Outline
The concluding course evaluation in this class is based on a final project. This project includes a numerical implementation of the methods discussed in the class, their extension, or new approaches falling under topics covered in the course. The project will further include appropriate tests to verify the implementation. Students should submit a proposal including their intended project outline. The project will be submitted as a report and presented in the week of the final exams. Students can work individually or in a team. However, the project's workload must be proportional to the number of students in the group.
Deliverables
Deliverables for the final project include a proposal, a checkpoint report, implemented source code, a final report, and a presentation.
Proposal
Students need to submit a proposal outlining their project topic of choice. The proposal is expected to be no more than one page (format provided below) and should delineate the topic of choice, what they plan to do, what open questions they need to address, and what would be delivered at the end of the course. The topic can be related to the student's research, from a publication, or one of the suggested papers by the instructor (provided below).
For your convenience, following are some guidelines on what should be addressed in the proposal:
For your convenience, following are some guidelines on what should be addressed in the proposal:
- Describe the scope of the intended project.
- What are objectives and goals?
- What analytical work and implementation is required?
- What tests would you use to verify your implementation?
- What steps do you need to take from your current state to the completion of the project?
- What are your knowledge gaps and how you intend to fill them?
- If your team is more than one person, include a collaboration plan that describes each members expected contribution.
- What are the milestones in your project? Include a rough timeline.
Checkpoint
A checkpoint report is to ensure students' progress in the final project. It includes a report on the project's progress, accomplished milestones, and the next step. The report is not expected to be more than 2-3 paragraphs. Students should schedule a 5-10 minute meeting with the instructor to discuss their progress. The grading would be either all or none based on if the student has reasonable progress.
Source code
At the end of the semester, students need to submit their source code for the instructor's review. All group members must be present in the review session. Each meeting will take about 20 minutes. Each group member should indicate their contribution to the implementation. Grading for source code review would be based on the soundness and clarity of the implementation. Students are encouraged to comment appropriately in their code.
Final report
Each group should submit its project in the form of a final report. The final report should include an introduction that motivates the problem, a description of the implemented method, a mathematical explanation, and algorithms used in the project. Then, a separate section must present numerical experiments to verify the approach. Finally, a conclusion on the project and how it relates to the class goals needs to be included. Students should add all references used in the project in the bibliography section. If in groups, students should indicate each member's contribution in the acknowledgment section.
Final presentation
The final projects are to be presented during the week of the final exams. Presentations are conducted online. Each student needs to select a timeslot (to be determined) for their presentation. All group members must present a part of the final presentation. The presentation length should be proportional to the number of group members. Each student should present their work in 10-12 minutes. Presentations are graded individually based on the student's coherency and in a group for the clarity of their presentation.
Grading
The grading for the final project would be
- Proposal: 15%
- Checkpoint: 10%
- Final report: 35%
- Source code: 20%
- Presentation: 20%
Deadline
The deadlines for the final project are as follows:
- Proposal due: September 15 (Extended: September 20)
- Proposal finalized: September 22 (Extended: October 3)
- Checkpoint: October 27
- Reports and source code due: November 22
- Source code review: November 29-December 6
- Presentation: December 12-16
Suggested papers
- Dynamic graphs: Leonardi, Nora, and Dimitri Van De Ville. "Tight wavelet frames on multislice graphs." IEEE Transactions on Signal Processing61, no. 13 (2013): 3357-3367.
- Latent link prediction: Kipf, Thomas, Ethan Fetaya, Kuan-Chieh Wang, Max Welling, and Richard Zemel. "Neural relational inference for interacting systems." In International Conference on Machine Learning, pp. 2688-2697. PMLR, 2018.
- Dynamic graphs: Rossi, Emanuele, Ben Chamberlain, Fabrizio Frasca, Davide Eynard, Federico Monti, and Michael Bronstein. "Temporal graph networks for deep learning on dynamic graphs." arXiv preprint arXiv:2006.10637 (2020).
Harvard - Node clustering for material science: Webb, Michael A., Jean-Yves Delannoy, and Juan J. De Pablo. "Graph-based approach to systematic molecular coarse-graining." Journal of chemical theory and computation 15, no. 2 (2018): 1199-1208.
- Graph representation learning: Simonovsky, Martin, and Nikos Komodakis. "Graphvae: Towards generation of small graphs using variational autoencoders." In International conference on artificial neural networks, pp. 412-422. Springer, Cham, 2018.
- Graph generative models: Kang, Seokho, and Kyunghyun Cho. "Conditional molecular design with deep generative models." Journal of chemical information and modeling 59, no. 1 (2018): 43-52.
- Graph wavelet design: Hammond, David K., Pierre Vandergheynst, and Rémi Gribonval. "Wavelets on graphs via spectral graph theory." Applied and Computational Harmonic Analysis 30, no. 2 (2011): 129-150.
- Graph generative models: Zou, Dongmian, and Gilad Lerman. "Encoding robust representation for graph generation." In 2019 International Joint Conference on Neural Networks (IJCNN), pp. 1-9. IEEE, 2019.
- Graph regression for chemical property prediction: Gilmer, Justin, Samuel S. Schoenholz, Patrick F. Riley, Oriol Vinyals, and George E. Dahl. "Neural message passing for quantum chemistry." In International conference on machine learning, pp. 1263-1272. PMLR, 2017.
- Graph regression for chemical property prediction: Hy, Truong Son, Shubhendu Trivedi, Horace Pan, Brandon M. Anderson, and Risi Kondor. "Predicting molecular properties with covariant compositional networks." The Journal of chemical physics 148, no. 24 (2018): 241745.
- Graph learning: Battaglia, Peter, Razvan Pascanu, Matthew Lai, and Danilo Jimenez Rezende. "Interaction networks for learning about objects, relations and physics." Advances in neural information processing systems 29 (2016).
- Graph regression for chemical property prediction: Kearnes, Steven, Kevin McCloskey, Marc Berndl, Vijay Pande, and Patrick Riley. "Molecular graph convolutions: moving beyond fingerprints." Journal of computer-aided molecular design 30, no. 8 (2016): 595-608.