EECS 445: Introduction to Machine Learning
Winter 2015
Instructor:
Prof. Jenna Wiens
Office: 3609 BBB
wiensj@umich.edu
Graduate Student Instructor:
Srayan Datta
Office: 3349 North Quad (**office hours location 3941 BBB**)
srayand@umich.edu
Course Information:
Lectures
Monday & Wednesday, 1:30pm-3:00pm, 1010 DOW
Discussions
Friday 11:00am-12:00am, 1010 DOW
Course Materials & Textbook
Course materials will be posted on the course CTools site (https://ctools.umich.edu/portal)
Recommended Textbooks (optional):
Chris Bishop, “Pattern Recognition and Machine Learning”, Springer, 2007.
Kevin Murphy, “Machine Learning: A Probabilistic Perspective”, MIT Press, 2012.
Course Description
The course is a programming-focused introduction to Machine Learning. Increasingly,
extracting value from data is an important contributor to the global economy across a range
of industries. The field of Machine Learning provides the theoretical underpinnings for dataanalysis as well as, more broadly, for modern artificial intelligence approaches to building
artificial agents that interact with data; it has had a major impact on many real-world
applications.
The course will emphasize understanding the foundational algorithms and “tricks of the
trade” through implementation and basic-theoretical analysis. On the implementation side,
the emphasis will be on practical applications of machine learning to computer vision, data
mining, speech recognition, text processing, bioinformatics, and robot perception and
control. Real data sets will be used whenever feasible to encourage understanding of practical
issues. On the theoretical side, the course will give a undergraduate-level introduction to the
foundations of machine learning topics including regression, classification, kernel methods,
regularization, neural networks, graphical models, and unsupervised learning.
EECS 445: Introduction to Machine Learning
Page !1
Prerequisites
EECS 281 In addition, students should have familiarity with linear algebra (MATH 217,
MATH 417) and probability (EECS 401).
Office Hours
GSI: 3941 BBB Wednesday 3:00pm-4:30pm, Thursdays 3:00pm-4:30pm
Instructor: 3609 BBB, Tuesdays 10:30am-Noon, or by appointment
Course Tools
Information about the course including assignments and supplementary readings will be
posted on CTools (https://ctools.umich.edu/portal). You are expected to check the site
frequently, although you usually will be automatically notified by e-mail when new materials
are posted. We will be using Piazza for class discussion. Find our class page at: https://
piazza.com/umich/winter2015/eecs445001w15/home
Grading
Exams
Midterm on March 11th (20%)
Final exam during finals weeks (35%)
Homework
4 Problem Sets (10%)
3 Mini-Projects (30%)
In-Class Problems (4%)
Course Evaluation (1%)
There will be four problem sets and three mini-projects assigned over the course of the
semester to strengthen understanding of fundamental concepts and provide an opportunity for
hands-on learning using real datasets. In-class problems will be handed out and solved during
lectures, graded for effort.
Submitting homework:
Bi-weekly homework assignments are due Fridays at 9am on the dates noted in the course
schedule. Scanned copies of your homework should be submitted via CTools.
Grading policy:
Solutions for problem sets will be posted exactly three days after the due date at 9am.
Students will be responsible for grading their own problem sets. Scanned copies of the
graded/corrected assignments are due at the same time as the next assignment. The goal
behind this policy is for students to take an active role in addressing their own
misconceptions and in evaluating their own performance. Mini-projects will be graded by the
course staff.
Late submission policy for homework and projects:
You can be up to three days late, automatically losing 10% for each 24 hour period starting
immediately (e.g., 1 min late means 10% off, 25 hours late means 20% off, and so on). No
submissions will be accepted after the three days unless accompanied with a note from the
Dean.
EECS 445: Introduction to Machine Learning
Page !2
Honor Code
The Honor Code outlines certain standards of ethical conduct for persons associated with the
College of Engineering at the University of Michigan. The policies of the Honor Code apply
to graduate and undergraduate students, faculty members, and administrators. Read about the
UM Honor Code here: (http://www.crlt.umich.edu/faculty/honor.html). There is also an
Engineering Honor Code: (http://www.engin.umich.edu/students/honorcode/code/). In this
class, as in many others at the University, you will be expected to include and sign the Honor
Pledge on each assignment you submit. The Honor Pledge is as follows:
I have neither given nor received unauthorized aid on this assignment, nor have I
concealed any violations of the Honor Code.
The Honor code is based on these tenets:
o Engineers must possess personal integrity both as students and as professionals. They
must be honorable people to ensure safety, health, fairness, and the proper use of
available resources in their undertakings.
o Students in the College of Engineering community are honorable and trustworthy
persons.
o The students, faculty members, and administrators of the College of Engineering trust
each other to uphold the principles of the Honor Code. They are jointly responsible for
precautions against violations of its policies.
o It is dishonorable for students to receive credit for work that is not the result of their own
efforts.
Among other things, the Honor Code forbids plagiarism. To plagiarize is to use another
person's ideas, writings, etc. as one's own, without crediting the other person. Thus, you must
credit information obtained from other sources, including web sites, e-mail or other written
communications, conversations, articles, books, etc.
On team assignments, the co-authors listed on the submission should include only those team
members who have contributed their fair share to the assignment. If you allow a teammate's
name to appear on an assignment to which he/she has not contributed fairly, then you are
violating the Honor Code.
Handling Data with Integrity
You may not falsify or misrepresent methods, data, results, or conclusions, regardless of their
source.
Unfair Advantage
You may not possess, look at, use, or in any way derive advantage from the solutions of
homework, exams or papers prepared in prior years, whether these solutions were former
students’ work products or solutions that have been made available by University of
Michigan faculty or on the internet, unless this section’s faculty expressly allows the use of
such materials.
EECS 445: Introduction to Machine Learning
Page !3
Disability Policy
If you have any disability as defined under the Americans with Disabilities Act that might
interfere with your ability to participate in class, or to turn in assignments on time or in the
form required, please contact your instructor and the Office of Students with Disabilities at
the start of the term so that arrangements can be made to accommodate you
Tentative Course Schedule (Subject to Change)
If you have any disability as defined under the Americans with Disabilities Act that might
inte
Lecture
Topics Covered
01/07
Introduction
Machine Learning: What & Why?
Linear Classification
01/09
PS1 - Out
01/12
Supervised Learning
Learning Linear Classifiers, Perceptron Algorithm
01/14
Supervised Learning
Linear Classifiers Non-Separable Case, Gradient
Descent
01/21
Supervised Learning
Linear Regression - Empirical Risk & Least Squares,
Regularization
01/23
PS1 - Due, Project1-Out
01/26
Supervised Learning
Support Vector Machines; Primal Formulation;
Geometric Margin
01/28
Supervised Learning
Dual Formulation; Kernels
02/02
Supervised Learning
Feature Construction; Selection: Filter, Wrapper,
Embedded Methods
02/04
Performance Evaluation
Confusion Matrices; AUROC; F-score; Calibration
02/06
Project1- Due, PS2 - Out
02/09
Supervised Learning
Decision Trees; Entropy
02/11
Ensemble Methods
Bagging; Random Forest
02/16
Ensemble Methods
Boosting; Adaboost
02/18
Recommender Problems
Collaborative Filtering
02/20
PS2- Due, Project 2 - Out
02/23
Unsupervised Learning
Introduction to Clustering; K-means; Hierarchical
Clustering
02/25
Unsupervised Learning
Spectral Clustering; Clustering as a Graph Cut Problem;
Graph Laplacian
03/06
Project 2 - Due, PS3 - Out
EECS 445: Introduction to Machine Learning
Page !4
Lecture
Topics Covered
03/09
Midterm Review
03/11
Midterm
03/16
Generative Models
Gaussian Mixture Model; EM1
03/18
Generative Models
EM2
03/20
PS3 - Due, Project 3 Out
03/23
Generative Models
Bayesian Nets
03/25
Generative Models
Hidden Markov Models; Inference Problems; Viterbi and
Forward-backward algorithms
03/30
Generative Models
Hidden Markov Models; Baum-Welch; Model Selection
04/01
Graphical Models
Latent Dirichlet Allocation; Applications & Extensions
04/03
Project 3 Due, PS4 - Out
04/06
Special Topics
Deep Learning, Multi-Layer Neural Nets 1
04/08
Special Topics
Deep Learning, Multi-Layer Neural Nets 2
04/13
Special Topics
Reinforcement Learning
04/15
Applications
Examples from Research & Industry
04/17
PS4 - Due
04/20
Final Review
EECS 445: Introduction to Machine Learning
Page !5