As mutation operator we have chosen the non-uniform mutation with parameter $ b=5$ [Mic92] as its dynamical nature makes it very suitable for a wide variety of problems [HL00].

The individuals $ \beta^m$ generated by this mutation are obtained as follows:

$\displaystyle \beta^m_i=\left\{ \begin{array}{lll} \beta_i + \triangle(t,b_i-\b...
...u = 0 \\ \beta_i - \triangle(t,\beta_i-a_i) & si & \tau = 1 \end{array} \right.$ (12)


$\displaystyle \triangle(t,y)=y(1-r^{(1-{t \over g_{max}})^b})$ (13)

where $ t$ is the generation, $ g_{max}$ is the maximum number of generations, $ \tau$ is a random value, $ \tau \in \{0,1\}$, $ r$ is a random number in the interval $ [0, 1]$ and $ b$ is a parameter that determines the degree of dependence of the mutation with regards to the number of iterations. Equation 13 gives values in the interval $ [0, y]$. The probability of obtaining a value near 0 increases as the algorithm progresses. This operator performs a uniform search in the initial stages of the evolution, and a very localized search in the final stages.

Domingo 2005-07-11