The goal of this assignment is propose and carry out an interesting research project using neural nets and deep learning. I have suggested project ideas in class, and several of you have described projects you would like to work on. My hope for this assignment is that you will produce novel results that will be the basis of a research article for a conference or journal.

Your project can be one of two sorts: (1) an idea for a novel architecture, learning algorithm, training procedure, etc.; and (2) a novel application of deep learning to a research problem in some area of computer science, cognitive science, or engineering.

You may work solo or with one other person. If you have a big enough project, you may work with 2 others, but speak to me first.

If you'd like to run project ideas past me, I'm happy to give you feedback.  Write up a brief 2-5 paragraph description of what you want to do and email it to me at my university address (not to piazza).

Hand in a compact paper that describes your results. Aim for a conference or journal style article, wherein you motivate the problem, describe the approach in detail, and systematically present results, then discuss the implication of these results.

This write up will be due in class on December 16 when we have our final exam time slot. Because I have only a few days after the due date to compute semester grades, there will be no extensions for this assignment.  Hand in what you have on December 16.

I'd like a subset of folks to present their results in class during our final meeting. If you'd be interested in presenting, contact me. I will consider in-class presentations as extra credit work.