Information Technology Reference
In-Depth Information
1 Introduction
The work in this topic shows how acquiring a model-centred view to reformula-
ting Learning Classifier Systems (LCS), a rule-based method for machine lear-
ning, provides an holistic approach to their design, analysis and understanding.
This results in a new methodology for their design and analysis, a probabilistic
model of their structure that reveals their underlying assumptions, a formal de-
finition of when they perform optimally, new approaches to their analysis, and
strong links to other machine learning methods that have not been available
before. The work opens up the prospects of advances in several areas, such as
the development of new LCS implementations that have formal performance
guarantees, the derivation of representational properties of the solutions that
they aim for, and improved performance.
Introducing the work start with a short overview of machine learning, its
applications, and the most common problem types that it is concerned with.
An example that follows highlights the difference between ad-hoc and model-
centred approaches to designing machine learning algorithms and emphasises
the advantages of the latter. This is followed by a short introduction to LCS,
their applications and current issues. Thereafter, the research question of this
work is introduced, together with the approach that is used to approach this
question, and a short overview of the chapters that are to follow.
1.1
Machine Learning
Machine learning (ML) is a sub-field of artificial intelligence (AI) that is con-
cerned with methods and algorithms that allow machines to learn. Thus, rather
than instructing a computer explicitly with regards to which aspects certain data
is to be classified, about relations between entities, or with which sequence of
actions to achieve certain goals, machine learning algorithms allow this know-
ledge to be inferred from a limited number of observations, or a description of
the task and its goal.
Their use is manifold, including speech and handwriting recognition, object
recognition, fraud detection, path planning for robot locomotion, game playing,
