Cooperative Evolutionary Optimization for Large-Scale Problems

Cooperative evolutionary optimisation has emerged as a powerful paradigm for addressing large-scale optimisation problems characterised by thousands or even millions of decision variables. At its core, this approach divides a high-dimensional problem into smaller, more tractable subcomponents, which are then evolved in parallel or in an interleaved manner. By exploiting variable interactions through careful decomposition schemes, cooperative frameworks can dramatically reduce computational complexity and mitigate the curse of dimensionality. These methods find application in diverse domains, from tuning parameters in deep neural networks and designing resilient supply chains to scheduling complex cloud workflows and optimising aerodynamic shapes. Recent advances have focused on adaptive decomposition, dynamic group learning and targeted modification of bottleneck variables, enhancing convergence speed and solution quality. Moreover, hybrid strategies that combine population-based metaheuristics with machine-learning models or surrogate approximations have extended the reach of cooperative coevolution to real-time and resource-constrained settings. The global significance of these techniques lies in their capacity to deliver near-optimal solutions for grand challenges in engineering, logistics and scientific computing, where traditional monolithic algorithms falter.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Building on conventional differential grouping, the dual differential grouping (DDG) method enhances variable decomposition by detecting both additive and multiplicative separability. This enables cooperative coevolutionary frameworks to tackle a broader class of large-scale functions and has demonstrated superior performance on benchmark suites and neural network parameter tuning. A gene-targeting differential evolution (GTDE) algorithm introduces a biologically inspired operator that probabilistically identifies and modifies bottleneck dimensions in the current best solution. By breaking through these critical constraints, GTDE accelerates convergence and outperforms state-of-the-art competitors on competition test suites. Another notable advance is the adaptive granularity learning distributed particle swarm optimisation (AGLDPSO), which partitions the swarm into subpopulations and employs locality-sensitive hashing and logistic regression to adjust subpopulation size dynamically. This strategy balances exploration across vast search spaces with exploitation of promising regions, yielding robust performance on both classical benchmarks and real-world scheduling tasks. Together, these contributions underscore the trend towards more flexible decomposition, targeted search and adaptive collaboration among evolutionary agents.

Cooperative Evolutionary Optimization for Large-Scale Problems publication trend

The graph below shows the total number of articles in cooperative evolutionary optimization for large-scale problems across all publications each year (not limited to Nature Index journals).

Technical terms

Cooperative coevolution: A framework that decomposes a high-dimensional problem into subcomponents optimised by separate evolutionary processes that share information.

Variable decomposition: The process of partitioning decision variables into groups based on their interactions or separability properties.

Differential grouping: A technique for detecting dependencies among variables by analysing differences in function evaluations under perturbations.

Particle swarm optimisation: A population-based metaheuristic in which candidate solutions (particles) move through the search space by learning from individual and collective experience.

Bottleneck dimension: A variable or subset of variables whose optimisation significantly restricts overall solution quality and convergence speed.

References

1. Dual Differential Grouping: A More General Decomposition Method for Large-Scale Optimization. IEEE Transactions on Cybernetics (2023).
2. Dynamic Group Learning Distributed Particle Swarm Optimization for Large-Scale Optimization and Its Application in Cloud Workflow Scheduling. IEEE Transactions on Cybernetics (2019).
3. Adaptive Granularity Learning Distributed Particle Swarm Optimization for Large-Scale Optimization. IEEE Transactions on Cybernetics (2021).
4. Gene Targeting Differential Evolution: A Simple and Efficient Method for Large-Scale Optimization. IEEE Transactions on Evolutionary Computation (2022).