include/lshkit/matrix.h File Reference

Dataset management. A dataset is maintained as a matrix in memory. More...

#include <fstream>
#include <boost/dynamic_bitset.hpp>
#include <lshkit/matrix-io.h>

Go to the source code of this file.

Namespaces
namespace	lshkit
Classes
class	lshkit::Matrix< T >
	Matrix. More...
class	lshkit::Matrix< T >::Accessor
	An accessor class to be used with LSH index. More...
Typedefs
typedef Matrix< float >	lshkit::FloatMatrix

---

Detailed Description

Dataset management. A dataset is maintained as a matrix in memory.

LSHKIT is a main memory indexing data structure and doesn't assume any particular file format. However, the standalone tools come with the LSHKIT package use the following simple binary file format.

The file contains N D-dimensional vectors of single precision floating point numbers.

In the beginning of the file are three 32-bit unsigned integers: ELEM-SIZE, SIZE, DIM. ELEM-SIZE is the size of the element, and currently the only valid value is 4, which is the size of float. SIZE is the number of vectors in the file and DIM is the dimension.

The rest of the file is SIZE vectors stored consecutively, and the total size is SIZE * DIM * 4 bytes.

Such binary files can be accessed using lshkit::Matrix<float>.

---