CSE 4311 - Spring 2025 - Assignments

All assignments must be submitted via Canvas.

This schedule is tentative, regarding both the dates and the topics. The initial plan is to have, in most cases, an assignment due every two weeks. However, some assignments may be broken into smaller components that are due every week. Any changes to this schedule will be communicated via e-mail, on Canvas, and on the course website.

Any assignment that includes the word "tentative" in its title or description is not finalized. When the contents of an assignment are finalized, the word "tentative" is removed from the title. When an assignment is posted on Canvas, it can be considered as finalized. However, even if an assignment is "finalized", the instructor reserves the right to make changes, so as to fix errors, add clarifications, etc. Such late changes will be announced via e-mail.

• Assignment 1.
  • Due date: Thu 01/23, 5:00pm.
  • Topics: review of prerequisite material on algorithms, algebra, calculus, programming skills, understanding of syllabus.

• Assignment 2.
  • Due date: Thu 02/06, 5:00pm.
  • Topics: Computing the output of perceptrons and neutral networks, gradient descent.

• Assignment 3.
  • Due date: Thu 02/20, 5:00pm.
  • Topics: Training neural networks with backpropagation.

• Assignment 4.
  • Due date: Thu 03/06, 5:00pm.
  • Topics: Keras, convolutional neural networks.

• Assignment 5.
  • Due date: Thu 03/27, 5:00pm.
  • Topics: Transfer learning for image classification, fully connected and recurrent neural networks time series classification.

• Tentative Assignment 6.
  • Due date: Thu 04/10, 5:00pm.
  • Topics: Text Classification, RNNs.

• Tentative Assignment 7.
  • Due date: Thu 04/24, 5:00pm.
  • Topics: Sequence-to-sequence translation with Encoder-Decoder RNNs, text classification with transformers and positional embeddings.

• Tuesday 04/29, 11:59pm: Deadline for all homework resubmissions (that will be considered towards achieving semester grades of B and below).

Assignment Policies

• Programming language: Programming assignments have to be done in Python. Python code needs to run on Colab. Typically, code that runs on Anaconda (Python version 3.8.3, numpy version 1.18.5 or later) also runs on Google Colab. If we cannot run your code on either Anaconda or Google Colab, your code will not considered correct. Any exceptions to these requirements must be provided via e-mail from the instructor or the teaching assistant.

• On using existing libraries and tools: The instructor recommends that students do not use existing libraries (except numpy, tensorflow, and keras) or existing tools in the programming assignments. The course slides provide sufficient information for students to complete their programming assignments without using special libraries. At the same time, if students want to use such special libraries or other built-in tools, they are allowed to. However, if these libraries or built-in tools do not provide results that match the assignment specifications, then the solutions will be counted as wrong. The instructor will not help the student before the deadline in determining whether any specific libraries/tools are compliant with the assignment specifications, since the recommended approach is to not use such libraries and tools. Oftentimes, existing libraries and tools implement variations of the methods and formulas that are incompatible with the assignment specifications, and any such variations will not be considered correct.

• Submissions: All assignments must be submitted via Canvas.

• Deadline extensions: No deadline extensions for the entire class will be provided, except in extenuating circumstances and/or when mandated by UTA. Students are encouraged to submit requests for individual extensions through the Student Access and Resource Center (SAR Center). The instructor will accommodate any such requests that have been approved by the SAR Center. In the absence of approval by the SAR Center, extensions for individuals can only be granted for emergencies documented in writing, and only when the student has clearly made efforts to inform the instructor as early as possible. The instructor may still refer such requests to the SAR Center before making a decision.

• Extra credit: Little or no extra credit will be provided. Any extra credit opportunities that are provided will be available for all students.

• Multiple submissions: If you make multiple online submissions for the same assignment, only the latest submission will be graded.

• Verifying your submission: After you submit your solutions, you should download them and make sure that you submitted the correct files. Every semester, several students ask for leniency, claiming that they did the assignment, but accidentally submitted the wrong files. These claims are often legitimate, but, unfortunately, no grade leniency will be accorded to such claims. It is each student's responsibility to doublecheck their submissions.

• In case of Canvas problems: If, for whatever reason, you cannot submit on Canvas, e-mail your solution to the instructor and the teaching assistant, from your UTA account, BEFORE the submission deadline. The time stamp on your e-mail message will serve as proof that you did the work on time. You still have to offer a convincing explanation as to why you were not able to submit online.

• Student conduct: Each student is expected to work on each assignment INDIVIDUALLY and submit his or her own work. The instructor will report to the Office of Student Conduct all violations of this policy, and all cases that are suspect of such violations.

Late submission policy:

• All assignments are graded out of 100 points. Assignments submitted late will be penalized, at a rate of 1 penalty point per hour. The submission time will be the time shown on Canvas. Any assignment submitted more than 100 hours late will receive no credit.
• Exceptions to late submission penalties will only be made for emergencies documented in writing, in strict adherence to UTA policy. For all such exception requests, the student must demonstrate that he or she made all efforts to notify the instructor as early as possible.
• Computer crashes, network crashes, software or hardware failure, Canvas failure, e-mail failure, will NOT be accepted as justification for late submissions. If you want to minimize chances of a late submission, aim to submit early. You can always revise your submission till the deadline.
• Sometimes students submit the wrong files online. Unfortunately, no credit or waiver of late penalties can be provided in such cases.
• If you find yourself in an emergency situation and can not deliver a homework on time, immediately inform the instructor and teaching assistant, via e-mail. Even if you have a valid reason for delivering late an assignment, you must make a convincing case that you have notified the instructor and teaching assistant as early as possible.

If you want to minimize chances of a late submission, aim to submit early. You can always revise your submission till the deadline.

Assignment Resubmission Policy:

The goal of the resubmission policy is to reward students who persist in understanding the material, even after they have received a low homework score. At the same time, the resubmission policy emphasizes the expectation that, for A-level performance, solutions should be correct at the time of first submission.

These are the rules of the resubmission policy for this semester:

• Students can resubmit any assignment on Canvas until the end of Tuesday, April 29. Please submit on Canvas, under the appropriate "HW #?? Resub" assignment, and please send an e-mail to the instructor and the TA to let them know (since these submissions can occur at any time, we do not regularly monitor for them).
• For the purposes of making a grade of A for the semester, the resubmission will NOT be considered. As mentioned earlier, the idea is that A-level performance should mean getting it right the first time.
• For the purposes of making a grade of B, C, or D for the semester, the resubmission score will replace the original submission score.