Member Functions | List of all members
tmbutils::interpol2D< Type > Struct Template Reference

Get a smooth representation of a data matrix. More...

#include <interpol.hpp>

Public Member Functions

 interpol2D (matrix< Type > data, vector< Type > x_range, vector< Type > y_range, interpol2D_config< Type > cfg=interpol2D_config< Type >())
 Construct interpolation object. More...
 
double operator() (double x, double y, int nx=0, int ny=0)
 Scalar evaluation with double types. More...
 
ad operator() (ad x, ad y, int nx=0, int ny=0)
 Scalar evaluation with ad types. More...
 
- Public Member Functions inherited from TMBad::global::Operator< ninput, noutput >
void dependencies_updating (Args<> &args, Dependencies &dep) const
 Default implementation of OperatorPure::dependencies_updating()
 
void * operator_data ()
 Return operator specific dynamic information (optional)
 
OperatorPureother_fuse (OperatorPure *self, OperatorPure *other)
 How to fuse this operator (self) with another (other)
 
void print (print_config cfg)
 Print this operator (optional)
 

Additional Inherited Members

- Static Public Attributes inherited from TMBad::global::DynamicOperator< 2, 1 >
static const bool dynamic
 
static const int max_fuse_depth
 
- Static Public Attributes inherited from TMBad::global::Operator< ninput, noutput >
static const bool add_forward_replay_copy
 Should this operator replay it self by invoking the copy CTOR ?
 
static const bool add_static_identifier
 Should this operator have a static identifier ?
 
static const bool allow_remap
 Is it safe to remap the inputs of this operator? More...
 
static const int dependent_variable
 Is output of this operator a dependent variable ?
 
static const bool dynamic
 Does this operator require dynamic allocation ?
 
static const bool elimination_protected
 Protect this operator from elimination by the tape optimizer ?
 
static const bool have_dependencies
 Have dependencies member been defined ?
 
static const bool have_eval
 Does this class have an eval member from which to define forward ?
 
static const bool have_forward_incr_reverse_decr
 Have forward_incr and reverse_incr members been defined ?
 
static const bool have_forward_mark_reverse_mark
 Have forward_mark and reverse_mark members been defined ?
 
static const bool have_forward_reverse
 Have forward and reverse members been defined ?
 
static const bool have_increment_decrement
 Have increment and decrement members been defined ?
 
static const bool have_input_size_output_size
 Have input_size and output_size members been defined ?
 
static const bool implicit_dependencies
 Does this operator have implicit dependencies? More...
 
static const int independent_variable
 Is output of this operator an independent variable ?
 
static const bool is_constant
 Is this a constant operator ?
 
static const bool is_linear
 Is this a linear operator ?
 
static const int max_fuse_depth
 How many times can this operator be doubled ?
 
static const int ninput
 Number of operator inputs.
 
static const int noutput
 Number of operator outputs.
 
static const bool smart_pointer
 Is this operator a 'smart pointer' (with reference counting) ?
 
static const bool updating
 This operator may update existing variables ? More...
 

Detailed Description

template<class Type>
struct tmbutils::interpol2D< Type >

Get a smooth representation of a data matrix.

This 2D interpolation can be put on the AD tape. Interpolation data is shared among instances so you can do interpolation many times without duplicating the data.

In addition the operator is thread safe, whether used with manual parallelization (parallel_accumulator etc) or automatic parallization (TMBad::autopar).

The actual interpolation is based on a kernel smoothing approach: To obtain the interpolated value at coordinate (x,y), the coordinate is first transformed to continous matrix index coordinates using the supplied x_range and y_range. The distance to each matrix entry d(i,j) is then calculated, and a corresponding 'weight' is given as a function (kernel) of the distance divided by the 'interpolation radius' (R), i.e. W(i,j)=kernel(d(i,j)/R). The final interpolated value becomes sum(W*F)/sum(W) where F is the data matrix.

We note that the interpolation radius (interpol2D_config::R) is relative to matrix index coordinates (and is thus not related to the x_range and y_range).

Note
An error is triggered if some data are not constant
Examples:
TMBad/interpol.cpp.

Definition at line 145 of file interpol.hpp.

Constructor & Destructor Documentation

§ interpol2D()

template<class Type >
tmbutils::interpol2D< Type >::interpol2D ( matrix< Type >  data,
vector< Type >  x_range,
vector< Type >  y_range,
interpol2D_config< Type >  cfg = interpol2D_config<Type>() 
)
inline

Construct interpolation object.

Parameters
dataMatrix of data values.
x_rangeVector of length 2 corresponding to first and last data row.
y_rangeVector of length 2 corresponding to first and last data column.
cfgControl parameters

Definition at line 165 of file interpol.hpp.

Member Function Documentation

§ operator()() [1/2]

template<class Type >
double tmbutils::interpol2D< Type >::operator() ( double  x,
double  y,
int  nx = 0,
int  ny = 0 
)
inline

Scalar evaluation with double types.

Parameters
xx-coordinate
yy-coordinate
nxx-derivative order
nyy-derivative order

Definition at line 193 of file interpol.hpp.

§ operator()() [2/2]

template<class Type >
ad tmbutils::interpol2D< Type >::operator() ( ad  x,
ad  y,
int  nx = 0,
int  ny = 0 
)
inline

Scalar evaluation with ad types.

Parameters
xx-coordinate
yy-coordinate
nxx-derivative order
nyy-derivative order

Definition at line 202 of file interpol.hpp.


The documentation for this struct was generated from the following file:
License: GPL v2