Overview
This report studies how metaheuristic algorithms can support cancer-related machine learning tasks. The analysis separates training-level optimization, such as learning-rate and weight tuning, from representation-level optimization, such as segmentation thresholding and feature selection.
Key Features
- Compares Whale Optimization for neural-network weight optimization in breast cancer detection.
- Explains Bat Algorithm learning-rate optimization for lung cancer prediction workflows.
- Reviews Firefly-based segmentation for brain tumor classification.
- Analyzes Bat-optimized feature selection with Extreme Learning Machine classification.
- Discusses computational overhead, convergence sensitivity, reproducibility limits, and clinical generalization concerns.
Evidence
Interactive Slide Deck
Metaheuristics Presentation
Slide 1 of 28
The strongest takeaway is that metaheuristics can supervise different parts of an AI pipeline, not only neural-network parameters. They can optimize how a model learns or what representation the model receives, but that flexibility often increases compute cost and tuning burden.
