Title :
A Prototype CVNS Distributed Neural Network Using Synapse-Neuron Modules
Author :
Khodabandehloo, Golnar ; Mirhassani, Mitra ; Ahmadi, Majid
Author_Institution :
Dept. of Electr. & Comput. Eng., Univ. of Windsor, Windsor, ON, Canada
fDate :
7/1/2012 12:00:00 AM
Abstract :
Continuous valued number system (CVNS) has the fulfilment of being used for building resistive-type distributed neural networks (DNNs). CVNS DNNs benefit form scalability and higher noise immunity compared to the conventional DNNs. In this paper, CVNS is used to design synapse-neuron modules to cope with the environment resolution. These modules can be employed as building blocks of distributed neural networks. A 4-3-2 network is designed and trained to justify the functionality of the proposed structure. This prototype is simulated and laid out in 0.18 μm CMOS technology.
Keywords :
CMOS analogue integrated circuits; CMOS digital integrated circuits; neural nets; 4-3-2 network; CMOS technology; CVNS DNN; continuous valued number system; environment resolution; noise immunity; prototype CVNS distributed neural network; resistive-type distributed neural networks; synapse-neuron modules design; Biological neural networks; CMOS integrated circuits; Neurons; Noise; Prototypes; Random access memory; Simulation; CVNS digits; Common node; distributed neural network; resolution of environment; sigmoidal neuron; synapse-neuron module; truncated CVNS;
Journal_Title :
Circuits and Systems I: Regular Papers, IEEE Transactions on
DOI :
10.1109/TCSI.2011.2177108
