Biomedical Engineering Reference
In-Depth Information
over the range of 0 to 65,535, using the same numerical starting point
(seed).
One of the most common methods of implementing a pseudorandom number generator is through
the use of a Linear Congruential Generator (LCG) algorithm, which produces a series of numbers in
which each successive term is computed from its predecessor. As such, the LCG produces a
pseudorandom sequence because each number isn't independent of all earlier numbers.
The arbitrary starting point for the LCG algorithm is called the
seed
. Because an LCG algorithm
computes successive terms from predecessor terms, including seed values, a new seed value results
in a different random number sequence. Conversely, the same seed will result in the same
pseudorandom sequence. As a result, one way to help ensure that the same pseudorandom sequence
isn't generated at each run of a simulation is to use a new seed value.
Another limitation of pseudorandom number generators is that the series of random numbers repeats
with a finite period. Furthermore, in some implementations of the LCG algorithm, the numbers
between zero and the period aren't fully represented in the output of the generator. If the nature of
the output of the system-supplied pseudonumber generator is unacceptable, there are a number of
possible solutions. The simplest is to add a randomizing shuffle to the output of the LCG to alter the
deterministic nature of the series. The effect of this shuffle is to randomize the output of the LCG.
However, the simple shuffle can't overcome the problem of a sparse period, because a shuffle will not
produce additional numbers, but instead alters only the sequence of the numbers produced. If a
richer period is required, then a substitute for the LCG will have to be developed.
Algorithms
Modeling in bioinformatics is a multidisciplinary activity that borrows algorithms from statistics,
mathematics, Artificial Intelligence (AI), and even robotics. For example, robotics algorithms are
being used to explore the manipulation of proteins by chaperone molecules. Instead of defining a
rotation or unfolding of a protein in 3D space, the space is split into
n
-dimensions. As a result, the
movements of molecules can be described with simple linear functions that are much less
computationally intensive than vector algebra. In addition to the esoteric algorithms that are useful in
niche areas of bioinformatics, there are several algorithms that have general applicability in modeling
