Dataset Class Reference

Reads a full dataset; limited to those that fit in memory. More...

#include <Dataset.hpp>

List of all members.

Public Member Functions

 Dataset ()
 count of rows in each class 1,2... ninc[0] unused
 Dataset (string fname, double enc0, double enc1)
 Read from a file.
const vector< double > & row (size_t i) const
 Look at an indexed row of data.
size_t getNClasses () const
 Return number of classes.
size_t getNRows () const
 Return number of rows / instances.
size_t getTargetClass (size_t ind) const
 Return target class 1..N.
size_t getNRowsInClass (size_t c) const
 Return the number of rows that belong to a given class.
const vector< double > & prototype (size_t row)
 Return a binary vector encoding the target class; the real values for the encoding must be given as a constructor parameter; default is -1 for non-class elements and +1 for class elements.
void toStream (ostream &ost) const
 Dump data to a stream.

Protected Attributes

vector< vector< double > > invec
vector< size_t > targetc
 input vectors
vector< vector< double > > prototypes
 target classes 1,2,...
size_t nclasses
 target prototypes
vector< size_t > ninc
 number of classes; discovered upon load

Detailed Description

Reads a full dataset; limited to those that fit in memory.

A pointer to this must be shared among all users. As of yet, works only for classification datasets, where the last value on each row is the integral class label and all other values are real-valued features.


Constructor & Destructor Documentation

Dataset::Dataset ( string  fname,
double  enc0,
double  enc1 
)

Read from a file.

TODO: differentiate classification / approx?


Member Function Documentation

size_t Dataset::getNClasses (  )  const

Return number of classes.

TODO: classification vs. approx?

size_t Dataset::getTargetClass ( size_t  ind  )  const

Return target class 1..N.

NOTE: class indices are base 1 while data indices are base 0!

const vector< double > & Dataset::row ( size_t  i  )  const

Look at an indexed row of data.

TODO: Likely to replace this with iterators that can be generated for splits. Or, rather, objects of type DataSplit.

void Dataset::toStream ( ostream &  ost  )  const

Dump data to a stream.


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

Generated on 1 Sep 2016 for Momulper - Multiobjective MLP for Classification by  doxygen 1.6.1