Carbonell, J. (1986). Derivational analogy: A theory of reconstructive problem solving and expertise acquisition. In Michalski, R., Carbonell, J., & Mitchell, T. (Eds.) Machine Learning: An Artificial Intelligence Approach. Morgan Kaufman Publishers: San Mateo, CA.

@InBook{,
  ALTauthor = 	 {Jaime Carbonell},
  ALTeditor = 	 {Michalski, R.,
Carbonell, J., & Mitchell, T.},
  title = 	 {Machine Learning: An
Artificial Intelligence Approach},
  chapter = 	 { Derivational analogy: A theory of reconstructive
problem solving and expertise acquisition},
  publisher = 	 {Morgan Kaufman Publishers},
  year = 	 {1986},
  OPTaddress = 	 {San Mateo, CA},
}

Author of the summary: Jim Davies, 2002, jim@jimdavies.org

Cite this paper for:

• Derivational Analogy
• derivational trace

• Search through operators: Good when there is little knowledge
• Plan instantiation: When plans are appropriate
• General Plan use: abstract plan gets used
• Analogy: When plans are relevent but not exact. [p636]

Derivational Analogy: Stores the goals and justifications at different levels for the actions taken, to facilitate plan adaptation. I think that's why the term "derivational" is used. Imagine quicksort implemented in Pascal and LISP. Their similarity is at the general specification, not in the actual code. The actual code really wouldn't help.

The derivational trace is stored which contains decisions, justifications, and failure-cause propogations, as well as problems, steps, and the solution.[640]

Also described are ways to get rules out of the cases in memory [644]

Summary author's notes:

• Page numbers are from the reprint in Shavlik and Dietterich's Readings in Machine Learning.
• This is distinguished from my work in that I only save the solution, and not the derivation, and I generate states in the solution, in addition to the actions taken.