Other possibilities #

There are many ways you can conclude this semester by studying the interaction of machine learning and mathematics; my project stubs address only a few and there are many others. Some ideas:

• Learn about any of the the many other classical machine learning algorithms.
• Investigate the the use of machine learning to suggest theorems in mathematics.
• Investigate the Rashōmon effect in machine learning. As an example, this paper constructs four different machine learning models on the same data set, each offering a completely different interpretation of what is going on. Also see this article or
• What do the trajectories for SGD and ADAM actually look like? Replicate the experiments in this paper, or see what happens in a different domain.
• How do you steal someone else’s production models? Start with the video (it is a really good talk!) and then read the paper.