تعديلات مطلوبة من مجلة لاستكمال إجراءات النشر

هذه تعديلات مطلوبة من مجلة لاستكمال إجراءات النشر

Reviewer 1

1. VASI mainly combines established techniques (PSO, WSO, Lévy flight, Gaussian search, and adaptive switching). The manuscript does not clearly demonstrate a sufficiently novel optimization principle.

2. Only Sphere, Rastrigin, and Ackley functions are used. These simple benchmarks are not enough to establish the superiority of a new meta-heuristic. Modern CEC benchmark suites should be included.

3. The comparison is mainly against PSO and WSO. Strong methods such as Bayesian optimization, Differential Evolution, CMA-ES, and Random Search should be considered.

4. The individual contributions of adaptive switching, Lévy flight, Gaussian refinement, and ensemble global-best mechanisms are not demonstrated.

5. Only four hyperparameters are optimized. It is unclear whether such a small search space justifies a complex meta-heuristic optimizer.

6. Each candidate is evaluated for only 64 mini-batches. This may optimize early-training performance rather than final model performance.

7. The manuscript does not clearly establish that histopathology images were split at the patient/whole-slide level. If patches from the same patient or slide occur in both training and testing sets, the very high reported accuracy may be inflated.

8. The reported improvements are not supported by appropriate repeated experiments, confidence intervals, or paired statistical tests.

9. The reported 99% accuracy and 0.9994 AUC are exceptionally high and require stronger evidence that they are not caused by data leakage or dataset-specific characteristics.

10. The optimization cost of VASI is not clearly reported. The statement that the improvements come at “no computational cost” is misleading because VASI requires many additional candidate evaluations.

11. Grad-CAM examples and retrospective classification results are insufficient to establish clinical utility or real-world diagnostic value.

12. Only two medical datasets and one main architecture are evaluated. The evidence is insufficient to support broad claims about generalization across medical-imaging tasks.

13. High confidence on correct predictions does not demonstrate that the model is well calibrated. Calibration metrics such as ECE or Brier score are needed.

14. The performance may depend strongly on the manually selected hyperparameter ranges, but this sensitivity is not investigated.

15. Encourage the authors to discuss new and recent optimization algorithms, such as "Secant Optimization Algorithm for efficient global optimization", and "Schrödinger optimizer: A quantum duality-driven metaheuristic for stochastic optimization and engineering challenges".

16. I notice that many references that are related to the research topic are somehow old. Please make the reference robust with a recent close paper such as :”Large scale salp-based grey wolf optimization for feature selection and global optimization”; “An innovative quadratic interpolation salp swarm-based local escape operator for large-scale global optimization problems and feature selection”; “An efficient SVM-based feature selection model for cancer classification using high-dimensional microarray data”; “An innovative time-varying particle swarm-based Salp algorithm for intrusion detection system and large-scale global optimization problems”; “An optimized computational framework for non-small cell lung cancer subtype classification and biomarker discovery”; “An optimization-based framework for gene expression-based diagnostic classification of Wilms tumor”; “Identification of Significant Risk Factors and Robust Cardiovascular Disease Prediction Using a CLPO-Optimization-Based Framework”; “miRNA-based breast cancer subtyping using AHALA multi-stage classification approach”.

Reviewer 2

Reviewer Comments:

The paper proposes VASI, which combines PSO exploitation, WSO exploration, Levy-flight escape, and Gaussian local refinement under an adaptive phase-switching mechanism.

This is a strong methodological contribution because it attempts to address premature convergence and insufficient exploration simultaneously. However minor problems exist:

1. Why authors selected only three test functions in Table1? These are useful but insufficient to comprehensively establish optimizer superiority.

2. Computational cost of VASI is a concern. Describe time complexity of proposed method.

3. Compare VASI experimental results with more SOTA methods. It would provide stronger evidence of generality.

4. The introduction/related review section should be enhanced by incorporating additional review studies like.

DOI: 10.2174/0118763863291395240516093102

https://ieeexplore.ieee.o...

https://openpublichealthj...

https://www.sciencedirect...