Title :
On the stability of bilayer associative memories
Author :
Sakaue, Takatoshi ; Matsuzaki, Manabu ; Miyata, Mnsnchika
Author_Institution :
Kanazawa Inst. of Technol., Ishikawa, Japan
Abstract :
A recurrent bilayer neural associative memory using mid-step type and ordinary sigmoid type neurons has less destroyed and spurious memories compared with a single layer memory using correlation learning. However, its stability has not been clarified. This paper shows that such a network is stable for some relaxation
Keywords :
content-addressable storage; recurrent neural nets; stability; bilayer associative memories; mid-step type neurons; recurrent neural associative memory; sigmoid type neurons; stability; Artificial intelligence; Associative memory; Learning systems; Linearity; Neurons; Stability;
Conference_Titel :
Circuits and Systems, 2001. ISCAS 2001. The 2001 IEEE International Symposium on
Conference_Location :
Sydney, NSW
Print_ISBN :
0-7803-6685-9
DOI :
10.1109/ISCAS.2001.921379
