Lecturer: Geoff Hulten TAs: Kousuke Ariga & Angli Liu CSEP 546 Machine Learning Lecturer: Geoff Hulten TAs: Kousuke Ariga & Angli Liu

Logistics Course location: https://www.washington.edu/classroom/SAV+260 Course website: https://courses.cs.washington.edu/courses/csep546/18au/ Canvas site: https://canvas.uw.edu/

Introducing Myself Geoff Hulten ghulten@cs.washington.edu https://www.linkedin.com/in/geoff-hulten-58136a1/ What I’ve worked on… Why I’m here…

Introducing our TAs Kousuke Ariga Research interests: Office hours: koar8470@cs.washington.edu Research interests: Office hours: Mondays 5:30-6:20 Second floor breakout in Allen Center Angli Liu anglil@cs.washington.edu Research Interests: My area of research is using external knowledge (knowledge bases, pre-trained embeddings, raw text, etc.) to help improve sequence models, with applications to low-resource language translation, and unsupervised named-entity recognition. Office Hours: < Let’s vote! >

Introducing the Class What types of jobs? Machine learning experience? Engineering? Data science? Management/PM? Other? Machine learning experience? This is my first class? Several classes / exploration? Do it for a living? Math? Not really? Can do? I think in math? Python? Never used? Some experience? No problem?

Overview of the Course 1) Learn important machine learning algorithms (the tools) 2) Learn how to produce models (use the tools) 3) Learn how to produce working systems (Machine Learning Engineering)

Lecture Overview 1 Overview of this Course Week Topics 1 Overview of this Course Overview of Machine Learning Evaluating 101: FP/FN and Confusion Matrices Logistic Regression 2 Feature Engineering (Text focus) Evaluating 201: Hold out, error bounds, cross validation ROC Curves and operating points 3 Decision Trees Overfitting and Underfitting Parameters in Modeling 4 Ensembles Machine Learning with an Abuser Deploying Machine Learning models Week Topics 5 Basics of Computer Vision Clustering and Dimensionality Reduction Instance Based Methods 6 Neural Networks An introduction to mistakes in machine learning Veterans Day Catch up on topics (maybe a video lecture)... 7 Design Patterns for Machine Learning Reinforcement Learning 8 Other important learning: Bayesian Networks, SVM, HMMs & CRFs Connecting Machine Learning to Users 9 Organizing models in large systems A quick review of the course Approaching your own machine learning project

Assignments Logistic Regression Feature Engineering (text) Decision Trees Ensembles Clustering Feature Engineering (Vision) Neural Networks Reinforcement Learning Model building & interpreting And more…

Evaluation Assignments: There will be ~8 weeks with assigned work. Each week’s work is worth ~10% of the final grade (1 point = 1%). You may hand each assignment in up to two weeks after it is assigned. Except for the last assignment, which is due before the start of the final lecture (so we can submit final course grades in a timely fashion). Clarity of communication is critical in machine learning, so your answers must be concise and easy to follow. If the TA can’t evaluate the answers in reasonable time they will have to give reduced credit. Exam: There will be an exam worth 20% of the final grade (although you must score at least 50% on the exam to pass the course). This will be online and timed, and will be available for 2-3 days. Final week of class? Finals week? This exam will be based on the assignments, so the best way to prepare will be to do the assignments carefully.

The Textbooks and why… All royalties to be donated…