The University of Tennessee at Knoxville

Department of Electrical Engineering & Computer Science

 

 

ECE 517: Reinforcement Learning in Artificial Intelligence

 

Fall 2015

 

Course Homepage: ECE-517

 

General Information

 

Lectures: T/TR 11:10 12:25 PM (MK 405)

Instructor: Dr. Itamar Arel

E-mail: itamar@eecs.utk.edu

Office: MK 608

Office Hours: T/Tr 2:00PM 3:00PM

Tel: (865)-974-3891

 

TA: Aaron Mishtal Office located in MK 606

Office hours: MWF: 12:00 2:00pm

email: amishtal@utk.edu

 

Course Description

 

Reinforcement learning (RL) is an exciting and relatively new machine learning discipline, which corresponds to a broad class of methods that allow a system to learn how to behave in environments that are incompletely known/specified based on reward signals. A key concept in RL is that the intelligent agent learns autonomously, based on acquired experience, rather than by being externally instructed or supervised. This course will address various topics pertaining to the theory and practice of RL.

 

Prerequisites

 

Background in linear algebra and probability theory, as well as knowledge of the Matlab programming, are required.

 

Reference Material

 

Lecture notes will be posted and made available at the course website. The majority of the material will appear in the lecture notes and supplemental reading material (papers). The course textbook will be:

m  R. Sutton and A. Barto, Reinforcement Learning: An Introduction, 2015, 2nd Ed. (available online at: http://webdocs.cs.ualberta.ca/~sutton/book/the-book.html)

Projects

 

The course will include two small projects and a final project. The final project will also involve an in-class presentation by the students.

 

Assignment Sets

 

There will be five assignment sets issued. Students are expected to complete all of the assigned problems. Assignment sets are to be submitted at the beginning of the class in which they are due.

 

Exams

 

There will be one in-class midterm exam. No make-up exams will be offered.

 

Syllabus

 

The following topics will be covered:

q  Foundations of machine learning & autonomous cognitive systems

q  Markov decision processes (MDP)

q  Neural networks

q  Dynamic programming

q  Temporal difference (TD) learning

q  Monte Carlo reinforcement learning methods

q  Eligibility traces

q  Hardware & software implementation considerations

q  Deep Reinforcement Learning

 

Final Grade

Average

A

90 100

B+

87 89

B

80 86

C+

77 79

C

70 76

D

60 69

F

0 59

 

 
Grading Policy (tentative)

 

Small projects (2 12.5 points) = 25%

Assignment sets (5 5 points) = 25%

Midterm exam = 20%

Final project = 30%

 

 

 

Students are encouraged to frequently visit the course website for announcements and updates.

 

Last Update: August 18, 2015