Enhancing Electronic Nose Performance by Feature Selection Using an Improved Grey Wolf Optimization Based Algorithm

. 2020 Jul 22;20(15):4065. doi: 10.3390/s20154065

Algorithm 2 Adaptive restart GWO

1:
Initialize the number of iterations for optimization $N_{i t e r}$
2:
Initialize the positions of n grey wolves $X_{i}$ , $i = 1, 2, \dots, N$
3:
Calculate the fitness value of each grey wolf
4:
Choose the best three grey wolves as $X_{α}$ , $X_{β}$ , $X_{δ}$ base on there fitness
5:
$t \leftarrow 0$ .
6:
$f i t n e s s_{N} \leftarrow$ the best fitness calculated from initialized wolves
7:
while $t < N_{i t e r} + 1$ do
8:
Update the position of the each wolf using Equation (6)
9:
Update $α$ , A, and C
10:
Update $x_{1}$ , $x_{2}$ , and $x_{3}$ using Equations (12)–(14) or get the position vector of each wolf
11:
Transform each wolf’s position into a binary vector using Equation (11) and $x_{1}$ , $x_{2}$ , $x_{3}$ or Equation (15) and wolf’s original position vector.
12:
Calculate the fitness of each Wolf
13:
$f i t n e s s_{L} \leftarrow f i t n e s s_{N}$
14:
$f i t n e s s_{N} \leftarrow$ best fitness in iteration t
15:
if $f i t n e s s_{N} \geq f i t n e s s_{L}$ then
16:
calculate $N_{r e s t a r t}$ using Equation (17).
17:
select $N_{r e s t a r t}$ wolves randomly from all search agents to reinitialize
18:
end if
19:
Update the first three grey wolves $X_{α}$ , $X_{β}$ , $X_{δ}$ base on fitness
20:
$t \leftarrow t + 1$ .
21:
end while
22:
return $b e s t f i t n e s s, X_{α}$