Project for CS 679

For those of you registered in CS 679 (the graduate version of Neural Networks), you will do a project instead of writing the final exam.

Find a Team

For this project, students will work in groups of three or four. You must find a group of three or four!

Expectations

In planning out your project, keep in mind that I expect at least a small amount of novelty. That novelty can come in the form of modifications to your implementation of someone else’s paper. Or, it could come in the form of a variation of a learning method we discussed. Or, it could come in the form of an application of a technique in a novel way.

Note that it is not sufficient to simply apply feed-forward neural-network learning to a new dataset. This is not a course on applied neural networks, so your project needs to contain an element of novelty in the learning method, or the architecture you choose. If you use a neural-network package (like PyTorch, Keras, or Tensorflow), then I would expect the architecture to be interesting, or you might use a cost function that includes multiple pieces, possibly involving different parts of your network.

Here is the project evaluation template that I use when grading your project report.

Deadlines for the project

Proposal Due: 5:00pm Friday March 8

The proposal is a short statement about what you plan to do for your project. It should be 1 to 2 pages in length, and outline what problem you want to solve, and/or what methods you want to use. Your project proposal will not be graded; its purpose is to make sure you are on the right track.

Project Report: 5:00pm Tuesday April 16

Your final project should be submitted online (into the Dropbox on D2L), and include a PDF of your project report. Only one member from each group needs to submit the project report. You should use the IEEE conference template to format your paper (you can get it from this web page, or on Overleaf). The report should not exceed 8 two-column pages in length, including figures and references.

Submit: Individual Contributions

You may submit a short statement describing your contribution to the project, explaining your impression of what percent of the work you did. Your contribution report is submitted privately to the instructor (on Piazza). Project marks for individuals might be adjusted from the group mark according to these contribution reports.