LAWE (Learning Algorithms from Worked Examples) is a computational model
of imitation learning. LAWE is able to learn a procedure by the user demonstrating
the steps.
LAWE is an example of Programming by Demonstration (PBD). This is also known by some authors as Programming by Example (PBE). I prefer the title "Programming by Demonstration because it makes explicit that the user demonstrates the steps of a procedure. This used to be known in AI as programming from traces. AI people tend to think of Programming by Example as programming from input-output example pairs which is completely different.
Programming by Imitation (PBI) is a phrase specific to the LAWE system and involves the user demonstrating the steps of the procedure but with no need for any question-answering or other forms of interaction. PBI is explicitly modelled on how humans learn by imitation and this makes LAWE an extremely friendly and easy to use system.
Slides on PBD
from one of my lectures.
Notes on PBD
from the same lecture.
The best site on PBD is Henry Lieberman's Programming by Example site and his Lieberary of downloadable examples:
Programming by Example
Example PBD Systems
PBD systems are often illustrated by Quicktime movies. Two well known examples are:
Allen Cypher's EAGER
Henry Liebermann's TINKER
Publications
Cognitive Science 1997 abstract (one page) (MS Word format)
AISB99 workshop on Imitation poster (four pages) (MS Word format)
AISB99 workshop on Imitation paper (9 pages) (MS Word format)
Paper from the journal "Cybernetics and Systems, 2001. (40 pages) (PDF format)
Last updated 5 Feb 2002