• Your presentation should introduce a working example of deep learning code, explain the theory, and illustrate the application. It's probably a good idea to prepare a short homework for the other students so they can come to your talk prepared. For example, if you use IPython notebook for your demo, you could upload the notebook beforehand and ask the students to carry out a task using the code in your notebook.

• You should use one of the datasets provided by the tutorial, and one other from outside the tutorial.*

• You may want to choose a presentation topic to match your final course project. This is optional.

• The core code we want to cover is the one in the Python tutorial. This is broken into the following topics.

1. Logistic Regression
2. Multilayer perceptron
3. Auto Encoders, Denoising Autoencoders
4. Stacked Denoising Auto-Encoders
5. Restricted Boltzmann Machines
6. Deep Belief Networks

A simple approach would be to pick one of these. You can choose another topic too, for example one of the systems that are featured in the deep learning demos we went through. Also see the section on Additional Readings.

• When you've settled on a preliminary topic, please edit the course Wiki and put down your name and topic. Then we'll see what gaps and conflicts arise and resolve these to come up with a final schedule for presentations.