Summary form only given. A class of models for an associative content-addressable memory is described. The models utilize layered neural networks analogous to those of the brain, but possess analogies to standard digital computer memories as well. When a cue is presented at the input layer, an intermediate processing layer retrieves a label associated with a stored word and the output layer displays the stored word. All computation is massively parallel. The algorithms have been constrained by limitations such as fan out so as to be implementable using analog VLSI technology currently undergoing intensive and promising development for other connectionist algorithms.
|Original language||English (US)|
|Number of pages||1|
|State||Published - 1987|
ASJC Scopus subject areas