PLearn 0.1
Public Member Functions | Static Public Member Functions | Static Public Attributes | Protected Member Functions | Static Protected Member Functions | Protected Attributes | Private Types
PLearn::NegCrossEntropySigmoidVariable Class Reference

#include <NegCrossEntropySigmoidVariable.h>

Inheritance diagram for PLearn::NegCrossEntropySigmoidVariable:
Inheritance graph
[legend]
Collaboration diagram for PLearn::NegCrossEntropySigmoidVariable:
Collaboration graph
[legend]

List of all members.

Public Member Functions

 NegCrossEntropySigmoidVariable ()
 Default constructor for persistence.
 NegCrossEntropySigmoidVariable (Variable *netout, Variable *target, real regularizer_=0, bool ignore_missing_=false)
virtual string classname () const
virtual OptionListgetOptionList () const
virtual OptionMapgetOptionMap () const
virtual RemoteMethodMapgetRemoteMethodMap () const
virtual
NegCrossEntropySigmoidVariable
deepCopy (CopiesMap &copies) const
virtual void build ()
 Post-constructor.
virtual void recomputeSize (int &l, int &w) const
 Recomputes the length l and width w that this variable should have, according to its parent variables.
virtual void fprop ()
 compute output given input
virtual void bprop ()
void setRegularizer (real r)
 Deprecated.

Static Public Member Functions

static string _classname_ ()
 NegCrossEntropySigmoidVariable.
static OptionList_getOptionList_ ()
static RemoteMethodMap_getRemoteMethodMap_ ()
static Object_new_instance_for_typemap_ ()
static bool _isa_ (const Object *o)
static void _static_initialize_ ()
static const PPathdeclaringFile ()

Static Public Attributes

static StaticInitializer _static_initializer_

Protected Member Functions

void build_ ()
 This does the actual building.

Static Protected Member Functions

static void declareOptions (OptionList &ol)
 Declare options (data fields) for the class.

Protected Attributes

real regularizer
 If > 0, will modify the cost function to: (1-t)(r*log(o)+(1-r)*log(1-o)) + t*(r*log(1-o)+(1-r)*log(o)) (t = target, o = output, r = regularizer = a small value)
bool ignore_missing
 Indication that missing targets should be ignored.

Private Types

typedef BinaryVariable inherited

Detailed Description

Definition at line 52 of file NegCrossEntropySigmoidVariable.h.


Member Typedef Documentation

Reimplemented from PLearn::BinaryVariable.

Definition at line 54 of file NegCrossEntropySigmoidVariable.h.


Constructor & Destructor Documentation

PLearn::NegCrossEntropySigmoidVariable::NegCrossEntropySigmoidVariable ( ) [inline]

Default constructor for persistence.

Definition at line 66 of file NegCrossEntropySigmoidVariable.h.

        : regularizer(0.0), ignore_missing(false) {}
PLearn::NegCrossEntropySigmoidVariable::NegCrossEntropySigmoidVariable ( Variable netout,
Variable target,
real  regularizer_ = 0,
bool  ignore_missing_ = false 
)

Definition at line 60 of file NegCrossEntropySigmoidVariable.cc.

References build_().

    : inherited(netout,target,1,1),regularizer(regularizer_), ignore_missing(ignore_missing_)
{
    build_();
}

Here is the call graph for this function:


Member Function Documentation

string PLearn::NegCrossEntropySigmoidVariable::_classname_ ( ) [static]
OptionList & PLearn::NegCrossEntropySigmoidVariable::_getOptionList_ ( ) [static]

Reimplemented from PLearn::BinaryVariable.

Definition at line 55 of file NegCrossEntropySigmoidVariable.cc.

RemoteMethodMap & PLearn::NegCrossEntropySigmoidVariable::_getRemoteMethodMap_ ( ) [static]

Reimplemented from PLearn::BinaryVariable.

Definition at line 55 of file NegCrossEntropySigmoidVariable.cc.

bool PLearn::NegCrossEntropySigmoidVariable::_isa_ ( const Object o) [static]

Reimplemented from PLearn::BinaryVariable.

Definition at line 55 of file NegCrossEntropySigmoidVariable.cc.

Object * PLearn::NegCrossEntropySigmoidVariable::_new_instance_for_typemap_ ( ) [static]

Reimplemented from PLearn::Object.

Definition at line 55 of file NegCrossEntropySigmoidVariable.cc.

StaticInitializer NegCrossEntropySigmoidVariable::_static_initializer_ & PLearn::NegCrossEntropySigmoidVariable::_static_initialize_ ( ) [static]

Reimplemented from PLearn::BinaryVariable.

Definition at line 55 of file NegCrossEntropySigmoidVariable.cc.

void PLearn::NegCrossEntropySigmoidVariable::bprop ( ) [virtual]

Implements PLearn::Variable.

Definition at line 146 of file NegCrossEntropySigmoidVariable.cc.

References PLearn::fast_exact_is_equal(), PLearn::Variable::gradientdata, i, ignore_missing, PLearn::BinaryVariable::input1, PLearn::BinaryVariable::input2, PLearn::is_missing(), PLERROR, regularizer, and PLearn::sigmoid().

{
    real gr = *gradientdata;
    for (int i=0; i<input1->size(); i++)
    {
        real output = sigmoid(input1->valuedata[i]);
        real target = input2->valuedata[i];
        if(!ignore_missing || !is_missing(target))
        {

            if (fast_exact_is_equal(regularizer, 0)) {
                // Standard cross entropy.
                input1->gradientdata[i] += gr*(output - target);
            } else {
                // Regularized cross entropy.
                if (fast_exact_is_equal(target, 0.0)) {
                    input1->gradientdata[i] += gr*((1-regularizer) * output - regularizer * (1-output));
                } else if (fast_exact_is_equal(target, 1.0)) {
                    input1->gradientdata[i] += gr*(regularizer * output - (1-regularizer) * (1-output));
                } else {
                    PLERROR("NegCrossEntropySigmoidVariable::bprop: target is neither 0 nor 1");
                }
            }
        }
    }
}

Here is the call graph for this function:

void PLearn::NegCrossEntropySigmoidVariable::build ( ) [virtual]

Post-constructor.

The normal implementation should call simply inherited::build(), then this class's build_(). This method should be callable again at later times, after modifying some option fields to change the "architecture" of the object.

Reimplemented from PLearn::BinaryVariable.

Definition at line 85 of file NegCrossEntropySigmoidVariable.cc.

References PLearn::BinaryVariable::build(), and build_().

Here is the call graph for this function:

void PLearn::NegCrossEntropySigmoidVariable::build_ ( ) [protected]

This does the actual building.

Reimplemented from PLearn::BinaryVariable.

Definition at line 91 of file NegCrossEntropySigmoidVariable.cc.

References PLearn::BinaryVariable::input1, PLearn::BinaryVariable::input2, and PLERROR.

Referenced by build(), and NegCrossEntropySigmoidVariable().

{
    if (input1 && input2) {
        // input1 and input2 are (respectively) netout and target from constructor
        if(input1->size() != input2->size())
            PLERROR("In NegCrossEntropySigmoidVariable: netout and target must have the same size");
    }
}

Here is the caller graph for this function:

string PLearn::NegCrossEntropySigmoidVariable::classname ( ) const [virtual]

Reimplemented from PLearn::Object.

Definition at line 55 of file NegCrossEntropySigmoidVariable.cc.

void PLearn::NegCrossEntropySigmoidVariable::declareOptions ( OptionList ol) [static, protected]

Declare options (data fields) for the class.

Redefine this in subclasses: call declareOption(...) for each option, and then call inherited::declareOptions(options). Please call the inherited method AT THE END to get the options listed in a consistent order (from most recently defined to least recently defined).

  static void MyDerivedClass::declareOptions(OptionList& ol)
  {
      declareOption(ol, "inputsize", &MyObject::inputsize_,
                    OptionBase::buildoption,
                    "The size of the input; it must be provided");
      declareOption(ol, "weights", &MyObject::weights,
                    OptionBase::learntoption,
                    "The learned model weights");
      inherited::declareOptions(ol);
  }
Parameters:
olList of options that is progressively being constructed for the current class.

Reimplemented from PLearn::BinaryVariable.

Definition at line 69 of file NegCrossEntropySigmoidVariable.cc.

References PLearn::OptionBase::buildoption, PLearn::declareOption(), PLearn::BinaryVariable::declareOptions(), ignore_missing, and regularizer.

{
    declareOption(ol, "regularizer", &NegCrossEntropySigmoidVariable::regularizer, OptionBase::buildoption, 
                  "If > 0, will modify the cost function to: \n"
                  "(1-t)(r*log(o)+(1-r)*log(1-o)) + t*(r*log(1-o)+(1-r)*log(o)) \n"
                  "(t = target, o = output, r = regularizer = a small value)\n");

    declareOption(ol, "ignore_missing", &NegCrossEntropySigmoidVariable::ignore_missing, OptionBase::buildoption, 
                  "Indication that missing targets should be ignored");

    inherited::declareOptions(ol);
}

Here is the call graph for this function:

static const PPath& PLearn::NegCrossEntropySigmoidVariable::declaringFile ( ) [inline, static]

Reimplemented from PLearn::BinaryVariable.

Definition at line 71 of file NegCrossEntropySigmoidVariable.h.

:
    static void declareOptions(OptionList & ol);
NegCrossEntropySigmoidVariable * PLearn::NegCrossEntropySigmoidVariable::deepCopy ( CopiesMap copies) const [virtual]

Reimplemented from PLearn::BinaryVariable.

Definition at line 55 of file NegCrossEntropySigmoidVariable.cc.

void PLearn::NegCrossEntropySigmoidVariable::fprop ( ) [virtual]

compute output given input

Implements PLearn::Variable.

Definition at line 109 of file NegCrossEntropySigmoidVariable.cc.

References PLearn::fast_exact_is_equal(), i, ignore_missing, PLearn::BinaryVariable::input1, PLearn::BinaryVariable::input2, PLearn::is_missing(), pl_log, PLWARNING, regularizer, PLearn::sigmoid(), and PLearn::Variable::valuedata.

{
    real cost = 0.0;
    for (int i=0; i<input1->size(); i++)
    {
        real output = sigmoid(input1->valuedata[i]);
        real target = input2->valuedata[i];
        if(!ignore_missing || !is_missing(target))
        {
            if (fast_exact_is_equal(output,0.0)) {
                if (fast_exact_is_equal(target, 1.0)) {
                    PLWARNING("NegCrossEntropySigmoidVariable::fprop: model output is 0 and target is 1, cost should be infinite !");
                    cost += -1e9;
                } // If target == 0.0 do nothing, cost is 0.
            } else if (fast_exact_is_equal(output, 1.0)) {
                if (fast_exact_is_equal(target, 0.0)) {
                    PLWARNING("NegCrossEntropySigmoidVariable::fprop: model output is 1 and target is 0, cost should be infinite !");
                    cost += -1e9;
                } // If target == 1.0 do nothing, cost is 0.
            } else {
                if (fast_exact_is_equal(regularizer, 0)) {
                    // Standard cross entropy.
                    cost += target*pl_log(output) + (1.0-target)*pl_log(1.0-output);
                } else {
                    // Regularized cross entropy.
                    cost += target*((1 - regularizer) * pl_log(output) + regularizer * pl_log(1.0 - output)) +
                        (1.0-target)*((1 - regularizer) * pl_log(1.0-output) + regularizer * pl_log(output));
                }
            }
        }
    }
    valuedata[0] = -cost;
}

Here is the call graph for this function:

OptionList & PLearn::NegCrossEntropySigmoidVariable::getOptionList ( ) const [virtual]

Reimplemented from PLearn::Object.

Definition at line 55 of file NegCrossEntropySigmoidVariable.cc.

OptionMap & PLearn::NegCrossEntropySigmoidVariable::getOptionMap ( ) const [virtual]

Reimplemented from PLearn::Object.

Definition at line 55 of file NegCrossEntropySigmoidVariable.cc.

RemoteMethodMap & PLearn::NegCrossEntropySigmoidVariable::getRemoteMethodMap ( ) const [virtual]

Reimplemented from PLearn::Object.

Definition at line 55 of file NegCrossEntropySigmoidVariable.cc.

void PLearn::NegCrossEntropySigmoidVariable::recomputeSize ( int l,
int w 
) const [virtual]

Recomputes the length l and width w that this variable should have, according to its parent variables.

This is used for ex. by sizeprop() The default version stupidly returns the current dimensions, so make sure to overload it in subclasses if this is not appropriate.

Reimplemented from PLearn::Variable.

Definition at line 103 of file NegCrossEntropySigmoidVariable.cc.

{ l=1, w=1; }
void PLearn::NegCrossEntropySigmoidVariable::setRegularizer ( real  r)

Deprecated.

Definition at line 176 of file NegCrossEntropySigmoidVariable.cc.

References PLWARNING, and regularizer.

{
    PLWARNING("NegCrossEntropySigmoidVariable::setRegularizer() has been deprecated, use the setOption() method instead");
    this->regularizer = r;
}

Member Data Documentation

Reimplemented from PLearn::BinaryVariable.

Definition at line 71 of file NegCrossEntropySigmoidVariable.h.

Indication that missing targets should be ignored.

Definition at line 62 of file NegCrossEntropySigmoidVariable.h.

Referenced by bprop(), declareOptions(), and fprop().

If > 0, will modify the cost function to: (1-t)(r*log(o)+(1-r)*log(1-o)) + t*(r*log(1-o)+(1-r)*log(o)) (t = target, o = output, r = regularizer = a small value)

Definition at line 60 of file NegCrossEntropySigmoidVariable.h.

Referenced by bprop(), declareOptions(), fprop(), and setRegularizer().


The documentation for this class was generated from the following files:
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Defines