Final Project Instructions
Overview
The final project is your opportunity to synthesize everything you’ve learned in the course into a meaningful and original application. Your goal is to tackle a real-world or scientific problem, combining domain knowledge with computational modeling and machine learning in a rigorous, creative, and insightful way.
This project is not just about performance metrics; it’s about:
- Framing a well-posed scientific question
- Designing and analyzing machine learning models in context
- Iterating thoughtfully by analyzing results, going back to the data, and refining your model
- Demonstrating a deep understanding of your problem with a clear plan for future work
Final Presentation
The final presentation of your work will be on Wednesday, April 30 at 10:00am. The format is 5-minute presentation + 5-minute Q&A; you will get points for asking questions as well. The presentation should be roughly structured as:
- Motivation, question, and background
- Methodology and approach
- Key results and analysis
- Reflections and future directions Plan your time accordingly. 5 minutes is not a lot of time, so make sure to practice your presentation.
Final Report
- Due: Sunday, May 4 (at 23:59)
- Length: 6 pages (max), in a conference-style format (e.g., IEEE, NeurIPS, ICML)
- Contents: (order and structure is flexible. Choose what is most appropriate for your project.)
- Abstract
- Introduction and Motivation
- Related Work (with citations)
- Problem Formulation
- Methodology
- Results & Analysis
- Conclusion & Future Work
- References
- Link to GitHub Repository
The GitHub repo must contain: all relevant code, scripts, and notebooks. It should also have a README file with: setup instructions, description of the project structure, and clear instructions for reproducing results. Find tutorials online for how to set up a GitHub repo if you need to.
Evaluation Criteria
Your final project will be evaluated based on the following criteria:
Criteria | Description |
---|---|
Creativity, Novelty, and Relevance | Is the project original, insightful, or tackling a meaningful problem? |
Report Quality | Is the report well-written, clearly structured, and within the 6-page limit? |
Literature Awareness | Have you reviewed and referenced relevant prior work appropriately? |
Technical Rigor | Is there a solid mathematical formulation, and is it suited to the dataset/context? Express yourself mathematically whenever possible. |
Use of Domain Knowledge | Have you incorporated prior knowledge into the model design and hypothesis space? |
Scientific Thinking | Did you iterate, test hypotheses, and refine your model based on results? |
Model Justification | Are the chosen methods justified in relation to the problem and data? |
Results & Analysis | Are the results critically analyzed and compared across models or approaches? |
Future Directions | Are thoughtful improvements or extensions proposed? |
Code Quality & Reproducibility | Is the code clean, documented, reproducible, and accessible via GitHub? |
Delivery & Follow-through | Did you fulfill the goals promised in your original proposal? |
Timeliness | Late submissions will be penalized |
Final Remarks
- Start early and ask questions early.
- Your depth of understanding matters more than flashy results.
- Ask questions. Meet with me or share roadblocks early if you get stuck.
Let this project be a chance for you to explore, create, and think like a scientist.