Course Emphasis and Goals

The goal of machine learning is to build computer systems that automatically solve problems using sample data and past experience. This course provides an overview of the state-of-art algorithms along with their theoretical and practical aspects. In this course, students will have an opportunity to have hands-on experience to understand the basic principles of these machine learning algorithms.

Description

Bayesian decision theory, parametric methods, nonparametric methods, decision trees, linear discrimination, multilayer perceptrons, unsupervised learning and clustering, hidden Markov models, and reinforcement learning.

Prerequisites

• Familiarity with the basic probability theory and statistics.
• Knowledge of a programming language (Java, C/C++, Matlab, etc.) to write reasonably non-trivial programs.

Grading Policy

Homework: 35%
Presentation: 30%
Midterm: 35%

Exams

The midterm exam will be closed book. You may bring only 2 A4 sheets (back and front = 4 pages) of your handwritten notes to the exam.

Presentations

Each student will present a journal paper in the class. Students are expected to select the papers that they will present. The papers could be selected from diverse areas (such as bioinformatics, robotics, computer vision, etc.) but they should contain some kind of applications in which machine learning techniques are used and they should be approved by the instructor. For approval, students are supposed to submit the reference and a copy of their selected papers to the instructor by March 16, 2009. These papers will be announced on the course web page.

Each student will have 15-20 minutes to present his/her paper. We will have a discussion period of 5-10 minutes after the presentation. Students are supposed to submit their presentations to the instructor and these presentations will be graded. Besides, each student is expected to read the papers selected by the other students and to participate the discussion session. The participation of the students will also be graded.

As a presenter, each student is expected to read the paper entirely, deeply understand the paper, and relate the techniques that are used in the paper to the topics that we will have seen in the class. Thus, it is important for you to prepare your presentations accordingly. Besides, the quality of the presentation is also important.

As a participant of a discussion session, each student is expected to ask relevant questions to the presenter. For that, it is important for you to read the paper entirely before coming the presentation and relate the techniques that are used in the paper to the topics that we will have seen in the class.

Homework Assignments and Late Policy

Homeworks will be programming assignments that require students to implement machine learning algorithms. Students are expected to work in groups of two for the assignments. For each assignment, students will be expected to write a detailed report on their findings by running their algorithms on given data sets. Additionally, there will be a demo session for each assignment. Each student is supposed to be ready at this demo session; it will be a part of the grading. In these sessions, students will make a small demonstration on how they run their algorithms to obtain the results and they are expected to answer questions about the implementation of their algorithms.

Assignments are expected to be turned in by 17:00 on the due date. For the late assignments, each group will be given a total of three grace days (whole or partial) for the whole semester. Once these late days have been exhausted, no late assignments will be accepted. As an example, if Group A submits their 1st assignment 29 hours late, they will have used two late days and have only one day left. If Group A then submits their 3rd assignment 5 hours late, they will have used their remaining late day. If Group A submits their 4th assignment 1 minute late, this assignment will not be accepted.

Academic Integrity

Copying or communicating during an exam is cheating. Students caught cheating on an exam will be subject to disciplinary action, as explained in the "Student Disciplinary Rules and Regulation" ( http://www.provost.bilkent.edu.tr/procedures/AcademicHonesty.htm).

Students in the same group are expected to work together. On the other hand, students in different groups are not allowed to discuss the solutions of programming assignments or to get help to write their codes and reports. Students caught cheating on assignments will also be subject to disciplinary action.