Comments on learning and adaptive machines for pattern recognition

Abstract

By learning and adaptive machines I mean special purpose machines with internal components having adjustable values. The use of such machines for the solution of several kinds of problems has been proposed. These include the design of switching functions, the design of classification machines (classification machines is meant to imply recognition, prediction, decision and classification machines), automatic manufacturing of certain devices, self optimization of decision machines with time variable input statistics, improving the reliability of digital processes, and automatic wiring and testing of microcomponents. Most of the work to date has been on using learning machines to design switching functions and classification machines.