Classical ML Algorithms

Classical ML Algorithms for Safety

Traditional ML algorithms provide intuition for understanding deep learning safety.

Linear Models

• Linear & logistic regression
• Interpretability advantages
• Limitations and failure modes
• Safety applications: Baseline models, interpretable classifiers

Tree-Based Methods

• Decision trees and random forests
• Feature importance and explainability
• Overfitting and generalization
• Safety applications: Rule extraction, decision analysis

Support Vector Machines

• Maximum margin classification
• Kernel methods and feature spaces
• Robustness properties
• Safety applications: Certified defenses, boundary analysis