Static Nonlinear Equation Solving

Yunong Zhang,Lin Xiao,Zhengli Xiao,Mingzhi Mao

2016-01-01

Abstract:A special class of neural dynamics (termed zeroing dynamics, ZD) is generalized, developed and investigated in this chapter for online solution of static nonlinear equation f (x) = 0 mostly for comparison and connection purposes. The gradient dynamics (GD) is thus also developed and exploited for solving online such a nonlinear equation. Conventionally and geometrically speaking, the GD model evolves along the surface descent direction (specifically, the tangent direction) of the square-based energy-function curve; but, how does the ZD model evolve? Trying to answer this question, the corresponding discrete-time ZD (DTZD) model is then developed and investigated. In terms of the constant nonlinear equation solving, Newton iteration is found once more to be a special case of the DTZD model. It is also discovered that, if a nonlinear equation possesses a local minimum point, the neural state of the CTZD model, starting from some initial value close to it, may move toward the local minimum point and then stop with warning information. In contrast, the neural state of the GD model falls into the local minimum point (with no warning). Furthermore, inspired by Wu's work, the improved CTZD models are proposed by defining two modified error functions, which handle such a local-minimum problem effectively.