Modeling, Evaluation, and Optimization

This unit advances core machine learning skills by focusing on model types, probabilistic reasoning, evaluation, and optimization techniques. Students will learn to implement and tune supervised and unsupervised models (including SVMs, k-NN, k-means, and CNNs), apply dimensionality reduction and ensembling, and address inference, metric selection, and algorithmic considerations to improve generalization and reliability.

Learning Objectives

• Analyze training, validation, and test procedures and compute appropriate error and scoring metrics (e.g., accuracy, precision/recall, F1, ROC-AUC) and use cross-validation to assess and report model generalization
• Apply probabilistic methods and algorithm-specific tuning (e.g., probabilistic classifiers, k-means tuning, SVM parameter selection, and gradient-descent-based boosting) to improve model performance through systematic hyperparameter search
• Demonstrate end-to-end implementation of common ML tasks—such as image classification with CNNs, text processing, dimensionality reduction with SVD, and k-NN similarity searches—using appropriate preprocessing, feature engineering, and code structure (classes) for reproducible results
• Evaluate algorithmic and inference-related issues by identifying common pitfalls, interpreting model behavior, selecting proper error metrics, and articulating the relationship between AI and machine learning to ensure ethical, reliable, and well-generalized solutions