Page 28 - Biosystems Engineering

Micr oarray Data Analysis Using Machine Learning Methods 9

The fuzzification unit converts these inputs into fuzzy sets based up
on fuzzy values, such as “low,” “medium,” and “high.” In this unit,
membership functions defined on the input variables are applied to
their actual values to determine the degree of truth for each rule
premise. The knowledge base contains membership functions defin-
ing fuzzy values and a set of fuzzy rules. The inference unit executes
these fuzzy rules. The truth value for the premise of each rule is com-
puted and applied to the conclusion part of each rule. When all the
rules are executed, a fuzzy region will be created for the output vari-
able y. With the process of defuzzification, a crisp value of the output
will be generated as a solution.
1.2.4 Genetic Algorithms
Genetic algorithms are global optimization algorithms that originated
from mechanics of natural genetics and selection (Holland 1975;
Goldberg 1989). They provide a method of problem solving that is
based on genetic evolution. Based on probabilistic decisions, they
exploit historic information to guide the search for better solutions in
the problem space. Using a direct analogy of natural evolution,
genetic algorithms work with a population of chromosomes where
each chromosome encodes a possible solution to the problem as a
string of bits (0100101010), a list of real values (0.2, 6.5, 4.1, 1.3), a
string of characters (B2, A3, C1, A1), or some other representation.
GAs start by randomly creating a population of chromosomes. The
chromosomes in the population are evaluated by a fitness function
indicating how good encoded solutions appear with respect to the
problem under consideration. The algorithm takes the chromosomes
with higher relative fitness (chosen as parents) and then creates a new
generation of chromosomes using genetic operators such as crossover
or mutation. In crossover, new offspring are created from two parents
by swapping a portion of their strings. In mutation, offspring are
identical to their parents, but they have random changes in portions
of their strings. The algorithm repeats the foregoing steps until a pre-
defined number of generations or fitness value is reached.
Genetic algorithms, therefore, begin with a random process and
arrive at an optimized solution. They are thus well suited for those
tasks that seek global optimization. They are highly effective in situ-
ations where many inputs interact to produce a large number of pos-
sible outputs or solutions. They are a robust search method requiring
little information to search effectively in a large or poorly understood
search space.

1.2.5 Particle Swarm Optimization
PSO is similar to GAs; each uses a population of potential solutions to
explore the search space. Although GAs are based on survival-of-the-
fittest approaches as in the theory of natural evolution, the PSO is an