CSC 390: Topics in Artificial Intelligence

Homework 5: Final Presentation

Due: in-class (see schedule on the calendar)

For this presentation, the goal is to present your final project topic in a way that is understandable and engaging for the rest of the class. It is an opportunity to practice oral presentation and communicating ideas in front of an audience. You are also welcome to solicit feedback or suggestions from the class. Note that no results are required for this presentation. In general, think about the final presentation as less about results and more about teaching the class something new in an effective manner.

The format is similar to the mid-semester presentations (15 minute slots, roughly 10-12 minute presentation and a few minutes for questions). The following sections can provide a template for your presentation.


Introduce your topic in a creative or visual way. Whenever you give a presentation, there will be those in the audience less interested in the topic than you are, who might question the "point" of your topic or thesis. Give them a reason to pay attention. Often this involves placing your topic in a larger context, using an image the audience can relate to, telling a personal story, or posing a question you'll answer later in the talk.

Make the high-level goals of your project very clear. What are you hoping to learn from the data or from the implementation of an algorithm? Set the project in a larger context - even if your results will not be completely novel or ground-breaking, how could someone build on your ideas in the future? Include key terminology as it makes sense for your project.

Supervised vs. Unsupervised Learning

At some point in your presentation, clearly explain how your project fits into the spectrum of supervised/unsupervised learning. There might be ways of analyzing your dataset that are supervised, but some parts that are unsupervised. Sometimes a certain feature of the data could be interpreted as a label, or could be thought of as simply another feature. It might make sense to put this part after an explanation of the data, or during the methods - include it whenever it makes sense for your project.


Explaining the dataset should be a significant fraction of your final presentation. Include a small example of what the data looks like if at all possible, ideally in a matrix form. Give concrete values for the number of datapoints (m), the number of features (p), and any labels. Clearly explain what the features are and how you plan to work with them.

If your project has more focus on implementing an algorithm, you can skip to the methods and spend more time there, then give an example of what data you might use to test your method at the end.


Give an overview of all the methods you plan to use on your dataset. If you're using methods we've discussed in class, you don't need to go over them again, but do explain how you will apply them to your dataset. For several projects, it is necessary to have a notion of distance between two datapoints. Explain the challenges and how you will define this distance. If you plan to use a method we haven't discussed in class, explain the algorithm at a high-level. Slides can be a great way to "layer" or "animate" an algorithm. Throughout the presentation, explain your thought process and how you moved past obstacles in the analysis.

If you are doing an implementation project, spend more time on the methods (you can even show some example of code/pseudocode and discuss implementation challenges).


You do not have to have any results for your final presentation. If you have preliminary results, that would be great to discuss. If not, discuss what you might expect the results to look like (figure, table, result of a comparison, etc).

Interpretation and Future Work

A very important part of your final project is the interpretation of the result. For example, what do the clusters mean? What do the hidden states mean in an HMM? It's almost always possible to run algorithm A on dataset X, but how can we make sense of the results? In this part, discuss how you might go about interpreting your results and what you hope to learn from them. At the end, discuss future work (sometimes that can be thought of as your TODO list) - you can also ask the class if anyone has ideas about challenges that are coming up in working with your data.

Email me your slides

Save your slides as a PDF file (no animations) and email it to me by 9:30am on the day you present (this will allow me to compile one set of slides so we don't have to keep switching to different computers). If you are unable to present on your assigned day, let me know as soon as possible. I will give extra credit for anyone willing to move up if we have last minute changes due to illness, etc.

Ask questions

The last part is to participate in the presentations of your classmates. Ask questions if you don't understand something they explain. Everyone should ask at least one question during our series of final presentations. Questions can help the presenter feel more comfortable and calibrate their pace.


Keep in mind the feedback from your mid-semester presentation and think about how you could improve. General advice: speak louder and slower than usual, speak to the audience and not to the board/slides, avoid lots of text on slides, pause half-way through for questions.