Introduction to Machine Learning - 10-701/15-781

Practical information

  • Lectures: Monday and Wednesday, 12:00PM to 1:20PM

  • Location: Baker Hall A51

  • Recitations: Tuesdays 5:00PM to 6:00PM

  • Location: Porter Hall 100 (January 22, 2013), Doherty Hall A302 (January 29, 2013 onwards)

  • Instructor: Barnabas Poczos (office hours 10:00am-12:00 noon Thursdays in Gates 8231) and Alex Smola (office hours 2-4pm Tuesdays in Gates 8002)

  • TA: Ina Fiterau, Mu Li, Junier Oliva, Xuezhi Wang, Leila Wehbe

  • Grading Policy: Homework (33%), Midterm (33%), Project (33%), Final (34%) with best 3 out of 4 used for score (final is mandatory).

  • Google Group: Join it here. This is the place for discussions and announcements.

Office hours and questions

If you have questions, you should do the following:

  • If it's a question that other students also might have, ask it on the Google Group first.

  • If this doesn't solve your question or if it's not something you'd like to discuss in public, send an e-mail with 10-701 in the subject.

  • Come to the office hours. By default they're open for all. And you're welcome to stick around while we're answering other students’ questions.

  • Alex has office hours on Tuesday 1-3pm in Gates 8002. Barnabas’ office hours Thursdays 10am-12pm in Gates 8231.


Homework, midterm, and project count 33% each. You can pick the best two out of three for grading (translation, you can flunk one and still get full points). Moreover, the final exam counts 34% and it is mandatory.


To satisfy the auditing requirement you must do one of the following:

  • Do the homeworks and get at least 50% of the points on average.

  • Take both exams and get at least 50% of the points in each.

Please send an email to the instructors that you'll be auditing the class and let them know beforehand what you're planning to do.

Waitlisted students

If you are waitlisted do not despair. Often students drop out and you will get a slot eventually. If there's space in the lecture theater, feel free to attend (obviously students who are registered have priority). Also, the videos of the lectures will all be up online, typically within 24-48 hours of the class (the exact timing depends on compression, network speed, and the time it takes YouTube to process the videos).

To ensure that you can get credits for your homework, you need to do the following: do the assignments and put them into a dated and sealed envelope with your name on it. Write waitlisted on the envelope. We will open and grade the assignments only once you are no longer waitlisted. There is no guarantee that we ever will do that. It will only happen if you are officially enrolled. That said, students who diligently submit assignments will get priority over students who don't. If you're still waitlisted by the time of the midterm, it's too late.


  • There will be 5 sets of assignments, due every second week. Please hand in the assignments at the beginning of the class on the due date (typically monday).

  • Place each problem onto a separate stack as this helps the TAs to grade them (different TAs specialize on different problems).

  • Assignments received after the due date receive zero credit. No excuses. This is draconian. However, only the best 4 out of 5 assignments count, so you can flunk one and still get full points.

  • Homeworks are due individually. Each student must hand in their own answers. If you collaborate with others, it is your responsibility to make sure you personally understand the solution.

  • You must indicate on each homework with whom you collaborated.

  • We strongly discourage you from copying solutions of your fellow students since you're depriving yourself of the experience of learning how to solve the problems on your own. In particular you won't learn useful things for the exams and projects this way. Or for that matter, from the course.

  • Likewise, since this is a graduate class, we expect you not to simply 'google’ your solutions (not that this would help you much anyway).

  • That said, feel free to discuss the solutions with others. You will likely benefit from that.


There are two exams. The midterm exam is on Monday, March 4, 2013. The final exam is on Monday, May 6, 2013, both of which are in lieu of the regular class.


Like any class project, it must address a topic related to machine learning and you must have started the project while taking this class. You will need to submit a project proposal (i.e. a paper draft) and present a poster. You need to present a final paper and a poster. The 6 best projects will get the opportunity to give brief talks about your work at the end of the class. The rest gets spotlights.

The final project should be completed in teams of 3 students. Teams of 2 or 4 are OK if there’s a good reason. Obviously, larger teams are expected to deliver more. Single projects are not OK unless you can prove that a) nobody would take you on their team and b) there is no project that you would like to work on. In an emergency, Alex and Barnabas can help overcome obstacles a) and b).

The report should describe the project. It should describe your work in a reproducible manner, i.e. in enough detail that someone competent could take the report and regenerate the results (after some work but no guesswork) reliably.