Title :
Associative memories with "killed" neurons: The methods of recovery
Author :
Reznik, A.M. ; Sitchov, A.S. ; Dekhtyarenko, O.K. ; Nowicki, D.W.
Author_Institution :
Inst. of the Math. Machines & Syst., Ukrainian Nat. Acad. of Sci., Kiev, Ukraine
Abstract :
We consider re-learning ability of a Hopfield-type network after killing some neurons. Neurons were "killed" by means of ification of corresponding rows and columns of the synaptic matrix. We show that one can restore recognition ability of this network using re-training with the vectors, which were memorized before. The number of vectors needed is equal to the number of deleted neurons. It does not depend on network\´s size and on volume of stored data.
Keywords :
content-addressable storage; neural nets; Hopfield-type network; associative memories; neurons; synaptic matrix; Adaptive filters; Associative memory; Convergence; Equations; Hopfield neural networks; Neural networks; Neurons; Stability; Vectors;
Conference_Titel :
Neural Networks, 2003. Proceedings of the International Joint Conference on
Print_ISBN :
0-7803-7898-9
DOI :
10.1109/IJCNN.2003.1223972
