Modeling, Tuning, and Evaluation

This unit advances core machine learning techniques for building, tuning, and evaluating models across supervised and unsupervised tasks. Students learn probabilistic methods, optimization and hyperparameter search, dimensionality reduction, basic deep learning for images and text, and practices to ensure generalization and avoid inference pitfalls. It builds on foundational concepts from earlier units and prepares learners to integrate models and ethical considerations into higher-level projects.

Learning Objectives

• Analyze training, validation, and testing workflows and select appropriate error and performance metrics to assess model quality and avoid inference pitfalls.
• Apply supervised and unsupervised algorithms (including k-means, k-NN, SVM, ensemble methods, and CNNs) and implement probabilistic approaches to perform classification and estimation tasks.
• Demonstrate techniques for improving generalization, including cross-validation, regularization, hyperparameter search, and model averaging, and use SVD for dimensionality reduction and feature engineering.
• Evaluate the role and limitations of algorithms and AI systems by identifying common pitfalls, ethical considerations, and appropriate uses of machine learning methods in applied projects.