Description and Objective:
Inferring the values of independent variables becomes
exponentially more difficult as the number of variables
increases, so it is only possible if we can use structural
knowledge. One way to encode this knowledge is through
graphical models. This generalizes the notion of Hidden
Markov Models, which have found applications ranging
from speech recognition to protein sequence matching.
Graphical models have much wider applications, including
protein networks, machine vision, decoding error-correcting
codes, etc. We will use the book Probabilistic Graphical
Models by Koller and Friedman to study the properties and
applications of these models, and whether they can
be used to infer causality rather than just correlation.
This course relates to Machine Learning and Artificial
Intelligence, but does not assume them as prerequisites.
We will review basic probability, but it will be helpful to
have had prior experience working in a probabilistic setting.
Prerequisites: Comp 15 and either MATH 22 or familiarity with basic probability theory.
Text:
The textbook for the course is Probabilistic Graphical Models. Daphne
Koller and Nir Friedman, MIT Press(2009) ISBN: 978-0-262-01319-2
Instructor:
Anselm Blumer
ablumer (at) cs dottufts dot edu
Halligan Hall, Room 214
Office Hours: Mondays 11-12 and 5-6, Tuesdays 12-1 and by appointment.
Home page
Communication:
Students are encouraged to communicate frequently with the instructor and
TA regarding any issues with the course. Students are encouraged to use
email and office hours frequently. Any announcements regarding the course
will be made via the course webpage or in class so be sure to check it frequently
and be sure to get material for any class you miss.
Homework:
Homework will be assigned regularly in the course. While reading assignments
will not be directly assigned it is important that students use the textbook
to supplement their understanding of the material presented in the lecture.
The majority of the assignments will be written assignments due on Wednesdays
at the beginning of class on the due date specified. This work can be handwritten
with the assumption that these assignments are legible. (A student may be
asked to type their assignments if grading is not possible.)
Late Homeowork:
Because of the size of the class and the amount of homework 15% of the total
number of points for the assignment will be deducted daily. No homework
will be accepted after one week.
Exams:
There will be no exams.
Grade Calculation:
95% Homework
5% Class participation
Feedback:
Your thoughts and concerns on this course are important. You are encouraged
to give feedback to the instructor throughout the term. As always students
will be asked to fill out a course evaluation at the end of the term.
Academic Misconduct:
Students should read the Tufts brochure on academic integrity located at:
http://uss.tufts.edu/studentaffairs/judicialaffairs/Academic%20Integrity%2010-11.pdf
A few highlights are presented to emphasize importance:
Absolute adherence to the code of conduct is demanded of the instructor, teaching fellow, and students. This means that no matter the circumstance any misconduct will be reported to Tufts University.
While students are encouraged to discuss course materials, no collaboration is allowed on homework. Specifically you may discuss assignments and projects verbally, but must write up or work on the computer alone. In addition any discussion should be documented. An example on the homework would be "Thanks to Ray for helping me understand Kolmogorov complexity." Another important example is citing a source, this could be "This information was adapted from www.boston.com"
While computers enable easy copying and collaboration both with other students and materials from the Internet, it is possible to use these same computers to detect plagiarism and collaboration.
If any student does not understand these terms or any outlined in The Academic Code of Conduct it is his/her responsibility to talk to the instructor.