Preview Workflow

Viewing: ECE 765 : Probabilistic Graphical Models for Signal Processing and Computer Vision

Last approved: Wed, 17 May 2017 19:52:29 GMT

Last edit: Wed, 17 May 2017 19:52:29 GMT

Change Type
Major
ECE (Electrical and Computer Engineering)
765
032459
Dual-Level Course
Cross-listed Course
No
Probabilistic Graphical Models for Signal Processing and Computer Vision
Probabilistic Graphical Models
College of Engineering
Electrical & Computer Engineering (14ECE)
Term Offering
Fall Only
Offered Every Year
Fall 2017
Previously taught as Special Topics?
Yes
2
 
Course Prefix/NumberSemester/Term OfferedEnrollment
ECE792Fall 201629
ECE792Fall 201535
Course Delivery
Face-to-Face (On Campus)

Grading Method
Graded/Audit
3
16
Contact Hours
(Per Week)
Component TypeContact Hours
Lecture3
Course Attribute(s)


If your course includes any of the following competencies, check all that apply.
University Competencies

Course Is Repeatable for Credit
No
 
 
Edgar Lobaton
Assistant Professor
Assoc

Open when course_delivery = campus OR course_delivery = blended OR course_delivery = flip
Enrollment ComponentPer SemesterPer SectionMultiple Sections?Comments
Lecture3535NoN/A
Open when course_delivery = distance OR course_delivery = online OR course_delivery = remote
Prerequisites: Programming experience (MATLAB, C++ or other object oriented language such as Python), linear algebra (MA 405 or equivalent), and probability (ECE 514, equivalent or instructor permission)
Is the course required or an elective for a Curriculum?
No
Techniques for machine learning using probabilistic graphical models. Emphasis on Bayesian and Markov networks with applications to signal processing and computer vision.

There are a number of applications in signal processing that make use of probabilistic graphical models. Tools such as Hidden Markov Models and Bayesian Networks are standard tools for researchers. This course aims to give the students a better understanding of these tools so they can properly apply them to their various research areas. These topics are brought up in courses such as pattern recognition in the ECE department, but they do not go into the level of detail specified here including computationally efficient approaches for inference and learning.


No

Is this a GEP Course?
GEP Categories

Humanities Open when gep_category = HUM
Each course in the Humanities category of the General Education Program will provide instruction and guidance that help students to:
 
 

 
 

 
 

 
 

 
 

 
 

Mathematical Sciences Open when gep_category = MATH
Each course in the Mathematial Sciences category of the General Education Program will provide instruction and guidance that help students to:
 
 

 
 

 
 

 
 

Natural Sciences Open when gep_category = NATSCI
Each course in the Natural Sciences category of the General Education Program will provide instruction and guidance that help students to:
 
 

 
 

 
 

 
 

Social Sciences Open when gep_category = SOCSCI
Each course in the Social Sciences category of the General Education Program will provide instruction and guidance that help students to:
 
 

 
 

 
 

 
 

 
 

 
 

Interdisciplinary Perspectives Open when gep_category = INTERDISC
Each course in the Interdisciplinary Perspectives category of the General Education Program will provide instruction and guidance that help students to:
 
 

 
 

 
 

 
 

 
 

 
 

 
 

 
 

Visual & Performing Arts Open when gep_category = VPA
Each course in the Visual and Performing Arts category of the General Education Program will provide instruction and guidance that help students to:
 
 

 
 

 
 

 
 

 
 

 
 

Health and Exercise Studies Open when gep_category = HES
Each course in the Health and Exercise Studies category of the General Education Program will provide instruction and guidance that help students to:
 
 

 
 

 
 

 
 

 
&
 

 
 

 
 

 
 

Global Knowledge Open when gep_category = GLOBAL
Each course in the Global Knowledge category of the General Education Program will provide instruction and guidance that help students to achieve objective #1 plus at least one of objectives 2, 3, and 4:
 
 

 
 

 
Please complete at least 1 of the following student objectives.
 

 
 

 
 

 
 

 
 

 
 

US Diversity Open when gep_category = USDIV
Each course in the US Diversity category of the General Education Program will provide instruction and guidance that help students to achieve at least 2 of the following objectives:
Please complete at least 2 of the following student objectives.
 
 

 
 

 
 

 
 

 
 

 
 

 
 

 
 

Requisites and Scheduling
 
a. If seats are restricted, describe the restrictions being applied.
 

 
b. Is this restriction listed in the course catalog description for the course?
 

 
List all course pre-requisites, co-requisites, and restrictive statements (ex: Jr standing; Chemistry majors only). If none, state none.
 

 
List any discipline specific background or skills that a student is expected to have prior to taking this course. If none, state none. (ex: ability to analyze historical text; prepare a lesson plan)
 

Additional Information
Complete the following 3 questions or attach a syllabus that includes this information. If a 400-level or dual level course, a syllabus is required.
 
Title and author of any required text or publications.
 

 
Major topics to be covered and required readings including laboratory and studio topics.
 

 
List any required field trips, out of class activities, and/or guest speakers.
 

College(s)Contact NameStatement Summary
College of SciencesJohn BlondinCOS does not have any comments or concerns regarding this action. -John
Dr. Lobaton will be teaching this course once a year as part of his regular course load. If demand for the course drops, then the course will be taught every other year.

This course will cover the fundamentals of probabilistic graphical models. The students will learn about Bayesian network representations and undirected graphical models, and how to perform inference and learning using these representations. Students will learn to apply these techniques to signal processing and computer vision.


Student Learning Outcomes

By the end of this course, the students will be able to:


1. Explain and implement strategies for model assessment


2. Apply standard machine learning approaches for classification and regression in data streams and at the pixel-level in images


3. Explain the basic concepts behind Bayesian Networks, Markov Networks, Dynamic Bayesian Networks, and Hidden Markov Networks


4. Explain the concepts behind exact and approximate inference in graphical models


5. Explain how parameter estimation and structure learning is done for graphical models


6. Train and apply graphical models for applications on signal processing and computer vision


Evaluation MethodWeighting/Points for EachDetails
Homework30There will individual biweekly homework assignments, which will require some programming and mathematical derivations. The final homework grade will be computed by averaging all assignments.
Quizzes30There will be biweekly in-class quizzes. They will be approximately 15-minutes long. The final grade will be computed by averaging all scores after dropping the lowest grade.
Project40There will be two group projects, one of which will focus on time series and the other on computer vision. The objective of these projects will be to gain experience applying existing graphical model toolboxes to solving projects in the specific subject areas. Students will implement a baseline approach using standard machine learning techniques, and then compare their results against their implementations using graphical models. Grading will be based on the instructor’s review of the reports, performance of the code provided by the teams, and peer reviews of their technical reports.
TopicTime Devoted to Each TopicActivity
Overview of Probability and Machine Learning2The basic concepts from probability theory and machine learning will be presented. Model Assessment will be discussed as well.
Bayesian Networks2Introduction to the concepts of Bayesian networks.
Template and Dynamical Bayesian Networks1Topics will include Hidden Markov Models and Kalman filtering.
Undirected Graphical Models2Introduction to Markov networks.
Exact Inference2Techniques such as variable elimination will be introduced and Clique Trees.
Approximate Inference1Techniques such as Belief Propagation and Markov Chain Monte Carlos Sampling Methods will be studied.
Parameter Estimation2Maximum Likelihood Estimation and Bayesian Estimation will be studied.
Structure Learning1Techniques for learning structures within graphical models.
Applications to Signal Processing2Applications to physiological signal processing and computer vision will be presented.
mlnosbis 4/3/2017: Suggest consultation with College of Sciences. Contact Dr. John Blondin (john_blondin@ncsu.edu) to arrange consultation and enter consultation summary in the consultation field.
1) Catalog description on the CIM form should match that in the syllabus. This is what will be used for the SIS catalog description.
2) Office hours should be listed on the syllabus
3) Topical outline should be listed on the syllabus

pjharrie - 4/3/17 - so are their quizzes or tests (see Student Evaluation section)? There should also be a much clearer description of what the projects will entail even if the topics aren't specifically known. What are the projects designed to accomplish?

ABGS Reviewer Comments:
- Syllabus shows 13 week length
- Cost of books is needed on syllabus
- Breakdown of weights for assessment could be more detailed.
- Concern about overlap with Statistics. Consultation with Statistics indicated no overlap. This seems like a very good course and while it is mostly a statistics course, the ST program does not have plans to offer a similar one.

Edgar Lobaton (ejlobato) - 4/26/17
- All above items have been addressed
- Syllabus has been updated to add up to 15 weeks
- Cost of book has been updated in Syllabus
- More details on break down of grading have been added
dgyu (Mon, 27 Mar 2017 22:33:21 GMT): Rollback: Rollback so that changes can be made.
Key: 14950