DocumentCode :
3118763
Title :
A photonic neuromorphic computational primitive for complex high bandwidth signal processing
Author :
Rosenbluth, David ; Olivieri, Marc ; Kravtsov, Konstantin ; Fok, Mable P. ; Prucnal, PAul R.
Author_Institution :
Adv. Technol. Lab., Lockheed Martin, Cherry Hill, NJ, USA
fYear :
2010
fDate :
21-23 Sept. 2010
Firstpage :
49
Lastpage :
50
Abstract :
The spike processing device is a full all-optical realization of the integrate-and-fire neuron model. While the fastest timescale on which biological spiking neurons operate is on the order of milliseconds, the experimentally demonstrated integrate-and-fire device operates on pico-second width pulses, and has an integration time constant of 180 ps, tunable in the range from 100 to 300 ps. Reconfiguration of device parameters enables it to perform a wide variety of signal processing and decision operations. An example of decisionmaking based on the number of input pulses is studied experimentally by changing the threshold value of the neuron. Its analog properties make it suited to efficient signal processing applications, while its digital properties make it possible to implement complex computations without excessive noise accumulation.
Keywords :
neural nets; signal processing; biological spiking neuron; full all-optical realization; integrate-and-fire device; integrate-and-fire neuron model; photonic neuromorphic computational primitive; signal processing; spike processing device; time 100 ps to 300 ps; Bandwidth; Neurons; Noise; Optical fibers; Optical signal processing; Performance evaluation; Photonics;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Avionics Fiber-Optics and Photonics Technology Conference (AVFOP), 2010 IEEE
Conference_Location :
Denver, CO
Print_ISBN :
978-1-4244-5312-2
Electronic_ISBN :
978-1-4244-5313-9
Type :
conf
DOI :
10.1109/AVFOP.2010.5637456
Filename :
5637456
Link To Document :
بازگشت