CS 4786/5786 - Machine Learning for Data Science

General Information

An introductory course in machine learning, with a focus on data modeling and related methods and learning algorithms for data sciences.

Prerequisites

• Probability Theory (BTRY 3080, ECON 3130, MATH 4710, or strong performance in ENGRD 2700 or equivalent)
• Linear Algebra (MATH 2940 or equivalent)
• CS2110 or equivalent programming proficiency

Topics Covered

CS 4786 focuses on Unsupervised Learning. Topics include:

• Dimensionality Reduction (PCA, CCA, Random Projection)
• Clustering (K-Means, Single-Link, Spectral)
• Probabilistic Modeling (Mixture Models, Hidden Markov Models, EM Algorithms)

Workload

6 assignments, 2 Kaggle projects, 0 prelims, 0 finals

If you have a background in ML or strong knowledge of probability, the assignments may be light, but otherwise start early just in case. Projects can be a heavy workload, but at least you have a team.

General Advice

• Go to office hours for assignments
• Start early on everything, especially the projects!
• If Kaggle’s format hasn’t changed, you get a number of submissions per day. Use them!
• Even if you do great on Kaggle, don’t skimp on the writing quality of your report.
• Even though there aren’t exams, try to review the lectures so you can get the most out of the class!
• Python has a lot of awesome libraries for ML like numpy and scikit-learn.

Testimonials

While the class was hard for me, the format of the class was enjoyable. Even though at one point I spent two days straight locked in my room trying to implement a working Hidden Markov Model, I learned a lot about machine learning the whole time.

Past Offerings