A Constructive Data Classification Version of the Particle Swarm Optimization Algorithm

Pseudocode 2

Constructive particle swarm clustering.

1. Procedure  [X, V, P, g, 𝜔, PLABELS ] = cPSC (Y, 𝑚, 𝜔, 𝑣_max, 𝜀, CLABELS)

2. Y//dataset

3. initialize X//initialize at random only one particle 𝑥_𝑖 ∈ [0,1]

4. initialize V//initialize at random, 𝑣_𝑖 ∈ [−𝑣_max, 𝑣_max]

5. initialize dist

6. 𝑡=1

7. while stopping criterion is not met

8.  for  𝑗=1 to 𝑛//for each input datum

9.   for 𝑖=1 to 𝑚//for each particle

10.     dist (𝑖) = distance (𝑥_𝑖,𝑦^𝑗)

11.    end for

12.   I = index (min (dist))

13.    [ PLABELS ] = label (𝑥_𝐼, CLABELS (𝑦^𝑗))//predicted label

14.   if distance (𝑥_𝐼,𝑦^𝑗) < distance (𝑝^𝑗_𝐼,𝑦^𝑗)

15.     𝑝^𝑗_𝐼 = 𝑥_𝐼

16.   end if

17.   if distance (𝑥_𝐼,𝑦^𝑗) < distance (𝑔^𝑗,𝑦^𝑗)

18.     𝑔^𝑗=𝑥_𝐼

19.   end if

20.   𝑣_𝐼(𝑡+1) = 𝜔∗𝑣_𝐼(𝑡) + φ ₁⊗(𝑝^𝑗_𝐼(𝑡)−𝑥_𝐼(𝑡)) + 𝜑₂⊗ (𝑔^𝑗(𝑡)−𝑥_𝐼(𝑡)) + 𝜑₃⊗ (𝑦^𝑗−𝑥_𝑖(𝑡))

21.   𝑣_𝐼 ∈[−𝑣_max, 𝑣_max]

22.   𝑥_I (𝑡+1) = 𝑥_𝐼(𝑡) + 𝑣_𝐼 (𝑡+1)

23.   𝑥_𝐼 ∈[0,1]

24.  end for

25.  if mod (𝑡,2)==0

26.   Eliminate particles from the swarm if necessary

27.   Test the stopping criterion

28.   Clone particles if necessary

29.  end if

30.  𝑡=𝑡+1

31.  𝜔=0.95∗𝜔

32. end while

33. end procedure