- You can work individually or in groups of 1-4 people. Larger groups will be graded more harshly than smaller groups.
- There are two main types of projects:
- Methodology implementation and testing: Implement a novel/difficult method from scratch and test with simulations. You will be judged primarily on the implementation and usability of the code.
- Data analysis: Implement a method on a dataset of your choosing, but the method should be novel to that dataset. You will be judged primarily on the selection of the methodology, the assessment of the method, and the presentation of the analysis.
You will be graded on the following criteria. To emphasize that you may 'excel' in different ways your lowest scores will be dropped, but it will be difficult to get the best grades in each.
- method implementation
- simulation / numerical experiments
- code efficiency and organization
- method novelty
- method selection
- method evaluation
- feature extraction and similarity measure construction
- data munging, extraction, merging, storage
- presentation of analysis / documentation of code
For example, a project that performs various binary classification methods on a social science dataset you may want to focus on data munging, method selection, method evaluation, feature extraction, and presentation of analysis. A project that implements L1 regularized support vector machines will focus on method implementation, method novelty, code efficiency, simulation, and documentation.