Design of RF MEMS based oscillatory neural network for ultra high speed associative memories


Baghelani M., EBRAHIMI A., Ghavifekr H. B.

Neural Processing Letters, cilt.40, sa.1, ss.93-102, 2014 (SCI-Expanded, Scopus)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 40 Sayı: 1
  • Basım Tarihi: 2014
  • Doi Numarası: 10.1007/s11063-013-9312-y
  • Dergi Adı: Neural Processing Letters
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Sayfa Sayıları: ss.93-102
  • Anahtar Kelimeler: Dynamic neuron, RF MEMS, Oscillatory neural network, Resonator
  • Atatürk Üniversitesi Adresli: Hayır

Özet

Microelectromechanical systems are utilized alongside with transistor amplifiers and resistive connections for implementing of oscillatory associative memories. Phase locking is studied in such a network and all requirements of the circuit level implementation are satisfied. A very high gain trans-impedance amplifier operating in 1 GHz in addition to a novel automatic amplitude control circuit is employed to remove amplitude dynamics of the system. Requiring resonator characteristics are extracted and calculated as well. A new method for initialization of the network is proposed. Each neuron consumes 1.08 mW from a 1.8 V power supply. The convergence time of a typical network trained by Hebbian rule is less than 1.5 ns which results in an ultra high speed analog signal processing system. © 2013 Springer Science+Business Media New York.