Learning from Data

Catalog description

Data shapes decisions in science, technology, medicine, policy, and everyday life. This course asks how we build models that help us understand data, explain patterns, and make reliable predictions. Students will learn how to frame different questions about data, design models to test hypotheses, and work with data in Python. Students will learn how to build and train models, how model complexity affects performance, and why concepts such as overfitting and generalization matter. Linear algebra ideas such as vectors, matrices, and eigenvalue decomposition help explain how models operate on data, while probability and statistics provide the tools needed to reason about randomness, uncertainty, and variability. Building on these foundations, the course covers linear and logistic regression, regularization, and rigorous model evaluation using cross-validation, bootstrapping, and permutation testing. Students will also learn methods for discovering structure in data, including principal components analysis (PCA), manifold learning, and clustering. Throughout the semester, students work with real datasets and practice the full modeling cycle: posing a question, building and training models, evaluating performance, and refining their approach while confronting realistic challenges such as noise, distribution shifts, and class imbalance.