Practical Machine Learning

Weekly on Saturdays, from March 26 to June 4, 2022.

Total Cost: $750

$100 registration deposit | $650 tuition

Tuition is due in full by the second day of class.

Whom it's for

This 11-session class is designed for students with a Python programming background to learn today's most salient machine learning techniques.

By the end of the class, students will be able to program neural networks, apply best practicies to data analytics, and be empowered to deploy multiple kinds of machine learning models.

We assume no machine learning background. We do assume knowledge of intermediate-level Python. The skills learned in Oak Stream's Python Crash Course serve as pre requsites. An A-/B+ in UO's CIS 211 (Computer Science II) or B/B- in UO's CIS 322 (Introduction to Software Engineering) also serve as indicators of sufficent programming background to take this class.

What to expect

This is a projects-driven class. The projects will help prepare you for professional employment and exploration of new research avenues. We ask you put a lot of time into this class, and we believe the payoff will be significant for each of you.

The projects in this class will be hard work. The class is somewhere around the junior undergraduate to beginning graduate school level, in terms of rigor. It is difficult to quote exactly how much time this course will take outside of lecture, since there is variation in background and programming skill. We expect those who have less developed programming skills will find this class to be a considerable effort, but also that they will have significant improvement by the end of the experience.

There will be several projects you will complete, nearly one for each day of class. Each project will train you on key machine learning concepts.

What you'll learn

Lecture topics, by day:
  1. Primer on machine learning and artificial intelligence.
  2. Linear regression & regressors.
  3. Support vector machines; k fold validation; error metrics.
  4. Trees; precision vs recall.
  5. Forests; ensemble methods; r^2 and explained variance.
  6. Neural networks; perceptrons, the xor problem.
  7. Neural networks; feed forward networks; Newton's method; optimization functions; backpropogation.
  8. Neural networks; feed forward networks; ReLUs and other activation functions.
  9. Neural networks; feed forward networks; dropout, skip connections; hyperparameter tuning; universal approximation theorem.
  10. Neural networks; convolutional neural networks.
  11. Ethics, explainability, auditability, and accountability; class imbalance problem; SHAP analysis; dimensionality reduction.

When it takes place

The class will take place weekly, for 11 meetings in total. We will be starting Saturday March 26, which is the last Saturday of spring break for the University of Oregon and Lane Community College.

We will begin at 11am and meet for 1.5 hours.

Where it's at

In person in Eugene, Oregon.

The location will be near (or on) the main campus of the University of Oregon. The classroom will be within a 5ish minute drive or a 30ish minute walk from UO's EMU building. Exact building and room details will be emailed to registered students one week before the class begins.

Registration Policies