Adaptive Analog VLSI Signal Processing and Neural Networks

Floating gate transistors;Neural networks;Adaptive filters;VLSI;Analog electronics

Dugger, Jeffery Don ; Electrical and Computer Engineering

University:Georgia Institute of Technology

Department:Electrical and Computer Engineering

关键词: Floating gate transistors;    Neural networks;    Adaptive filters;    VLSI;    Analog electronics;   

Others  :  https://smartech.gatech.edu/bitstream/1853/5294/1/dugger_jeffery_d_200312.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

Research presented in this thesis providesa substantial leap from the study of interestingdevice physics to fully adaptive analog networksand lays a solid foundation for future developmentof large-scale, compact, low-power adaptive parallelanalog computation systems.The investigation described here started withobservation of this potential learning capability and led to the first derivation and characterization ofthe floating-gate pFET correlation learning rule.Starting with two synapses sharing the same error signal,we progressed from phase correlation experimentsthrough correlation experiments involving harmonically related sinusoids,culminating in learning the Fourier series coefficients of a square wave cite{kn:Dugger2000}.Extending these earlier two-input node experiments to the general caseof correlated inputs required dealing with weight decay naturally exhibited by the learning rule.We introduced a source-follower floating-gate synapseas an improvement over our earlier source-degenerated floating-gate synapsein terms of relative weight decay cite{kn:Dugger2004}.A larger network of source-follower floating-gate synapses was fabricatedand an FPGA-controlled testboard was designed and built.This more sophisticated system provides an excellentframework for exploring applications to multi-input, multi-nodeadaptive filtering applications.Adaptive channel equalization provideda practical test-case illustrating the useof these adaptive systems in solving real-world problems.The same system could easily be applied to noise and echo cancellation in communication systems and system identification tasks inoptimal control problems.We envision the commercialization of these adaptive analog VLSIsystems as practical products within a couple of years.