• DocumentCode
    2442800
  • Title

    A Variable Step-Size Optimization Algorithm Based on a Special Numerical Method to Solve the System of Nonlinear Equations

  • Author

    Zhou You-hang ; Lin Jie-huang ; Zhang Jian-xun

  • Author_Institution
    Sch. of Mech. Eng., Xiangtan Univ., Xiangtan
  • fYear
    2008
  • fDate
    27-28 Dec. 2008
  • Firstpage
    459
  • Lastpage
    463
  • Abstract
    After being studied the expression of figure-counting methods, a special numerical method is presented to approach any given value. With a gradual decrease weight of each digit, under the given error, any given data could be described by the permutation and combination of digits (-1, 0, 1). Based on this special numerical method, a new variable step-size optimization algorithm is designed to solve the system of nonlinear equations. According to the idea of this algorithm, the gradual decrease weight is considered as a variable step, and the change of each variable should be tested with the rule of ldquoForward, backward or maintain a steprdquo to approach the optimization objective, and the choice principle of the related variables has been discussed. The computer simulated results show that this algorithm can overcome the problems such as sensitivity to initial values, poor convergence reliability, and it is easy and practical to design.
  • Keywords
    nonlinear equations; optimisation; figure-counting method; nonlinear equation; numerical method; permutation; variable step-size optimization algorithm; Algorithm design and analysis; Artificial intelligence; Artificial neural networks; Genetic algorithms; Mechanical engineering; Newton method; Nonlinear equations; Optimization methods; Simulated annealing; Testing; Optimization; Special numerical method; System of nonlinear equations; Variable step size;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Modelling, Simulation and Optimization, 2008. WMSO '08. International Workshop on
  • Conference_Location
    Hong Kong
  • Print_ISBN
    978-0-7695-3484-8
  • Type

    conf

  • DOI
    10.1109/WMSO.2008.16
  • Filename
    4757049