Optimising memory usage in n -tuple neural networks

摘要: Abstract The use of n -tuple or weightless neural networks as pattern recognition devices is well known (Aleksander and Stonham, 1979). They have some significant advantages over the more common biologically plausible networks, such multi-layer perceptrons; for example, been used a variety tasks, most popular being real-time recognition, they can be implemented easily in hardware standard random access memories. In operation, series images an object are shown to network, each processed suitably effectively stored memory called discriminator. Then, when another image system, it similar manner system reports whether recognises image; sufficiently one already taught? If able recognise discriminate between m -objects, then must contain -discriminators. This require great deal memory. paper describes various ways which requirements reduced, including novel method multiple discriminator recognition. By using this method, normally required handle -objects 2 — objects.