Create a new object.
this = lowdisc_new ( ) this = lowdisc_new ( method )
the current object
a 1-by-1 matrix of strings, the low discrepancy sequence (default = "sobolf"). The method can be equal to : "halton", "haltonf", "faure", "fauref", "reversehalton", "reversehaltonf", "sobol", "sobolf", "niederreiter-base-2", "niederreiterf". See below for details.
This function creates a new low discrepancy object.
The following is the list of possible values for method.
"-primeslist"
option.
By default, it is able to generate experiments in dimension at most 100.
This sequence is able to generate at most 2^52 - 1 = 4 503 599 627 370 495 experiments.
To extend the number of dimensions, please configure the "-primeslist"
option.
The maximum dimension which can be obtained is equal to the number of
different primes in the prime table.
For the Halton sequence, the skip
and
leap
options are fast, that is, large
values of the skip
or leap
parameter do not reduce the performance.
This is a macro-based algorithm.
This implementation is based on the book by Paul Glasserman,
"Monte-Carlo methods in Financial Engineering"."-primeslist"
option.
By default, it is able to generate experiments in dimension at most 100.
This sequence is able to generate at most 2^31 - 1 = 2 147 483 647 experiments.
To extend the number of dimensions, please configure the "-primeslist"
option.
The maximum dimension which can be obtained is equal to the number of
different primes in the prime table.
For the Halton sequence, the skip
and
leap
option is fast, that is, large
values of the skip
or leap
parameter do not reduce the performance.
This fast algorithm is based on a compiled C source code.
The implementation is a modification of the C source code by John Burkardt."-primeslist"
option.
By default, it is able to generate experiments in dimension at most 541.
This sequence is able to generate at most 2^52 - 1 = 4 503 599 627 370 495 experiments.
To extend the number of dimensions, please configure the "-primeslist"
option.
The maximum dimension which can be obtained is equal to the maximum
prime number in the prime table.
For the Faure sequence, the skip
and
leap
option is fast, that is, large
values of the skip
or leap
parameter do not reduce the performance.
This is a macro-based algorithm.
This implementation is based on the book by Paul Glasserman,
"Monte-Carlo methods in Financial Engineering"."-primeslist"
option.
By default, it is able to generate experiments in dimension at most 541.
This sequence is able to generate at most 2^31 - 1 = 2 147 483 647 experiments.
To extend the number of dimensions, please configure the "-primeslist"
option.
The maximum dimension which can be obtained is equal to the maximum
prime number in the prime table.
For the Faure sequence, the skip
and
leap
option is fast, that is, large
values of the skip
or leap
parameter do not reduce the performance.
This fast algorithm is based on a compiled C source code.
The implementation is a modification of the C source code by John Burkardt.
Original Fortran 77 version by Bennett Fox in "Algorithm 647: Implementation and Relative Efficiency of
Quasirandom Sequence Generators"."-primeslist"
option.
By default, it is able to generate experiments in dimension at most 100.
This sequence is able to generate at most 2^52 - 1 = 4 503 599 627 370 495 experiments.
To extend the number of dimensions, please configure the "-primeslist"
option.
The maximum dimension which can be obtained is equal to the number of
different primes in the prime table.
For the Reverse Halton sequence, the skip
and
leap
option are fast, that is, large
values of the skip
or leap
parameter do not reduce the performance.
This is a macro-based algorithm.
This implementation is based on the paper "Good permutations for deterministic scrambled
Halton sequences in terms of L2-discrepancy" by B. Vandewoestyne and R. Cools."-primeslist"
option.
By default, it is able to generate experiments in dimension at most 100.
This sequence is able to generate at most 2^31 - 1 = 2 147 483 647 experiments.
To extend the number of dimensions, please configure the "-primeslist"
option.
The maximum dimension which can be obtained is equal to the number of
different primes in the prime table.
For the Reverse Halton sequence, the skip
and
leap
option are fast, that is, large
values of the skip
or leap
parameter do not reduce the performance.
This fast algorithm is based on a compiled C source code.
This implementation is based on the paper "Good permutations for deterministic scrambled
Halton sequences in terms of L2-discrepancy" by B. Vandewoestyne and R. Cools.skip
and
leap
options are slow, that is, large
values of the skip
or leap
parameter leads to increased CPU time.
This is because generating the new vector implies to update iteratively
the lastq vector.
This is a macro-based algorithm.
The current implementation is a Scilab port of the source code in Matlab
by John Burkardt. The original source code was created by Bennett Fox in Fortran
in "Algorithm 647: Implementation and Relative Efficiency of
Quasirandom Sequence Generators".skip
and
leap
options are slow, that is, large
values of the skip
or leap
parameter leads to increased CPU time.
This is because generating the new vector implies to update iteratively
the lastq vector.
This fast algorithm is based on a compiled C source code.
The implementation is a modification of the C source code by John Burkardt.
Original Fortran 77 version by Bennett Fox in "Algorithm 647: Implementation and Relative Efficiency of
Quasirandom Sequence Generators".skip
and
leap
options are slow, that is, large
values of the skip
or leap
parameter leads to increased CPU time.
This is because generating the new vector implies to update iteratively
the nextq
vector.
The implementation is a Scilab port of the Matlab source code by John Burkardt.
Original Fortran 77 version by Paul Bratley, Bennett Fox, Harald Niederreiter
in "Algorithm 738: Programs to generate Niederreiter's low-discrepancy
sequences".skip
and
leap
option is slow, that is, large
values of the skip
or leap
parameter leads to increased CPU time.
This is because generating the new vector implies to update iteratively
the nextq vector.
The library generates two temporary files gfarit and gfplys when the sequence is started up.
This fast algorithm is based on a compiled C source code.
This is a C port of "Algorithm 738: Programs to generate Niederreiter's low-discrepancy
sequences" (1994) by Paul Bratley, Bennett Fox, Harald Niederreiter.
The C port has been done by John Burkardt in 2005-2009. The library has
been updated to present a uniform API.