“Empyrean”, an alternative paradigm for pattern recognition

作者:

Highlights:

摘要

It is pointed out that along with the measurement subproblem and the decision subproblem of pattern recognition, there exists the feature-extraction subproblem. A general structure is proposed, consisting of an “ascending hierarchy” of sequential machines which transduce the input word (measurement vector) to a set of pattern description vectors, and a descending hierarchy of finite state pattern recognizers (automata). An iterative procedure for constructing the latter for prescribed statistical performance is given, with certain heuristics employed in state reduction of the resulting automata to yield certain types of extrapolations. A comparison is made with the analogous approach of Hunt, Marin and Stone.(4)

论文关键词:

论文评审过程:Received 11 April 1968, Available online 16 May 2003.

论文官网地址:https://doi.org/10.1016/0031-3203(68)90005-8