Brain Storm Optimization Algorithms in Swarm Intelligence

Brain storm optimization (BSO) is an evolutionary computation technique inspired by collective human ideation processes, in which candidate solutions are treated as “ideas” that cluster and evolve in a population. As a branch of swarm intelligence, BSO harnesses decentralised interaction among individuals to explore complex search spaces and balance global exploration with local exploitation. Core mechanisms involve clustering the population into groups, generating new solutions through perturbations within or between clusters, and employing adaptive operators to prevent premature convergence. Recent advances have focused on improving convergence speed, solution accuracy and robustness through hybrid strategies, theoretical analysis of time complexity and the integration of adaptive learning methods. This has broadened the practical applications of BSO in engineering design, neural network training, combinatorial optimisation and real-time decision support systems worldwide.

Research from Nature Portfolio

A novel global-best brain storm optimization algorithm has been designed to address slow convergence and susceptibility to local optima. By introducing a chaotic difference step strategy based on multiple chaotic maps, the algorithm expands the search space and enhances diversity in the population. Concurrently, an opposition-based learning mechanism generates opposing candidate solutions to accelerate escapes from local minima. Comparative experiments on benchmark suites demonstrate that this enhanced global-best variant significantly outperforms the original algorithm and other recent improvements in both convergence rate and solution quality when tackling multidimensional continuous optimisation problems.

Research from all publishers

Running-time analysis of single-individual BSO variants has provided theoretical bounds on the expected time to reach target solutions. By modelling average gain per iteration, researchers have shown linear time complexity in equal-coefficient linear functions and quantified the impact of different mutation operators on convergence speed. These insights offer a formal foundation for algorithm design and parameter tuning in high-dimensional settings.

A hybridisation between BSO and a chaotic accelerated particle swarm optimiser has capitalised on the complementary strengths of both methods. Initial exploration is driven by BSO-based clustering, followed by rapid local exploitation via a minimalistic accelerated PSO. Benchmark studies reveal that this hybrid yields superior computational efficiency and solution accuracy across diverse unimodal and multimodal test functions compared with either algorithm alone.

An enhanced BSO variant integrates a modified Nelder–Mead simplex approach with an elite learning mechanism to guide population evolution. The Nelder–Mead component introduces a directional search operator that refines promising regions, while elite learning preserves high-quality solutions. A reinitialisation strategy further mitigates stagnation. Empirical results on standard benchmark suites and real-world prediction problems demonstrate marked improvements in convergence stability and optimisation precision.

Brain Storm Optimization Algorithms in Swarm Intelligence publication trend

The graph below shows the total number of articles in brain storm optimization algorithms in swarm intelligence across all publications each year (not limited to Nature Index journals).

Technical terms

Brain Storm Optimization (BSO): A swarm intelligence algorithm that models problem-solving as clustering and recombination of “ideas” within a population of candidate solutions.

Swarm Intelligence: A class of decentralised algorithms inspired by collective behaviour in natural systems, where simple agents interact to achieve complex global optimisation.

Clustering Strategy: The process of grouping individuals in BSO into different clusters to control local and global search movements.

Exploration–Exploitation Trade-off: The balance between sampling new regions of the search space (exploration) and refining existing promising solutions (exploitation).

Opposition-Based Learning: A technique that simultaneously considers current candidate solutions and their “opposites” to enhance diversity and accelerate convergence.

References

1. Global-best brain storm optimization algorithm based on chaotic difference step and opposition-based learning. Scientific Reports (2024).
2. Running-Time Analysis of Brain Storm Optimization Based on Average Gain Model. Biomimetics (2024).
3. A Brain Storm and Chaotic Accelerated Particle Swarm Optimization Hybridization. Algorithms (2023).
4. Enhanced Brain Storm Optimization Algorithm Based on Modified Nelder–Mead and Elite Learning Mechanism. Mathematics (2022).