The Ivy Optimization Algorithm (IVYA) represents an innovative nature-inspired metaheuristic derived from adaptive climbing and spreading behaviors of ivy plants. However, IVYA encounters significant limitations in population distribution quality and local exploitation effectiveness, resulting in premature convergence when navigating complex high-dimensional multimodal landscapes. To address these constraints, this study introduces the Adaptive Quadratic Memory-based Ivy Optimization (AQMIVY) algorithm through three synergistic mechanisms: Superior Point-Based Initialization Strategy (SPIS), Adaptive Quadratic Interpolation Strategy (AQI), and Memory-Guided Dual Attractor Dynamics (MGDAD). SPIS elevates the population quality through deterministic Good Point Set generation combined with Oppositional-Based Learning, ensuring uniform coverage with mathematically guaranteed low discrepancy distribution. AQI introduces intelligent refinement through dual quadratic interpolation models, balancing neighborhood exploitation with adaptive exploration. MGDAD maintains diversity through memory attractors enhanced by logarithmic perturbation mechanisms facilitating escape from local optima. AQMIVY's superiority was rigorously validated through extensive benchmark evaluations using CEC2017 (50D and 100D) and CEC2022 (20D) function suites against various state-of-the-art algorithms. Results demonstrate AQMIVY achieved first rank on both CEC2022 and CEC2017 functions with Friedman rankings of 1.97 to 2.36. Additionally, AQMIVY was successfully integrated into the VConVitICH-AQMIVY-RSVM framework for Intracranial Hemorrhage detection from CT scans, where it optimizes feature selection from fused Vision Transformer, ConvNeXt-V2, and VGG16 representations classified by Radial Basis Function SVM. Moreover, the SVM is also optimized using the proposed AQMIVY. The framework achieved classification accuracy of 98.83%, precision of 98.76%, recall of 98.74%, F1-score of 98.75%, and specificity of 99.76%, outperforming six state-of-the-art medical imaging models. Furthermore, Gradient-weighted Class Activation Mapping (Grad-Map) visualization confirmed anatomically consistent and clinically interpretable decision-making, validating AQMIVY's effectiveness in life-critical diagnostic applications.
Jia et al. (Tue,) studied this question.