*/
void buildIndex()
{
- std::vector<size_t> indices(feature_size_ * CHAR_BIT);
-
tables_.resize(table_number_);
for (unsigned int i = 0; i < table_number_; ++i) {
-
- //re-initialize the random indices table that the LshTable will use to pick its sub-dimensions
- if( (indices.size() == feature_size_ * CHAR_BIT) || (indices.size() < key_size_) )
- {
- indices.resize( feature_size_ * CHAR_BIT );
- for (size_t i = 0; i < feature_size_ * CHAR_BIT; ++i)
- indices[i] = i;
- std::random_shuffle(indices.begin(), indices.end());
- }
-
lsh::LshTable<ElementType>& table = tables_[i];
- table = lsh::LshTable<ElementType>(feature_size_, key_size_, indices);
+ table = lsh::LshTable<ElementType>(feature_size_, key_size_);
// Add the features to the table
table.add(dataset_);
* @param feature_size is the size of the feature (considered as a ElementType[])
* @param key_size is the number of bits that are turned on in the feature
*/
- LshTable(unsigned int /*feature_size*/, unsigned int /*key_size*/, std::vector<size_t> & /*indices*/)
+ LshTable(unsigned int /*feature_size*/, unsigned int /*key_size*/)
{
std::cerr << "LSH is not implemented for that type" << std::endl;
assert(0);
// Specialization for unsigned char
template<>
-inline LshTable<unsigned char>::LshTable( unsigned int feature_size,
- unsigned int subsignature_size,
- std::vector<size_t> & indices )
+inline LshTable<unsigned char>::LshTable(unsigned int feature_size, unsigned int subsignature_size)
{
initialize(subsignature_size);
// Allocate the mask
mask_ = std::vector<size_t>((size_t)ceil((float)(feature_size * sizeof(char)) / (float)sizeof(size_t)), 0);
+ // A bit brutal but fast to code
+ static std::vector<size_t>* indices = NULL;
+
+ //Ensure the Nth bit will be selected only once among the different LshTables
+ //to avoid having two different tables with signatures sharing many dimensions/many bits
+ if( indices == NULL )
+ {
+ indices = new std::vector<size_t>( feature_size * CHAR_BIT );
+ }
+ else if( indices->size() < key_size_ )
+ {
+ indices->resize( feature_size * CHAR_BIT );
+ for (size_t i = 0; i < feature_size * CHAR_BIT; ++i) {
+ (*indices)[i] = i;
+ }
+ std::random_shuffle(indices->begin(), indices->end());
+ }
+
// Generate a random set of order of subsignature_size_ bits
for (unsigned int i = 0; i < key_size_; ++i) {
- //Ensure the Nth bit will be selected only once among the different LshTables
- //to avoid having two different tables with signatures sharing many dimensions/many bits
- size_t index = indices[0];
- indices.erase( indices.begin() );
+ size_t index = (*indices)[0];
+ indices->erase( indices->begin() );
// Set that bit in the mask
size_t divisor = CHAR_BIT * sizeof(size_t);