A hybrid differential evolution particle swarm optimization algorithm, known as MDE-DPSO, promises to resolve longstanding issues associated with optimization algorithms such as premature convergence to local optima.
Particle Swarm Optimization (PSO) and Differential Evolution (DE) algorithms are widely utilized for solving various optimization challenges. While PSO is known for its simplicity and fast convergence, it often falls short, particularly when the datasets involve multiple peaks or complex landscapes, leading to what is termed 'premature convergence'. Addressing this, the research team spearheaded by Xu H. has developed MDE-DPSO, which integrates dynamic strategies to bolster the searching capability of particles and improve their performance across various benchmarks.
The core of MDE-DPSO lies in its innovative approach to adjusting inertia weights and acceleration coefficients, which enhances the algorithm's adaptability to different optimization stages. "Our algorithm employs a parametric strategy incorporating new inertia weights and acceleration coefficients to more effectively balance the ability of global and local search," state the authors. This approach enables particles to dynamically widen their search range, helping them escape from local optima.
One of the standout features of MDE-DPSO is its hybridization of mutation strategies borrowed from differential evolution. By utilizing both DE/rand/1 and DE/current-to-pbest/1 strategies, MDE-DPSO can adapt its search behavior based on the performance of each particle, ensuring greater diversity across solutions and bolstering the chances of identifying optimal ones.
To validate its efficacy, MDE-DPSO was rigorously tested against 100 benchmark functions from recognized optimization suites, including CEC2013, CEC2014, CEC2017, and CEC2022. The experimental results reveal the algorithm's exceptional performance, marking substantial improvements compared to fifteen other competitive algorithms.
"...the integration of dynamic strategies and mutation operators enhances the exploration capability of MDE-DPSO," the authors conclude, reflecting the algorithm's ability to navigate complex landscapes more efficiently than its predecessors. Notable performance gains across 264 evaluations underline MDE-DPSO's competitive edge, with results indicating its viability as a more effective optimization tool.
The MDE-DPSO algorithm introduces significant improvements to optimization processes, making it particularly relevant for real-world applications where complex, multi-dimensional data sets are prevalent. Future research is expected to explore its application potential across diverse fields, from engineering optimization to path planning challenges.