Guide to pattern recognition using random-access memories

About 12 years of work with a specific type of learning pattern-recognition system are reviewed. The principles and characteristics of the scheme, which is based on random-access-memory implementation, are discussed in some detail. Methods of improving performance and cost-optimising pattern recognisers are presented, together with case studies in a variety of fields including the recognition of alphanumerics, chemical data and faults in digital circuit boards.