DocumentCode :
948926
Title :
Gaussian activation functions using Markov chains
Author :
Card, Howard C. ; McNeill, Dean K.
Author_Institution :
Dept. of Electr. & Comput. Eng., Manitoba Univ., Winnipeg, Man., Canada
Volume :
13
Issue :
6
fYear :
2002
fDate :
11/1/2002 12:00:00 AM
Firstpage :
1465
Lastpage :
1471
Abstract :
We extend, in two major ways, earlier work in which sigmoidal neural nonlinearities were implemented using stochastic counters. 1) We define the signal to noise limitations of unipolar and bipolar stochastic arithmetic and signal processing. 2) We generalize the use of stochastic counters to include neural transfer functions employed in Gaussian mixture models. The hardware advantages of (nonlinear) stochastic signal processing (SSP) may be offset by increased processing time; we quantify these issues. The ability to realize accurate Gaussian activation functions for neurons in pulsed digital networks using simple hardware with stochastic signals is also analyzed quantitatively.
Keywords :
Markov processes; neural chips; probability; signal processing; transfer functions; Gaussian activation functions; Gaussian mixture models; Markov chains; bipolar stochastic arithmetic; neural networks; neural transfer functions; pulsed digital networks; sigmoidal neural nonlinearities; signal to noise limitations; stochastic counters; stochastic signal processing; unipolar stochastic arithmetic; Arithmetic; Counting circuits; Gaussian noise; Neural network hardware; Neurons; Signal analysis; Signal processing; Stochastic processes; Stochastic resonance; Transfer functions;
fLanguage :
English
Journal_Title :
Neural Networks, IEEE Transactions on
Publisher :
ieee
ISSN :
1045-9227
Type :
jour
DOI :
10.1109/TNN.2002.804285
Filename :
1058080
Link To Document :
بازگشت