Information Technology Reference
In-Depth Information
development stage. Regarding development models, this test data generation
issue arises in all models covered by the taxonomy. “D.2.5 Testing and Debug-
ging” would be the main subject descriptor of such problem and “Testing Tools”
would work as implicit descriptor.
Considering now the OPTIMISATION perspective, this problem could be
characterised as mono-objective, since it composes two measures (approach level
and branch distance) in a single evaluation function. Additionally, it works with
a usually continuous instance space formed of input vectors. It is also an un-
constrained and linear problem. Finally, the Structural Test Data Generation
problem can be seen as a simple instantiation of the BASIC OPTIMISATION
PROBLEM described earlier, where, given a target structure t in p , the problem
involves simply searching for an input vector
D , representing elements in
the instance space U , with minimum value given by evaluation function fit ( t,
i
),
representing f ( u ) in the description of the BASIC OPTIMISATION PROBLEM.
Finally, the Software Cost Estimation problem is associated with the Software
Planning development phase and all development models (Table 7). The most
adequate mains and implicit subject descriptors would be “D.2.9 Management”
and “Cost Estimation”, respectively.
Similarly to the Test Data Generation problem, this problem is an instantia-
tion of the BASIC OPTIMISATION PROBLEM. In this case, the problem seeks
solutions represented by well-formed functions, forming the instance set U ,look-
ing for a solution with minimum value given by function f ( u ), associated with
measures such as minimum squared error or correlation coecient . Additionally,
the problem should be classified as mono-objective, continuous, unconstrained
and nonlinear.
i
10 Next Steps: Getting Started
This section is primarily aimed at those who have not used SBSE before, but
who have a software engineering application in mind for which they wish to ap-
ply SBSE. Throughout this section the emphasis is unashamedly on obtaining
the first set of results as quickly as possible; SBSE is attractive partly because it
has a shallow learning curve that enables beginner to quickly become productive.
There is an excitement that comes with the way in which one can quickly as-
semble a system that suggests potentially well optimised solutions to a problem
that the experimenter had not previously considered.
By minimising the time from initial conception to first results, we seek to
maximise this excitement. Of course subsequent additional work and analysis
will be required to convert these initial findings into a suciently thorough
empirical study for publication. The goal of the section is to take the reader
from having no previous work on SBSE to the point of being ready to submit
their first paper on SBSE in seven simple steps. The first four of these steps are
sucient to gain the first results (and hopefully also the excitement that comes
with the surprises and insights that many authors have experienced through
using SBSE for the first time).
 
Search WWH ::




Custom Search