Information Technology Reference
In-Depth Information
Case-Based Reasoning (CBR) is a relatively developed branch in Artificial
Intelligence. It is a kind of reasoning based on previously practical
experiences. From traditional point of view, reasoning is a process of
conclusion drawing by “cause-effect” chains, which has been used by many
expert systems. CBR is a quiet different view, in which knowledge is
implicitly contained in cases, not in the form of rules explicitly. These cases
record all sorts of relevant contexts in the past. CBR solves new problems by
adapting previously successful solutions to similar problems, rather than by a
reasoning chain. For a new problem, CBR retrieves the most relevant case
from memory or in the case base, and then does some revision to the case to
produce a suitable solution to the present problem.
CBR is rational because there are two characteristics in the reality:
regulation and repetition. Wholly speaking, there are certain regulations exist
in the world, and actions happening under similar conditions will produce the
similar results. As it goes, “reality takes shape in the memory alone”, and the
past experiences may carry some hints to the future.
The researches on CBR originate from the investigations on the
mechanism of reasoning and learning from a cognitive perspective. From the
children's simple activity to experts' cautious decision, human affairs are
usually accomplished with the help of people's recollection unconsciously or
consciously. The mankind often acts according to experiences, and people are
intelligent systems in some sense, so it is intuitive to applying such kind of
reasoning based on experiences to researches and applications of Artificial
Intelligence. Generally speaking, CBR has made the following contributions to
Artificial Intelligence:
(1) knowledge acquisition
This is a difficult process, often referred to as the
knowledge
in knowledge-based systems. The most
challenges to implement knowledge-based systems are the elicitation of an
explicit model of the domain and the implementation of knowledge often in the
form of rules, which require teamwork between domain experts and knowledge
engineers. Sometimes creating an explicit model of the domain knowledge is
extremely difficult. CBR does not require an explicit domain model, and the need
for knowledge acquisition in CBR can be limited to establishing how to
characterize cases.
acquisition
bottleneck
