Modified discrete particle swarm optimization algorithm based on inver-over operator

Though the discrete particle swarm optimization (DPSO) can make the best of the local and global optima of particles, it converges slowly with low precision. The Guo Tao algorithm converges with fast high precision, but it is blindfold to learn from the other particles. A modified discrete particle swarm optimization algorithm is presented based on the inver-over operator (IDPSO). To prevent premature convergence, the local sub-optimum particle swarm is introduced into IDPSO. Particles learn from the particles in the local sub-optimum particle swarm instead of their local optima. Three new parameters are introduced into IDPSO. Learning selection probability is introduced to select the particle to be learned. A generation threshold is introduced to define when to learn from the global particle. Local sub-optimum particle swarm ratio is introduced to define the size of the sub-optimum particle swarm. Selecting principles of these parameters is detailed discussed and the general reference scopes are given. Experiments are carried out on the traveling salesman problem and the results show that the modified IDPSO achieves good results compared with the Guo Tao algorithm and the general DPSO. The proposed algorithm improves both the convergence speed and solution precision.