Thoughts on Particle Swarm Optimization Algorithms

I. Particle SwarmAlgorithmDifferences from real models:

The particle swarm optimization algorithm is designed based on the bird's feeding model. The model differs from the real bird's feeding process.

1. discretization and continuity are different:When a bird is flying, it is a continuous flight, and all the "lines" on the road constitute a position for food. The motion of particles in the model is discrete, it is a position that consists of "points. If the particle speed in the model is too large, it may "fly over" the target location. If the speed is too small, it may lead to slow convergence. Therefore, setting the proper speed is the key to the particle swarm algorithm.

2. Differences between verifiable and non-verifiable:In the process of feeding birds, it can be determined whether the true value is found, but there are two situations in the particle swarm optimization algorithm:

① Verifiable:In the process of solving the zero value of the monotonic function, the obtained solution x can be substituted into F (x) to verify whether f (x) = 0 exists;

② Not verifiable:When solving the minimum value of a function, because it is searched in the solution space, there is no way to verify whether it is the minimum value, only the minimum value of the particle flight "discrete point set" can be found.

③ Improvement based on specific problems:It can be verified that if the value to be converged is incorrect, the problem can be re-searched for divergence, and the next random structure is set to generate the exclusion component of the minimum point; an unverifiable problem can only be caused by a relatively small moving inertia component so that the speed is not too high. This ensures convergence to the minimum value at a time, or multiple operations.Program.