Computer Sciences Dept.

Flexible Pattern Recognition

Leonard Uhr

This paper presents and describes a sequence of three computer programs that examine what "flexibility" might mean in the context of pattern recognition. Flexibility is a vague, but important, concept, and it is something that artificial intelligence programs have been accused of being without. Various possible meanings of the concept are discussed and programmed. Essentially, flexibility is taken to point to a rich set of methds, which are decided upon and changed, as appropriate. In pattern recognition, this means making a sequence of parallel characterizations, where the program decides, as a function of what it has learned so far about the pattern instance it is trying to recognize, what might be there, and what characterizers should therefore be applied next, and where.

Download this report (PDF)

Return to tech report index

Computer Science | UW Home