Back to articles
Volume: 28 | Article ID: art00017
Refractory Neural Nets and Vision – A Deeper Look
  DOI :  10.2352/ISSN.2470-1173.2016.15.IPAS-188  Published OnlineFebruary 2016

In an earlier paper, it was shown that the neuron’s refractory period (the period of time after the neuron has fired before it can fire again) can serve as a short term local memory. In particular, if an array of refractory neurons (the retina) trains over an image, is then offset, the trained pixel comparisons to the offset pixels are done globally across the entire array. The refractory period is biologically based, and so is the offset; the offset is done by ocular microtremors. Together, they provide a tool that can do grey scale boundary and texture segmentation. This paper significantly extends the capabilities of refractory neural nets by pointing out that refractory neurons can be arranged into XOR gates. We have a ‘pixel-predictor’ and use an XOR gate to compare the sensing of a pixel to the prediction for that pixel. If the two are the same, then nothing comes up from the gate. If they are different, then a signal comes out and a modification is made to the pixel-predictor. These predictors can be done at multiple levels of coarseness which effectively give us a multilayer classifier, i.e., a deep learning capability.

Subject Areas :
Views 3
Downloads 0
 articleview.views 3
 articleview.downloads 0
  Cite this article 

Thomas C Fall, "Refractory Neural Nets and Vision – A Deeper Lookin Proc. IS&T Int’l. Symp. on Electronic Imaging: Image Processing: Algorithms and Systems XIV,  2016,

 Copy citation
  Copyright statement 
Copyright © Society for Imaging Science and Technology 2016
Electronic Imaging
Society for Imaging Science and Technology
7003 Kilworth Lane, Springfield, VA 22151 USA