COMPSCI 682 Neural Networks: A Modern Introduction

Note

There will be three assignments throughout the semester, focusing on theoretical and practical issues related to deep learning. The assignments are due just before midnight, at 11:55 pm, unless otherwise specified. The tentative due dates are listed below.

You will submit the assignments electronically through Gradescope (do not email your assignments). You will be automatically enrolled in Gradescope at the start of the semester.

More details on the submission process will be provided with each homework.

Late day policy:

  • Free late days: You can use 7 late days in total, with up to 3 late days for one assignment. Beyond 3 late days the assignment will not be counted at all. Once you have used all 7 late days, penalty is 25% for each additional late day. We will use your latest submission for grading and for calculating your late day usage. There is no bonus if you don't use late days at all. There is no need to ask for permission for late days – we will automatically figure late days based on the latest gradescope submissions. Delay of even a single minute counts as a full late day, so do not wait till the last minute to submit.
  • Documented late days: Beyond the seven “free” late days, we will only provide “verified” late days as required by university policy, with documentation (i.e. illness documented by a doctors note).
  • Undocumented late days: If you submit your homework late for any reason not covered by the above two policies, you may include at the top of your submission a justification for why it was submitted late. When assigning grades at the end of the semester, the instructor will consider all these justifications and may, at his sole discretion, waive some or all of the penalties applied. The instructor may also waive penalties for submissions that are only very slightly late (e.g. 15 seconds after the deadline). No feedback will be given before the end of the semester about how these decisions will be made. Please make sure to include this in the submission itself and not, e.g., by emailing the instructor.

Re-grading policy: Errors in grading of assignments and exams can occur despite the best efforts of the course staff. If you believe you’ve found a grading error, submit a re-grade request on gradescope. Re-grade requests must be submitted no later than one week after the assignment is returned. Note that re-grading may result in your original grade increasing or decreasing as appropriate.

AssignmentDue DateDescription
Assignment #1 Thursday, Sept 26, 11:55pm Image Classification, kNN, SVM, Softmax, Neural Network
Assignment #2 Thursday, Oct 24, 11:55pm Fully-Connected Nets, Batch Normalization, Dropout, Convolutional Nets
Assignment #3 Tuesday, Nov 26, 11:55pm Image Captioning, Self-Supervised Learning, Style Transfer