Assignment 5: Project proposal

Posting date: Sep. 20th
Due date: Sep. 27th
GitHub classroom’s submission link.

Description

In this assignment, you will write a proposal for your final class project. This will be a short written report of around 2 or 3 pages.

Submission format: You are strongly encouraged to use LaTeX for your project submission — this is good practice for your research life. However, I will also accept a submission in PDF format if you do choose to use other programs for creating your document.

There are two possible kinds of final projects.

If you are not sure about your project, it is your responsibility to contact me (and sooner is better than later), to discuss.

Alternative 1: Implement a recent vis paper

The first possibility for a final project involves you implementing a technique from a recent visualization paper. If the paper is an evaluation of an existing technique, your final project can be a replication of the original result and reanalysis.

Here’s some links you can use for inspiration:

There are other conferences that you might want to search for and use for inspiration: CHI, EuroVis, Pacific Vis, SIGGRAPH, NIPS, ICML, KDD, ICDM.

If you choose to go this route, your project proposal will be an extended abstract describing the paper you chose and your plan for implementation.

Possible ideas:

Implement a library that can support Szafir’s corrections for color perception.
Implement a system to visualize the behavior of MLP based on Relevance Layer Propagation.
Implement (and possibly improve) squares, a system for understanding multi-class supervised learning techniques.

Alternative 2: Identify an open vis problem, and make progress towards solving it.

The second possibility is more challenging, more rewarding, and more advantageous for your progress throughout grad school life. If you choose this alternative, you will work in a research problem of your choosing.

By the end of the course (that is, when you present your project), you will have produced an artifact that is evidence of your progress towards solving the research problem. This can be an implementation prototype, a sketch of a theoretical result, or a preliminary user study.

If you choose to go this route, your project proposal will be an extended abstract describing the research problem you want to solve, why and how you think it relates to data visualization, and your plan for solving the problem.

(In case you think this is overly ambitious, this style of class project has actually yielded a few results.)

Possible ideas:

Design a gamut correction mechanism that makes sense for visualization purposes. Correcting out-of-gamut problems by projecting to the closest color in RGB is easy but wrong. Doing it precisely in Lab color space is likely to be expensive, complicated to implement, and still non-intuitive. What are the solutions?
Design a system for automatic design of color scales for complex data modalities. As we discussed in class, we have separate systems for designing categorical scales,
How do you visualize the behavior of a GAN?
How do you visualize maps that change over time? OpenStreetMap maintains a version history, and it would be fascinating to have a way to visualize the many different versions. Specifically, it would be really fun to try to match changes in a given area to specific groups of users, etc. Here’s a tool that could serve as inspiration.

Assessment

You will be assessed on the quality of your writing and your research plan.

Note that I reserve the right to ask you to change your final project. The reason for this is simple: some fraction of students tend to choose projects that are too simple, and some fraction of students tend to choose projects that are too ambitious.