p

p

p

h other to arrive at the global optimal solution. The objective

has three saddle points with three objective function values,

ଵ^{, ݕ}ଶ^{, and ݕ}ଷ^{. One of them (ݕ}ଵ^{) is the global optimal (minimum)}

There are five initialised model parameters, i.e., ݔଵ, ݔଶ, ݔଷ, ݔସ

The Newton’s method will generate five models. Each of the

pdates its model parameters leading to these three saddle points.

e saddle point of ݕଵ is reached, the other two saddle points are

way. Therefore, it is less likely to miss the global minimum in

Fig. 8.1. The demo of the competition between multiple candidates.

(a) (b)

e demo of evolutionary computation. The dots stand for the original candidates.

and for the evolved candidates. (a) Before evolution. (b) After evolution.

ose the limit of the candidate size is three (not five) for the

shown in Figure 8.1. As shown in Figure 8.2(a), none of these

didates can help reach the global minimum ݕଵ. The evolutionary

ion can use an evolutionary operator to generate new candidates

e three candidates. The new candidates may lead to some better

ints so that the Newton’s method can be used to reach the global