Advanced Data Science for Biomedical Engineering

Class Description

This course will focus on the fundamental concepts, learning paradigms, and problems in modern data science. It will provide students with a precise characterization of problems and algorithms for learning from data, from problem formulations to data models to algorithms. We will cover supervised and unsupervised problems, as well as topics on modern aspects of the application of data science in the 21st century, including aspects of interpretability and fairness.

Who is this for? This course is "advanced" in the sense that it will presume undergraduate-level courses in statistics, linear algebra, programming, and basic introductory classes to data science. It is ideal for undergraduate students wanting to specialize in data science, or for graduate students looking for a theory-oriented introduction to data science.

Syllabus/Topics

• What is Data Science and Machine Learning?
• Fundamentals of Decision Theory
• Fundamentals of Supervised Learning and Generalization
• Feature learning and data representation
• Unsupervised learning: PCA, clustering, sparse representations
• First-order optimization methods
• Interpretability in machine learning
• Fairness in machine learning
• Notions of causal inference
• Applications to BME: neuroscience microscopy, protein folding, digital pathology

Class Format & Evaluation

The class combines lectures, hands-on exercises, and group discussions. Lectures are once a week, 2.5 hrs (with a break). TA session once a week complements the topics covered in the lectures and facilitates weekly (mandatory) quizzes. A small number of coding exercises are due through the first part of the semester, and a sizeable, research-like project (in small groups) is due at the of the semester.

Evaluation: Quizzes 40% · Project presentation 40% · Coding assignments 20% · Participation (bonusu) 10% ·