• DocumentCode
    3140071
  • Title

    Can we work around Complex Numerical Methods? An Insight

  • Author

    Chandana, Sandeep ; Mayorga, Rene V.

  • Author_Institution
    Fac. of Eng., Regina Univ., Sask.
  • fYear
    2006
  • fDate
    38838
  • Firstpage
    2120
  • Lastpage
    2123
  • Abstract
    This paper aims to provide an insight into the concept of doing away with complex numerical methods; the authors propose a generalized neural approximator based replacement for numerical methods. The presented work outlines a non-conventional (and non-numerical) technique of analyzing the stability thresholds for one type of self-sustaining oscillations. The revered van der Pol equation has been studied in this work with an intention to predict an accurate stability criterion under the effect of a set of parameters and estimation techniques with the help of a novel neural network based approximation module. Such a system, for the presented case, boasts of generalized approximation capability and of a relatively flexible-easy-to-alter method
  • Keywords
    inference mechanisms; mathematics computing; neural nets; numerical stability; uncertainty handling; complex numerical methods; estimation techniques; neural approximator based replacement; neural network based approximation module; nonconventional technique; self-sustaining oscillations; stability thresholds; Computer networks; Differential equations; Electron tubes; Laboratories; Modems; Neural networks; Parameter estimation; Resistors; Stability analysis; Stability criteria; Neural Networks; Numerical Methods; Self-Sustaining Oscillations; Stability Threshold;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Electrical and Computer Engineering, 2006. CCECE '06. Canadian Conference on
  • Conference_Location
    Ottawa, Ont.
  • Print_ISBN
    1-4244-0038-4
  • Electronic_ISBN
    1-4244-0038-4
  • Type

    conf

  • DOI
    10.1109/CCECE.2006.277425
  • Filename
    4054853