Case-Based CBR : Capturing and Reusing Reasoning About Case Adaptation

1
Case-Based CBR Capturing and Reusing Reasoning
About Case Adaptation
David B. Leake, Andrew Kinley, David
Wilson International Journal of Expert Systems,
1997, Vol10, No2 p197213
???? ????? ????? ??M8702048 ??88/3/10
2
Why need case-based CBR

• CBR is receiving widespread use in applications,
  especially in retrieve and propose advisory
  system.
• Is difficult in choose a method for
  poorly-understand domain
• The difficulty of codifying the domain may also
  make it impractical to build the rule-based
  systems usually used for case adaptation the
  deep knowledge required to adapt cases may be
  precisely the same knowledge the would be needed
  to reason form scratch.

3
Case-based CBR

• The framework integrates case-based adaptation
  into the CBR system. As an intelligent
  component that automatically records and
  reapplies adaptations that the system initially
  performs form scratch.
• Uses derivational analogy to store and replay the
  reasoning.
• Develop method to monitor, record, and reapply
  traces of its own reasoning.

4

• Using Introspective reasoning-reasoning and
  learning about the reasoning process
• Acquires adaptation knowledge form experience
  with autonomous adaptation and by interactively
  capturing the adaptation process of a human user.
• The questions of case-based CBR
• System design
• Knowledge acquisition
• Effectiveness of learning

5
A study of case-based adaptation

• Ex DIAL disaster response for natural and
  man-made disasters.
• Include
• schema-based story understander process input
• response plan retriever
• simple evaluator for candidate response plans
• adaptation component to adapt retrieved plans

6
Basic sequence of processing step

• Case-based adaptation
• Rule-based adaptation(introspective planning)
• Plan evaluation
• Storage

7
Acquiring adaptation cases

• How a cased-based adaptation component can
  capture autonomous adaptations from the CBR
  systems own experience with adaptations
  performed form scratch
• How the cased-based adaptation components case
  library can be augmented by capturing the
  sequence of steps in a users cases adaptation
  process
• How the systems adaptation process can be
  redirected interactively through a simple
  interactive interface

8
Capture autonomous adaptations

• Starts with general rules about information
  search and about the constraints that must be
  satisfied by the case adaptation process.
• Store the trace of reasoning done
• Ex schoolchildren do not have union.

9

• Initial rule-based adaptation process provide
• flexibility to deal with novel adaptation
  problems
• a way to generate specific adaptation knowledge
• Rule-based adaptation process treats the
  adaptation problem as primarily one of finding
  the information needed to apply a set of very
  general transformations-substitutions, deletions,
  and additions to the structure of a retrieval
  solution-by search memory

10
Capture user adaptations

• Use the derivational approach to adaptation as
  the basis for an interface to facilitate direct
  acquisition of adaptation cases for a human user
• Have 3 problems cause adaptation to fail
• Resource limit exceeded
• Overspecific constraints
• Overgeneral constraints

11
Perspective on case-based adaptation

• Design choose
• transformational vs. derivational CBR
• case representation
• case organization
• adaptation of adaptation cased
• knowledge acquisition effort

12

• effects on adaptation efficiency
• adaptation learning can be important as case
  learning
• simple domain-independent methods can be
  sufficient for case-based CBR processes
• comparison to other method for addressing the
  adaptation problem.
• Ramifications of adaptation learning for
  case-based maintenance