Graphics Reference
In-Depth Information
8.4.3.7 Fixed+Wrapper
•
Random Mutation Hill Climbing (RMHC)
[
147
]—It randomly selects a subset
S
from
TR
which contains a fixed number of instances
s
(
s
. In each
iteration, the algorithm interchanges an instance from
S
with another from
TR
-
S
. The change is maintained if it offers better accuracy.
=
%
|
TR
|
)
8.5 Related and Advanced Topics
Research in enhancing instance and PS through other data reduction and learning
methods is common and in high demand nowadays. PS could represent a feasible
and promising technique to obtain expected results, which justifies its relationship
to other methods and problems. This section provides a wide review of other topics
closely related to PS and describes other works and future trends which have been
studied in the last few years. In each subsection, we provide a table that enumerates,
in not an exhaustive way, the most relevant methods and papers in each of the topics.
Although we do not extend them, it is included for informative purposes for the
interested reader.
8.5.1 Prototype Generation
Prototype generation methods are not limited only to select examples from the train-
ing set. They could also modify the values of the samples, changing their position in
the
d
-dimensional space considered. Most of them use merging or divide and conquer
strategies to set new artificial samples [
27
], or are based on clustering approaches
[
12
], LVQ [
98
] hybrids, advanced proposals [
102
,
113
] and evolutionary algorithms
based schemes [
25
,
151
,
152
]. A complete survey on this topic is [
153
].
Table
8.2
itemizes the main prototype generation methods proposed in the litera-
ture.
8.5.2 Distance Metrics, Feature Weighting and Combinations
with Feature Selection
This area refers to the combination of IS and PS methods with other well-known
schemes used for improving accuracy in classification problems. For example, the
weighting scheme combines the PS with the FS [
40
,
147
] or Feature Weighting [
55
,
129
,
163
], where a vector of weights associated with each attribute determines and
influences the distance computations.
