The main objectives for the project presentation part of the class are: (1) you (as audience) will get an overview of current research problems in neural networks, (2) you (as a presenter) will get feedback for your project, (3) you (as a presenter) will have a chance to practice giving conference talks.
The project talks are 15 minutes + 5 minutes for questions, just like talks in most conferences. I will be moderating the session, and I will be ruthless, so make sure your talk fits the time slot. In the talk, you should give a brief motivation (What is the problem? Why is it important? Why is it difficult?), discuss your approach / architecture, present your results, evaluate the results (How well does it work? Why or why not? What can you do to fix it?) and discuss future work and extensions. Like in the topic talk, use lots of illustrations and examples and be provocative. If you want, you can meet with the TA and go over your talk, but a meeting is not required.
You should try to complete your project by the time you present it to class so that you can use the rest of the time to write it up. In any case, you must present at least preliminary results and a complete plan for wrapping the project up. The schedule of talks will be in the class schedule in due time. The talks are scheduled a few weeks before the end of the class, when we know who is working on what and with whom.