Table 3 Comparison of Core Mechanisms Between StMA and Mainstream Algorithms.
From: A sterna migration algorithm-based efficient bionic engineering optimization algorithm
Algorithm | Biological behavior model | Global/local cooperation mechanism | Means of maintaining diversity | Adaptive control parameters | Termination condition design |
|---|---|---|---|---|---|
StMA | Three-stage migration modeling of the common tern | Spatial compression and perturbation feedback | Distribution density and variance control | Sigmoid Function& β(t), η(t) | Multiple termination criteria (stagnation/diversity/maximum iterations) |
CFDE | Competitive framework differential evolution | Strategic individual selection and competition | Competitive elimination mechanism | Dynamic scaling factors | Maximum number of iterations |
THRO | Tianji’s horse racing strategy | Dynamic individual matching strategy | Greedy strategy for maximizing benefits | Racing intensity parameters | Maximum number of generations |
CIA | Crested ibis foraging behavior | Group coordination and information interaction | Exploration strategies based on foraging success/failure | Foraging efficiency factors | Maximum number of iterations |
GFA | Gyro fireworks explosion mechanism | Multi-stage multiple search strategies | Spiral rotation and aggregation contraction | Gyro effect parameters | Maximum number of steps |
FBO | Frigate bird soaring behavior | Two-stage search strategy | Random search direction and radius | Soaring altitude control | Maximum number of iterations |
FCO | Fishing cat hunting strategy | Four-phase search strategy | Lurking, perceiving, rapid diving, precise trapping | Hunting success rate factors | Maximum number of generations |
HBO | Corporate rank hierarchy structure | Fitness-based hierarchical arrangement | Intra-hierarchy competition and promotion mechanism | Heap size and depth control | Maximum number of iterations |
SMP-JaQA | No biological metaphor is employed | A self-adaptive multi-subpopulation framework is adopted | Multiple subpopulations are naturally distributed across the search space | Apart from population size and the maximum number of evaluations, the algorithm contains no user-tunable parameters | Max-NFEs |
EW-DHOA | No biological metaphor is employed | Two sub-populations run in parallel | Heterogeneous operators edge-difference truncation natural spatial distribution of two sub-populations | No manual parameter tuning is required throughout the entire process | Max-NFEs |
