datapoint< FEATURE_SIZE > Struct Template Reference

List of all members.

---

Detailed Description

template<int FEATURE_SIZE>
struct datapoint< FEATURE_SIZE >

This structure represents a datapoint.

A datapoint is a group of pixels with ternary values (much brighter than the centre, much darker than the centre or similar to the centre pixel). In addition to the feature descriptor, the class and number of instances is also stored.

The maximum feature vector size is determined by the template parameter. This allows the ternary vector to be stored in a bitset. This keeps the struct a fixed size and removes the need for dynamic allocation.

Definition at line 146 of file learn_fast_tree.cc.

Public Member Functions
	datapoint (const string &s, unsigned long c, bool is)
	datapoint ()
Ternary	get_trit (unsigned int tnum) const
Public Attributes
unsigned long	count
bool	is_a_corner
Static Public Attributes
static const unsigned int	max_size = FEATURE_SIZE
Private Member Functions
void	pack_trits (const string &unpacked)
void	set_trit (unsigned int tnum, Ternary val)
Private Attributes
bitset< max_size *2 >	tests

---

Constructor & Destructor Documentation

template<int FEATURE_SIZE>

datapoint< FEATURE_SIZE >::datapoint	(	const string &	s,
		unsigned long	c,
		bool	is
	)			[inline]

Construct a datapoint.

Parameters:

	s	The feature vector in string form
	c	The number of instances
	is	The class

Definition at line 152 of file learn_fast_tree.cc.

References datapoint< FEATURE_SIZE >::pack_trits().

    :count(c),is_a_corner(is)
    {
        pack_trits(s);
    }

template<int FEATURE_SIZE>

datapoint< FEATURE_SIZE >::datapoint

(

)

[inline]

Default constructor allows for storage in a std::vector.

Definition at line 160 of file learn_fast_tree.cc.

    {}

---

Member Function Documentation

template<int FEATURE_SIZE>

Ternary datapoint< FEATURE_SIZE >::get_trit

(

unsigned int

tnum

)

const [inline]

Extract a trit (ternary bit) from the feture vector.

Parameters:

tnum

Number of the bit to extract

Returns:

The trit.

Definition at line 172 of file learn_fast_tree.cc.

References Brighter, Darker, datapoint< FEATURE_SIZE >::max_size, Similar, and datapoint< FEATURE_SIZE >::tests.

    {
        assert(tnum < size);
        if(tests[tnum] == 1)
            return Brighter;
        else if(tests[tnum + max_size] == 1)
            return Darker;
        else
            return Similar;
    }

template<int FEATURE_SIZE>

void datapoint< FEATURE_SIZE >::pack_trits

(

const string &

unpacked

)

[inline, private]

This code reads a stringified representation of the feature vector and converts it in to the internal representation.

The string represents one feature per character, using "b", "d" and "s".

Parameters:

unpacked

String to parse.

Definition at line 203 of file learn_fast_tree.cc.

References Brighter, Darker, fatal, datapoint< FEATURE_SIZE >::set_trit(), Similar, and datapoint< FEATURE_SIZE >::tests.

Referenced by datapoint< FEATURE_SIZE >::datapoint().

        {
            tests = 0;
            for(unsigned int i=0;i < unpacked.size(); i++)
            {
                if(unpacked[i] == 'b')
                    set_trit(i, Brighter);
                else if(unpacked[i] == 'd')
                    set_trit(i, Darker);
                else if(unpacked[i] == 's')
                    set_trit(i, Similar);
                else
                    fatal(2, "Bad char while packing datapoint: %s", unpacked);
            }
        }

template<int FEATURE_SIZE>

void datapoint< FEATURE_SIZE >::set_trit	(	unsigned int	tnum,
		Ternary	val
	)			[inline, private]

Set a ternary digit.

Parameters:

	tnum	Digit to set
	val	Value to set it to.

Definition at line 222 of file learn_fast_tree.cc.

References Brighter, Darker, datapoint< FEATURE_SIZE >::max_size, Similar, and datapoint< FEATURE_SIZE >::tests.

Referenced by datapoint< FEATURE_SIZE >::pack_trits().

        {
            assert(val == Brighter || val == Darker || val == Similar);
            assert(tnum < max_size);

            if(val == Brighter)
                tests[tnum] = 1;
            else if(val == Darker)
                tests[tnum + max_size] = 1;
        }

---

Member Data Documentation

template<int FEATURE_SIZE>

unsigned long datapoint< FEATURE_SIZE >::count

Number of instances.

Definition at line 163 of file learn_fast_tree.cc.

template<int FEATURE_SIZE>

bool datapoint< FEATURE_SIZE >::is_a_corner

Class.

Definition at line 164 of file learn_fast_tree.cc.

template<int FEATURE_SIZE>

const unsigned int datapoint< FEATURE_SIZE >::max_size = FEATURE_SIZE [static]

Maximum number of features representable.

Definition at line 166 of file learn_fast_tree.cc.

Referenced by datapoint< FEATURE_SIZE >::get_trit(), and datapoint< FEATURE_SIZE >::set_trit().

template<int FEATURE_SIZE>

bitset<max_size*2> datapoint< FEATURE_SIZE >::tests [private]

Used to store the ternary vector Ternary bits are stored using 3 out of the 4 values storable by two bits.

Trit n is stored using the bits n and n + max_size, with bit n being the most significant bit.

The values are

• 3 unused
• 2 Brighter
• 1 Darker
• 0 Similar

Definition at line 185 of file learn_fast_tree.cc.

Referenced by datapoint< FEATURE_SIZE >::get_trit(), datapoint< FEATURE_SIZE >::pack_trits(), and datapoint< FEATURE_SIZE >::set_trit().

---

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

• learn_fast_tree.cc

---