This session has ended. For next session, see Machine Learning Spring 2015.

Title: Machine Learning 

Course number: 91.422 (undergraduate) 91.545 (graduate) 

Session: Spring 2014

Instructor: Kate Saenko, saenko@cs.uml.edu  

TA: Edward Luo, edwardluonlp@gmail.com

Location: Olsen Hall, Room 406

Meeting Times: Tue/Thu 4-5:15pm

Office Hours: 

Kate: Olsen 223, Mon 10:00-12:00, Tue 13:00-14:00, and by appointment

Edward: Olsen 212A, Tue/Thu 11:30-13:00

Catalog listing   

This introductory course gives an overview of machine learning techniques used in data mining and pattern recognition applications. Topics include: foundations of machine learning, including statistical and probabilistic methods; generative and discriminative models; linear regression; Bayesian methods; parametric and non-parametric classification; supervised and unsupervised learning; clustering and dimensionality reduction; anomaly detection; and applications to very large datasets.

Prerequisites

This is an upper-level undergraduate/graduate course. Students should have completed the following courses (or their equivalents):

  • 92.322 Discrete Structures II (or Linear Algebra)
  • 92.386 Probability & Statistics I
  • 91.102 Computing II

In addition, students must complete and pass Quiz 0 on prerequisite math knowledge – Probability & Statistics, Discrete math, Calculus, basic Matrix Algebra given in the first week of class. Students who cannot pass Quiz 0 are strongly discouraged from taking the class.

the course can be taken as part of the two-course project sequence in combination with either 91.420 Artificial Intelligence, 91.530 Topics: Natural Language Processing, or 91.423 Computer Vision.

Textbooks

The required textbook is:

The course will also follow and build on Andrew Ng’s online course lectures, available at

                  https://class.coursera.org/ml-003/lecture/preview

Other recommended supplemental textbooks are:

Software and Hardware

The course will use the Octave scientific programming environment, a free version of Matlab. To facilitate programming assignments, software libraries in the corresponding programming language will be made available to the students.

Meeting Format

The class will meet twice a week for a 75 min lecture, taught by the instructor.

Syllabus Overview

  1. Introduction to Machine Learning. Review of mathematical concepts.
  2. Probabilistic Approaches. Maximum Likelihood Principle.
  3. Multivariate linear regression. Bayesian regression. Bias-variance trade-off.
  4. Logistic regression, One-vs-all classification, Regularization.
  5. Neural Networks. Convolutional Neural Networks.
  6. Practical advice for applying learning algorithms: debugging, feature/model design, setting up experiment structure, evaluating methods.
  7. Support Vector Machines (SVMs) and the intuition behind them.
  8. Active learning. Bootstrapping.
  9. Unsupervised learning: clustering and dimensionality reduction, anomaly detection.
  10. Ensemble classifiers: Boosting.
  11. Nearest neighbor models. Metric learning.
  12. Large-scale machine learning. Example applications of machine learning.

Student Learning Outcomes

After the completion of the course, the students should be able to understand 

  • the design, construction and evaluation of a machine learning/pattern recognition system
  • the mathematical foundations of several mainstream machine learning algorithms
  • major classes of approaches in machine learning/pattern recognition literature
  • the theoretical issues involved in learning algorithm design such as the curse of dimensionality
  • implementation of machine learning techniques in a scientific computing environment

These goals will be evaluated through quizzes, homeworks, and a final project.

Deliverables/Graded work

Students will be evaluated based on the following graded work (subject to change):

  • best 5 of 6 homeworks: 25% (5% each) 
  • class participation: 5%
  • final project: 30%, 
  • 3 in-class quizzes: 40% 

Students enrolled in the graduate section will be expected to submit additional work and/or achieve a higher percentage on the exams to receive the same grade as undergraduate section students. 

Late Policy

  • 20% off per day
  • up to 4 days
Important: if you cannot submit an assignment on time because of CS department server issues, please IMMEDIATELY 1) send an email to help@cs.uml.edu, 2) cc: the instructor and TA, 3) do not modify the timestamps on your files, i.e. timestamps should be from before the deadline, even if the files are submitted later. Only doing (1)-(3) can result in late charge reversal.

Academic Honesty Policy: Students are expected to honor all CS department and UMass Lowell policies related to academic honesty and integrity. Violators risk failing the course in addition to any actions taken by university administration. The default CS department policy is that a student who cheats will fail the course. The University's policy is described here. A definition of plagiarism is here. All work on exams must be the student's own work. All work on homework assignments must also be the student's own work, with the following exceptions: 1) hints provided by the instructor or TA may be used but must be acknowledged in writing in the student's work; 2) high-level hints from another student may also be used if a student is unable to make progress on an assignment problem on his/her own; this type of hint must also be acknowledged in writing in the student's work. Detailed collaboration among students on homework assignments is not permitted. Students cannot obtain homework answers from web sites. 

Religious Observance: UMass Lowell respects the religous observances of students. If religious obligations conflict with course due dates and/or examinations, students should notify the professor in writing well in advance of the due date.