Enhanced particle swarm optimization based on principal component analysis and line search

- A new information accumulation mechanism with principal component analysis is presented.
- Principal component analysis is for promising directions and line search is for enhanced search.
- Principal component analysis is analyzed and illustrated how to be utilized.
- Extensive comparisons with pPSA, PCPSO, CLPSO, OLPSO, and AIWPSO.

Particle swarm optimization (PSO) guides its search direction by a linear learning strategy, in which each particle updates its velocity through a linear combination among its present status, historical best experience and the swarm best experience. The current position of each particle can be seen as a velocity accumulator. Such a storage strategy is easy to achieve, however, it is inefficient when searching in a complex space and has a great restriction on the achieved heuristic information for the promising solutions. Therefore, a new PSO searching mechanism (PCA-PSO) is proposed based on principal component analysis (PCA) and Line Search (LS), in which PCA is mainly used to efficiently mine population information for the promising principal component directions and then LS strategy is utilized on them. PCA-PSO can inherit most of the velocity information of all the particles to guide them to the most promising directions, which have great difference in learning mechanism with usual PSOs. Experimental results and extensive comparisons with hybrid PSOs, pPSA, PCPSO, CLPSO, GL-25, and CoDE show that PCA-PSO consistently and significantly outperforms some PSO variants and is competitive for other state-of-the-art algorithms.