Neural Networks trained with the Belief Propagation Inspired (BPI) algorithm are able to learn a number of associations close to the theoretical limit in time that is sublinear in the number of input. Using binary synapses, implemented by a memristor, a single layer perceptron with BPI has been proposed. It well know that perceptrons with step function type nonlinearity can be implemented by a suitable class of Cellular Neural/Nonlinear Networks. This paper aims to present a statistical analysis on the learning efficiency of Memristor-based Cellular Nonlinear Networks (M-CNNs) with Belief Propagation Inspired (BPI) algorithm. Monte Carlo simulations permit to assess that the learning efficiency of M-CNNs with BPI is not regardless of the input signals given to train the perceptron.
[1]
Josef A. Nossek,et al.
An equivalence between multi-layer perceptrons with step function type nonlinearity and a class of cellular neural networks
,
1991,
1991., IEEE International Sympoisum on Circuits and Systems.
[2]
L. Chua.
Memristor-The missing circuit element
,
1971
.
[3]
Jacques-Olivier Klein,et al.
Robust neural logic block (NLB) based on memristor crossbar array
,
2011,
2011 IEEE/ACM International Symposium on Nanoscale Architectures.
[4]
Gianfranco Genta,et al.
Features and performance of some outlier detection methods
,
2011
.