Inductive Acquisition of Expert Knowledge
dc.contributor.advisor
Michie, Donald
en
dc.contributor.author
Muggleton, Stephen H.
en
dc.date.accessioned
2013-11-13T14:11:39Z
dc.date.available
2013-11-13T14:11:39Z
dc.date.issued
1986
dc.description.abstract
Expert systems divide neatly into two categories: those in which ( 1) the expert decisions result in
changes to some external environment (control systems), and (2) the expert decisions merely seek
to describe the environment (classification systems). Both the explanation of computer-based
reasoning and the "bottleneck" (Feigenbaum, 1979) of knowledge acquisition are major issues in
expert systems research. We have contributed to these areas of research in two ways. Firstly, we
have implemented an expert system shell, the Mugol environment, which facilitates knowledge
acquisition by inductive inference and provides automatic explanation of run-time reasoning on
demand. RuleMaster, a commercial version of this environment, has been used to advantage
industrially in the construction and testing of two large classification systems. Secondly, we have
investigated a new technique called sequence induction which can be used in the construction of
control systems. Sequence induction is based on theoretical work in grammatical learning. We
have improved existing grammatical learning algorithms as well as suggesting and theoretically
characterising new ones. These algorithms have been successfully applied to the acquisition of
knowledge for a diverse set of control systems, including inductive construction of robot plans and
chess end-game strategies.
en
dc.identifier.uri
http://hdl.handle.net/1842/8124
dc.language.iso
en
dc.publisher
The University of Edinburgh
en
dc.subject
computer science
en
dc.title
Inductive Acquisition of Expert Knowledge
en
dc.type
Thesis or Dissertation
en
dc.type.qualificationlevel
Doctoral
en
dc.type.qualificationname
PhD Doctor of Philosophy
en
This item appears in the following Collection(s)