Novelty detection on a very large number of memories stored in a Hopfield-style network

Summary form only given, as follows. The Hopfield-style network does not forget any stored memory. Its error-correcting performance, however, degrades rapidly with a large number of stored memories because of the rapid development of spurious memories. The author shows that in spite of this, the network performs well in the case of error detection, specifically novelty detection, even on very large collections of stored memories. The speed with which the network can detect novelty, even on very large collections of stored memories, is the main advantage of this scheme. Efficient distributed storage of large collections of memories is another important advantage