Title :
Optoelectronic devices in optoelectronic signal processors
Author_Institution :
Connecticut Univ., Storrs, CT, USA
Abstract :
A neural network is one architecture for an optoelectronic signal processor which operates in a dynamic parallel mode. The basic node is a circuit which has multiple inputs and a single output which is in general fanned out to several next stage inputs. A key feature is that an adjustable synaptic weight is applied to each input. Then the weighted inputs are summed and the sum is thresholded to produce a single on/off output signal. A neural network is made up of many such neurodes with a massive number of interconnections. Because of the interconnectivity requirements, it is attractive to use optical signals as the input and output signals to the neurode. Because of the weighting, summing and thresholding requirement it is attractive to use electronic device implementations. Therefore the neural signal processor is a natural candidate for optoelectronic device technology
Keywords :
integrated optoelectronics; optical information processing; optical interconnections; optical neural nets; optoelectronic devices; adjustable synaptic weight; basic node; dynamic parallel mode; electronic device implementations; interconnectivity requirements; multiple inputs; neural signal processor; neurodes; optical interconnections; optical signals; optoelectronic device technology; optoelectronic devices; optoelectronic signal processor; optoelectronic signal processors; output signal; stage inputs; summing and thresholding requirement; weighted inputs; Circuits; Detectors; HEMTs; MODFETs; Neural networks; Optical control; Optoelectronic devices; Signal processing; Smart pixels; Voltage control;
Conference_Titel :
Lasers and Electro-Optics Society Annual Meeting, 1997. LEOS '97 10th Annual Meeting. Conference Proceedings., IEEE
Conference_Location :
San Francisco, CA
Print_ISBN :
0-7803-3895-2
DOI :
10.1109/LEOS.1997.645486