Title :
Convergence in Iteratively Formed Correlation Matrix Memories
Author :
Reid, Richard J. ; Frame, J. Sutherland
Author_Institution :
Department of Computer Science, Michigan State University
Abstract :
Correlation matrix memories are formed iteratively given a single cue-response pair at a time to be stored in an associative anner. If the cues presented are not orthogonal, the memory provides imperfect recollections after a single training exposure; however, if the cues are linearly independent, repetitions of the exposures can lead to ultimate perfect memorization and recollection ability.
Keywords :
Associative memory, associative recall, correlation matrix memory, improving associative recollections, iteratively formed associations.; Bismuth; Computer science; Convergence; Eigenvalues and eigenfunctions; Lead; Machine intelligence; Mathematics; Associative memory, associative recall, correlation matrix memory, improving associative recollections, iteratively formed associations.;
Journal_Title :
Computers, IEEE Transactions on
DOI :
10.1109/T-C.1975.224314
