Information Technology Reference
In-Depth Information
4
The Basic GEA
in Problem Solving
In this chapter we will see how the basic gene expression algorithm can be
used to solve complex problems from very different fields. We will start by
exploring the workings of the fundamental parameters of the algorithm by
solving a simple cubic polynomial. Then we will continue our exploration
by analyzing the performance of the algorithm on a complex test function of
five arguments. Furthermore, we will also discuss the power of the algo-
rithm to extract knowledge from noisy data, not only by mining a noisy com-
puter-generated dataset but also by mining complex real-world data, includ-
ing mining a dataset with 51 input attributes in order to decide whether to
approve or not a credit card. We will also see how the basic GEA can be used
to diagnose diseases and classify different types of plants. Particularly inter-
esting is the three-class prediction of the iris data that will be tackled using
two different approaches: the first consisting of the conventional way of
partitioning the data into three separate datasets so that three different mod-
els are created and afterwards combined to make the final prediction; and the
second consisting of a three-genic system evolving three different models at
the same time, in which each model is responsible for classifying a certain
type of plant.
Also in this chapter we will learn how gene expression programming can
be used to find parsimonious solutions. Indeed, finding parsimonious solu-
tions is not only a matter of concern in mathematics but also, and perhaps
most importantly, in logic. And, in section 4.3, Logic Synthesis and Parsimo-
nious Solutions, we will see how gene expression programming can be used
to find parsimonious solutions with aplomb. For that purpose, special fitness
functions with parsimony pressure will be described and a simple suite of
Boolean functions will be solved in order to find their most parsimonious
representations using five well-known universal systems: the classical Boolean
system of ANDs, ORs, and NOTs; the NAND and NOR systems; the
