Title :
Spike correlation based learning for unsupervised neural lattice structures
Author :
Rouw, Eelco ; Hoekstra, Jaap ; Van Roermund, Arthur H M
Author_Institution :
Electron. Lab., Delft Univ. of Technol., Netherlands
Abstract :
Neural networks are suitable for implementation in nanoelectronics due to their fault tolerance. Other restrictions, a low fanin/fanout and restrictions to short connections force a different approach towards neural networks. This paper proposes techniques to overcome the topological restrictions using segmentation and bio-inspired local learning rules with spike encoding as well as a way to exploit the inherent stochastic nature of nanoelectronic devices in neural subsystems
Keywords :
correlation theory; neural nets; stochastic systems; unsupervised learning; fault tolerance; information encoding; lattice structure; local learning rule; nanoelectronic device; neural network; spike correlation; stochastic subsystem; topological segmentation; unsupervised learning; Artificial neural networks; Encoding; Fault tolerance; Integrated circuit interconnections; Laboratories; Lattices; Nanobioscience; Neural networks; Stochastic processes; Stochastic systems;
Conference_Titel :
Circuits and Systems, 2001. ISCAS 2001. The 2001 IEEE International Symposium on
Print_ISBN :
0-7803-6685-9
DOI :
10.1109/ISCAS.2001.921338
