Novelty: 20 pts
The project is innovative / creative / interesting / tackling a fundamental research question. Examples of this could include any of the following: a) exposing a previously unknown connection between two different methods, b) exploring a component of current methods that is not well understood, c) proposing an improvement to an existing method, or d) interpreting or analyzing the learned mechanisms and representations in an existing model.
The project is a creative/important application of deep learning that provides novel scientific insights into deep learning. Applying an existing method to a new dataset and reporting results is insufficient. Examples of ways to achieve sufficient novelty: a) exposing fundamental limitations of current deep learning methods and experimentally or theoretically analyzing why the method is underperforming, b) exploring how additional structure or knowledge from an application area can be built into deep learning methods to improve results.
Technical Soundness and Content: 30 pts
The methods are sound, the math is correct, and conclusions are justified by the results.
The experiments/analyses are well designed to test the scientific claims. Uncertainty is property quantified.
The project is well scoped for the course (i.e., not too ambitious to be finished within the timeframe and with available resources).
Clarity: 20 pts
The project proposal is well written and structured, and is made understandable to the reader. The figures are clear and help the reader to understand your analysis.
Is the motivation clear? Is there a discussion of implications and limitations? Is there a clear conclusion based on the experiments in the paper?
Literature Review: 20 pts
Is the proposed project well motivated by a gap in prior research? Please include citations to prior work/papers to indicate the gap in previous literature that your project addresses.
Formatting: 10 pts
You should include the following content:
- An introduction or motivation
- Background and related work with literature cited
- A description of your methods and experiments with figures showing the method or setup
- An analysis of the results of your experiments with figures showing the results
- A conclusion or discussion, with mention of limitations
