RandomizedLDS< LDS, PRS > Class Template Reference

Randomized (random shift) low-discrepancy sequence. More...

#include <ql/math/randomnumbers/randomizedlds.hpp>

Public Types

typedef Sample< std::vector
< Real > > 
sample_type
 

Public Member Functions

 RandomizedLDS (const LDS &ldsg, const PRS &prsg)
 
 RandomizedLDS (const LDS &ldsg)
 
 RandomizedLDS (Size dimensionality, BigNatural ldsSeed=0, BigNatural prsSeed=0)
 
const sample_typenextSequence () const
 returns next sample using a given randomizing vector
 
const sample_typelastSequence () const
 
void nextRandomizer ()
 
Size dimension () const
 

Detailed Description

template<class LDS, class PRS = RandomSequenceGenerator<MersenneTwisterUniformRng>>
class QuantLib::RandomizedLDS< LDS, PRS >

Randomized (random shift) low-discrepancy sequence.

Random-shifts a uniform low-discrepancy sequence of dimension $ N $ by adding (modulo 1 for each coordinate) a pseudo-random uniform deviate in $ (0, 1)^N. $ It is used for implementing Randomized Quasi Monte Carlo.

The uniform low discrepancy sequence is supplied by LDS; the uniform pseudo-random sequence is supplied by PRS.

Both class LDS and PRS must implement the following interface:

LDS::sample_type LDS::nextSequence() const;
Size LDS::dimension() const;
Precondition
LDS and PRS must have the same dimension $ N $
Warning:
Inverting LDS and PRS is possible, but it doesn't make sense.
Possible enhancements:
implement the other randomization algorithms
Tests:
correct initialization is tested.

Member Function Documentation

void nextRandomizer ( )

update the randomizing vector and re-initialize the low discrepancy generator