CSE 4392 - Special Topics: Neural Networks and Deep Learning
Description of Course Content:
This course offers an introduction to neural networks and deep learning. Topics include perceptrons, single-layer neural networks, multi-layer neural networks, Tensorflow and Keras, convolutional neural networks, transfer learning, deep learning methods for image classification, and sequential learning models for analyzing text. Auto-encoders and generative adversarial networks will be covered to some extent, as time permits. A strong programming and algorithmic background is assumed, as well as familiarity with linear algebra (vector and matrix operations).
Prerequisites: Admitted into an Engineering Professional Program. C or better in each of the following: CSE 3318 (Algorithms), CSE 3380 or MATH 3330 (Linear Algebra).
Student Learning Outcomes: After completing this course, students will be able to:
- Understand the concepts, and representation of the common neural network models and the most essential deep learning models and algorithms.
- Implement neural networks (from scratch).
- Use popular neural network libraries such as Tensorflow and Keras.
- Implement convolutional neural networks for image classification.
- Use transfer learning to improve classification accuracy.
- Implement and use neural networks for sequence learning, applicable to text.
- Understand the basic principles of auto-encoders.
Attendance is NOT mandatory for lectures. Attendance in lectures will NOT be used in calculating the semester grade. However, students are responsible for the material covered in the lectures. Lecture recordings will be available online on Microsoft Teams for students to review at any time.
Class participation during lectures is optional, and will not be considered for the course grade. At the same time, students are highly encouraged to participate, by asking questions, as well as answering questions by the instructor. Class participation can be an important resource for students who have difficulty understanding any part of the course material.
Students are expected to be professional and civil in their language and conduct:
For any student violating this policy, the instructor reserves the right to impose any grading penalties that the instructor considers appropriate, including a failing grade for the class, regardless of any other aspects of student performance. Examples of violations include language that is vulgar, insulting, disrespectful or threatening, making noise or talking with other students during lectures, disrupting lectures in any way, or making it difficult for other students to follow lectures in any way.
- During lectures.
- During office hours.
- In any oral, written or electronic communication with the instructor and TAs.
- In assignment submissions.
The semester score will be the weighted average of all assignment scores. The list of assignments specifies the weight for each assignment. There will be no exams (midterms or finals) in this course.
The final semester score, calculated based on the percentages listed above, will be converted to letter grades based on the following scale:
The instructor reserves the right to lower these thresholds, based on the distribution of scores. The instructor also reserves the right to lower a student's grade as penalty for violating the requirements of professional and civil conduct, as described in the student conduct section of this syllabus.
F: below 60%.
Any request for re-grading must be made within 5 days of receipt of that grade. Re-grading can lead to a higher or lower grade, depending on grading errors that are discovered.
There will be little or no extra credit. If there are extra credit opportunities, they will be included as part of the assignments, and they will be available to all students. There will be no make-up opportunities, and there will be no way for individual students to do extra work and improve their grade at the end of the semester.
IMPORTANT: It should be clear to every student that course grades will depend EXCLUSIVELY on the above grading criteria. Students should not request nor expect any other factor to be considered in computing the course grade. For example, factors that will NOT be considered are: need of a better grade to keep financial aid, to stay in the program, to qualify for a job offer, or to graduate. Students are expected to carefully monitor their own performance throughout the semester and seek guidance from available sources (including the instructor) if they are concerned about their performance and the course grade that they will earn. However, if the assignment scores are not good enough to warrant the desired grade at the end of the semester, there will be no other recourse for improving the grade.
The university withdrawal policy will be strictly adhered to. Up to the initial withdrawal date, all students will receive a W. After that date, the grade will be determined by the student's current average, and a WF or WP assigned as appropriate.
Expectations for Out-of-Class Study
Beyond the time required to attend each class meeting, students enrolled in this course should expect to spend an additional minimum of 9 hours per week of their own time in course-related activities, including reading required materials and completing assignments. Significantly more time may be needed for people having difficulties understanding the material, having a relatively weak mathematical or programming background, or having a relatively weak background in the prerequisite materials for this course.
University Policies and Services
All students enrolled in this course are expected to adhere to the UT Arlington Honor Code:
I pledge, on my honor, to uphold UT Arlington's tradition of academic
integrity, a tradition that values hard work and honest effort in the
pursuit of academic excellence.
I promise that I will submit only work that I personally create or
contribute to group collaborations, and I will appropriately reference
any work from other sources. I will follow the highest standards of
integrity and uphold the spirit of the Honor Code.
Instructors may employ the Honor Code as they see fit in their courses,
including (but not limited to) having students acknowledge the honor
code as part of an examination or requiring students to incorporate the
honor code into any work submitted. Per UT System Regents' Rule 50101,
paragraph 2.2, suspected violations of university's standards for academic
integrity (including the Honor Code) will be referred to the Office of
Student Conduct. Violators will be disciplined in accordance with
University policy, which may result in the student's suspension or
expulsion from the University.
The list of topics for future lectures is tentative, and will be updated as needed.
Emergency Phone Numbers: In case of an on-campus emergency, call the UT Arlington Police Department at 817-272-3003 (non-campus phone), 2-3003 (campus phone). You may also dial 911. The non-emergency number is 817-272-3381.
Back to the CSE 4392 home page.