ECE 544NA Fall 2016: Project

Note on Team Projects

You are allowed to work up to groups of 2. The project's scope and report length will be scaled by 2 if you decide to do so. (i.e. report length for groups of 2 will be 8 pages + references, and the expected amount of work will also be doubled)

Project Proposal (10 points), due 11/14/2016:

There are two options for the project:

Explore the revelvant models and techniques covered in this course on a problem of your interest. This could be releated to your research.
Reproduce an existing work related to the materials covered in this course.

The project proposal should contain the following parts:

Introduction and Project Descriptions
Proposed Method
Dataset Description
Proposed Experiments
Resource Feasibility
Tentative timeline and the necessary steps
Referenes

Handing in assignment on compass

The project proposal should be in pdf format. File name should be net_id_project_proposal (e.g yeh17_project_proposal.pdf)
For groups of two, only one team member should submit the project proposal. File name should be net_ids_project_proposal.pdf (e.g. yeh17_yeh17_project_proposal.pdf)

Written Report + Code Submission (90 points), due 12/7/2016:

The written report should be in the form of a conference submission, we will follow ICASSP 2016's submission format [link]. The written report should contain the following parts:

Background Section (15 points): Describing the content of at least one interesting article from the pattern recognition literature (machine learning, signal processing, or some related area)
Method Section (20 points): Overall method on how you achieved to solved the proposed problem, and a bit of original derivation that has some relevance to what you're trying to accomplish. This could just be writing outin more details of the derivation in the original paper
Experiment Section (15 points): Describing the experiments you ran and the results.
Conclusion and Future Work (5 points): Discussion and future work.
References (5 points): List of references.
Code Description + Code (25 points): Describe the code your wrote, and its revelance to the project. Note: You are allowed to use tool boxes, however you are NOT allow to simply run some github project and use the results. You are required to write your own code, significant similarity in code of unreferenced sources will result in loss of points.

Handing in assignment on compass

The project report should be in pdf format. File name should be net_id_project_report.pdf (e.g yeh17_project_report.pdf)
For groups of two, only one team member should submit the project report. File name should be net_ids_project_report.pdf (e.g. yeh17_yeh17_project_report.pdf)
The project code should be in zip format. File name should be net_id_project_code.zip (e.g yeh17_project_code.zip)
Note: Do not zip any data or large files alonge with the code. You should provided instructions on how to obtain the data, or provided a link on dropbox, google drive...etc.

List of topics

Kingma, D. P., & Welling, M. (2013). Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114. [link]
Goodfellow, Ian, et al. "Generative adversarial nets." Advances in Neural Information Processing Systems. 2014. [link]
He, Kaiming, et al. "Deep residual learning for image recognition." arXiv preprint arXiv:1512.03385 (2015). [link]
Kingma, Diederik, and Jimmy Ba. "Adam: A method for stochastic optimization." arXiv preprint arXiv:1412.6980 (2014). [link]
Jayanth Koushik, and Hiroaki Hayashi. "Improving stochastic gradient descent with feedback." arXiv preprint arXiv:1412.6980 (2016). [link]
Gatys, Leon A., Alexander S. Ecker, and Matthias Bethge. "Image style transfer using convolutional neural networks." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2016. [link]
Coates, Adam, and Andrew Y. Ng. "Learning feature representations with k-means." Neural Networks: Tricks of the Trade. Springer Berlin Heidelberg, 2012. 561-580. [link]
Glorot, Xavier, and Yoshua Bengio. "Understanding the difficulty of training deep feedforward neural networks." Aistats. Vol. 9. 2010. [link]
Ba, Jimmy Lei, Jamie Ryan Kiros, and Geoffrey E. Hinton. "Layer normalization." arXiv preprint arXiv:1607.06450 (2016). [link]
Goodfellow, Ian J., Jonathon Shlens, and Christian Szegedy. "Explaining and harnessing adversarial examples." arXiv preprint arXiv:1412.6572 (2014). [link]
Springenberg, Jost Tobias, et al. "Striving for simplicity: The all convolutional net." arXiv preprint arXiv:1412.6806 (2014). [link]