Monson H. Hayes
If you choose to do a project, this will be 90% of your grade. You can do it either individually or in a group of two. It is expected that the project will be extended in scope and difficulty if two work on a project together.
A good project is one that applies one or more machine learning algorithm or adaptive filtering algorithms covered in class, in novel ways to a dataset or an application such as recognition, tracking, or data mining. An excellent project is a research project that will result in a paper at a major conference, but this is not a requirement.
The project will provide you with a unique opportunity for exploring one or more areas of machine learning or adaptive filtering that we did not cover in depth. Some examples are graphical models, collaborative filtering, inductive logic programming, topic models and deep learning. You should choose a data set, apply machine learning techniques from these fields to it and compare their performance with the techniques covered in class.
If you want to tie the class project to your research project, you are strongly encouraged to do so. However, you should be able to demonstrate novelty. Simply applying an algorithm to a sub-problem in your research project is not acceptable.
Consider using sample project topics from the following sources: