CS446: Machine Learning

Spring 2017

Course Information


Meeting Times and Locations

Lecture: 3 or 4 hours credit, Tue/Thu 17:00 - 18:15, 1404 Siebel Center

Instructors
Professor:
Dan Roth
Office: 3322 SC
Office hours: Monday: 1:00-2:00 pm; or by appointment.
Phone: (217) 244-7068
E-mail: danr at illinois dawt edu
Teaching assistants:
  • Chase Duncan
    Office hours: Tues, 12pm - 1pm.
    E-mail: cddunca2

  • Subhro Roy
    Office hours: Wed, 4pm - 5pm
    E-mail: sroy9

  • Qiang Ning
    Office hours: Thurs, 3pm - 4pm.
    E-mail: qning2

  • Hao Wu
    Office hours: Fri, 1pm - 2pm
    E-mail: haowu4


  • All TA office hours will be held at the whiteboard by 3333 SC.

    Discussion Sections

    Teaching assistants will hold discussion sections four times a week to describe solutions to class exercises or homework, and to answer questions raised by students. Although we encourage you to attend, these sections are not mandatory, and all materials discussed in sections will be available online. Tentatively, the session will take place as follows. The plan is to start on the third week of the semester. Please plan to attend the following session based on your [last name initial].

    Monday: 4:00pm-5:00pm. Room 3405. Subhro Roy. [A-I]
    Wednesdays: 5:00pm-6:00pm. Room 3405. Hao Wu. [J-L]
    Thursdays: 4:00pm-5:00pm. Room 3405. Chase Duncan. [M-S]
    Fridays: 4:00pm-5:00pm. Room 3405. Qiang Ning. [T-Z]

    Course Description

    The goal of Machine Learning is to build computer systems that can adapt and learn from their experience. This course will study the theory and application of learning methods that have proved valuable and successful in practical applications. We review the theory of machine learning in order to get a good understanding of the basic issues in this area, and present the main paradigms and techniques needed to obtain successful performance in application areas such as natural language and text understanding, speech recognition, computer vision, data mining, adaptive computer systems and others. The main body of the course will review several supervised and (semi/un)supervised learning approaches. These include methods for learning linear representations, on-line learning methods, Bayesian methods, decision-tree methods, kernel based methods and neural netoworks methods, as well as clustering and dimensionality reduction techniques. We will also discuss how to model problems as machine learning problems and some open problems.

    Topics to be covered:

    Prerequisites

    Students are expected to have taken a class in linear algebra and in probability and statistics and a basic class in theory of computation and algorithms.

    Course Material

    Text: Tom Mitchell, Machine Learning, McGraw Hill, 1997
    I will not follow the text closely - a lot of the material covered isn't there, but it is still a very useful book to have. Some of the material covered is in the following books (but I do not recommend purchasing these now; they are all reserved in the Engineering Library):

    Optional books with relevant material:

    Lecture notes, course handouts, pointers to relevant papers, and other materials will be available as HTML and PDF documents from the course home page.
    http://L2R.cs.illinois.edu/~danr/Teaching/CS446-17/

    Online Discussion Platform

    We will use Piazza for our online discussion platform. You can use Piazza to ask the course staff and your classmates questions about the course material (rather than sending individual e-mails). You can register to class discussion space in Piazza at https://piazza.com/illinois/spring2017/cs446.

    Grading

    Course grades will be based on: 25% -- homework, 5% -- online quizzes, 30% -- mid-term, and 40% -- final. For student that wish to earn 4 hours credit, the final grade will be scaled accordingly (with the project being 25% of the grade).

    Reporting Mistakes:
    We encourage students to find mistakes in the lecture notes and the powerpoint slides. Students who report mistakes (email to TAs and the Professor) will get a small amount of credit.

    Homework:
    There will be 7 +/- 1 problem sets; the first few will involve programming and experimental work. Instructions on how to submit solutions will be available on the course home page later.

    Quizzes:
    There will be short quizzes nearly every week hosted on Compass 2g (in the "Quizzes" folder). The purpose of these quizzes is to get you to review the lectures from the previous week and to think about the involved content. They will be short (~10 minutes) and are open note. In general, they will be due on Mondays at 11:59 PM; consult the course schedule for further details.

    Late Policy:
    You have 96 hours of "credit" that you can use any way you want. You don't need to come to us and ask to submit the homework late. Just submit it when you are ready; we will accumulate your late time and allow up to 96 hours for the whole semester.

    Late submission will not be accepted once homework solutions are released online. Typically, we will wait for up to 96 hours after due date before we release the solutions. However, when there is a mid-term or a final coming up, we may want to release solutions earlier. This means that you may not be able to use all 96 hours of credit on assignments that are due near the mid-term or the final. When we release these assignments, we will inform you if late submission is not accepted.

    Do let us know if there are extreme situations where this lenient policy isn't satisfactory.

    Mid-term Exam:
    Thursday, March 16, in class.

    Final Exam:
    Tuesday 1:30pm, May 09 (during the final week)

    Projects:
    Students who wish to get 4 hours credit will have to submit a project and (possibly) give a short presentation. Details will be available later in the semester.

    Check the project proposal page for details on the submission of the proposal, status report and final project report.
    We will use Compass 2g to submit homework, evaluate and assign grades to homework, and allow you easy access and ability to know well you are doing.


    Dan Roth