Next: Overview
Up:
Previous: Volume:
  Contents
Subsections
- Overview
- About random numbers
- Generation of Random Numbers
- Congruential Generators
- which defines as being congruent
to n modulo m. It means that
is the remainder
- Provided an initial seed
is supplied,
a string of distinct values can be generated.
- By taking
we find a sequence
of numbers distributed on [0,1].
- In the linear congruential generator, only one previous random variable
is used to derive the next one, we can use more of the earlier generated numbers,
each associated with a specified multiplier - a compound generator. If the last
k generated numbers are used, we can write
- where
and the set of multipliers is a= (aa,..,a).
By taking a value of just 2 for k and suitable multipliers aand
a, a compound generator with acceptable result structure
can be obtained, described next.
- Shift-Register Generators
- Fibonacci generators
- We can generate a sequence of numbers, given that the first p numbers
are provided e.g. from a multiplicative congruential generator. The operation
denoted by
is conventionally the `exclusive-or'.
It may be however just simple addition or subtraction, and recent studies favour
the latter operation, giving rise to a subtracted Fibonacci generator.And especially
interesting development is the subtract-with-borrow generator of this type.
The algorithm is
- The `carry' coefficient,
, which must
be set to 0 or 1, arbitrarily, at initialization, is reassigned in each call
as follows; if the quantity in brackets is negative, so that m must be added
to it to carry out the modulo m operation,
is
set to zero otherwise it assumes the value unity. If m, p, and q satisfy certain
conditions, the period of this generator is m-m.
- Practical generators-Super Duper
- Unra generator in CECILE
- Testing random Number Sequences
Amaury LATAILLADE
2002-11-04