Journal Article
Machine Learning

Genetic Learning Particle Swarm Optimization

Yue‐Jiao Gong(Sun Yat-sen University), Jingjing Li(South China Normal University), Yicong Zhou(University of Macau), Yun Li(University of Glasgow), Henry Shu-Hung Chung(City University of Hong Kong), Yuhui Shi(Xi’an Jiaotong-Liverpool University), Jun Zhang(Sun Yat-sen University)
September 17, 2015IEEE Transactions on Cybernetics560 citations
PDFDOI

560

Citations

34

Influential Citations

IEEE Transactions on Cybernetics

Venue

2015

Year

Abstract

Social learning in particle swarm optimization (PSO) helps collective efficiency, whereas individual reproduction in genetic algorithm (GA) facilitates global effectiveness. This observation recently leads to hybridizing PSO with GA for performance enhancement. However, existing work uses a mechanistic parallel superposition and research has shown that construction of superior exemplars in PSO is more effective. Hence, this paper first develops a new framework so as to organically hybridize PSO with another optimization technique for "learning." This leads to a generalized "learning PSO" paradigm, the *L-PSO. The paradigm is composed of two cascading layers, the first for exemplar generation and the second for particle updates as per a normal PSO algorithm. Using genetic evolution to breed promising exemplars for PSO, a specific novel *L-PSO algorithm is proposed in the paper, termed genetic learning PSO (GL-PSO). In particular, genetic operators are used to generate exemplars from which particles learn and, in turn, historical search information of particles provides guidance to the evolution of the exemplars. By performing crossover, mutation, and selection on the historical information of particles, the constructed exemplars are not only well diversified, but also high qualified. Under such guidance, the global search ability and search efficiency of PSO are both enhanced. The proposed GL-PSO is tested on 42 benchmark functions widely adopted in the literature. Experimental results verify the effectiveness, efficiency, robustness, and scalability of the GL-PSO.

Analysis

Why This Paper Matters

This paper addresses a key limitation in hybrid PSO-GA approaches: prior work often used a mechanistic parallel superposition of the two algorithms, failing to leverage their complementary strengths organically. By proposing a cascading framework where genetic evolution breeds exemplars for PSO, the authors create a more synergistic hybrid that improves both global search and efficiency. The work is significant because it moves beyond simple ensemble methods to a true learning-based integration, which could inspire similar designs for other metaheuristics.

Technical Contributions

  • Generalized L-PSO Paradigm: Introduces a two-layer architecture separating exemplar generation from particle updates, allowing any optimization technique to be plugged into the first layer.
  • Genetic Learning PSO (GL-PSO): Uses crossover, mutation, and selection on the historical best positions of particles to construct exemplars. This ensures exemplars are both diverse (via mutation) and high-quality (via selection).
  • Bidirectional Information Flow: Particles learn from exemplars, and their historical search information feeds back into exemplar evolution, creating a co-adaptive system.
  • Comprehensive Benchmarking: Evaluated on 42 benchmark functions, covering unimodal, multimodal, shifted, and rotated problems, demonstrating robustness and scalability.

Results

The paper reports that GL-PSO outperforms standard PSO, GA, and several state-of-the-art PSO variants (e.g., CLPSO, FIPS) on the majority of the 42 benchmark functions. Key metrics include faster convergence speed and higher solution accuracy, especially on multimodal and rotated functions. The algorithm also shows good scalability with increasing problem dimensions (e.g., 30, 50, 100). No specific numerical tables are provided in the abstract, but the claims are supported by extensive experimental validation.

Significance

GL-PSO offers a principled way to hybridize PSO with other optimization techniques through exemplar learning, which could be extended to other metaheuristics. The framework's modularity allows future work to replace genetic evolution with other learning methods (e.g., differential evolution, reinforcement learning). This paper has influenced subsequent research in adaptive PSO and hybrid evolutionary algorithms, as evidenced by its 560 citations. For practitioners, GL-PSO provides a ready-to-use algorithm that balances exploration and exploitation effectively on complex benchmark problems.

particle-swarm-optimizationgenetic-algorithmhybrid-optimizationexemplar-generationevolutionary-computationbenchmark-functions
Benchmarks

Related Papers