Applied Model Training and Evaluation

This capstone unit integrates practical machine learning workflows, emphasizing reproducible execution, performance optimization, and responsible practice. Students will run and document experiments, engineer and ensemble models, validate data integrity, and assess ethical and operational implications.

Learning Objectives

• Analyze training workflows and experimental results to identify data leakage, estimate variance, and detect rare failure modes using validation holdouts and reproducible notebooks
• Apply hardware acceleration, dataset management, and execution procedures to run end-to-end experiments, verify data sources, and produce shareable runnable notebooks
• Demonstrate feature engineering, model comparison, and ensemble techniques (blending and stacking) to improve predictive performance and report comparative metrics
• Evaluate ethical risks, potential biases, and downstream consequences of models, and implement mitigation and communication strategies to avoid discriminatory practices