Alexander Lex - Student Mentoring

Interested in a VIS Project?

data visualization; visualization for molecular biology

Version 1, Dec 26, 2015

Visualization is an exciting and recently also an immensely popular field. If you're considering a project in visualization, take a look at my website and the Visualization Design Lab website and read a couple of papers to get an idea about the work we do. You should also check out the visualization class website, but you should already be familiar with that! And finally, here is a list of current and former student projects.

Over the years I've supervised many student projects. I've seen some really great work that turned into first-class research papers, but I've also seen students struggle and even occasionally fail. The projects that don't go too well often fail because of similar reasons. One of these is a lack of understanding of the expectations and the workload, an other one is a lack of structure. And while I try to be clear and transparent with my students, it is helpful to have a written document describing the process and my expectations. This is that document.

Why do a research project?

First, here is a list of wrong reasons to do a research project in visualization:
  1. You want credit for a research-class, and figure doing something like this is less work than a regular class. It is not. A project is probably more work and you have no way to hide in the masses: I will know how much time and effort you have put into the project. You need to stick to a schedule and be prepared to demonstrate progress constantly.
  2. You'd like to graduate with honors and thus have to write a thesis, but are not particularly interested in research and are not willing to put in a significant amount of work. Don't do this just for the grade, and be warned that this is not an easy path.
If you find your motivation falls into those categories, my advice is: don't do a research project. You and your adviser will be frustrated and you will get a bad grade. To contrast that, here is a list of good reasons to do a research project in visualization:
  1. You'd like to get research experience. If you're considering a PhD program you should definitely try to do as many research projects as you can. And you should excel at them. Not only will you find out whether this is the right thing for you, you also build a relationship with an adviser who then can write a good letter of recommendation for you, which is critical for admissions to PhD programs.
  2. You love to write code. Not every project has to be scientifically novel. If you like to develop interactive user interfaces using a modern web-stack, we can talk about an implementation-heavy project. This will be valuable for you because you will experience what it means to build larger and more complex software than you would in the class. In an implementation-focused project, I might ask you to port or re-implement an existing technique. This is typically not an option for thesis-oriented projects though, since they require independent conceptual work.
  3. You'd like to publish a paper. This is related to point #1, but here I'm assuming that you already know that research is something that interests you. A paper-worthy project is usually significantly larger in scope than a project under #1, and of course the result has to be novel. I expect a different level of dedication to the project if this is your goal. Also, transitions between #1 and #3 are possible, depending on success. You might want to publish a paper to strengthen your grad-school application, or you might just want to get the experience. You can either take the lead on a paper yourself, or, if I already know you, I can involve you in a research project in the group.
  4. Visualization is a great area to be in. While these other points could equally apply to any computer science project, if you work with me you will work on a visualization project. Modern visualization is at the intersection of human computer interaction, data science, and computer graphics. You will learn to develop new user interfaces and interaction techniques for cutting edge systems and might get to work with domain scientists, interesting datasets, etc. You will also have an opportunity to apply machine learning, statistics and algorithms in visualization. Finally, you'll get to work with the latest web-technology, which will be a great skill to have on the job marked.
I'm happy to work with students for all of these reasons, but I'm most excited when you want to publish a paper. Typically, however, students first work with me on #1 or #2 and then continue with another project in the #3 category. I have two rules that you must meet before I take you on as a project student:
  1. You have to have taken a visualization class and have done well in it. A one on one student-teacher relationship is not the right framework for getting to grips with the basics of visualization. You should know the theory and also have practical experience through homeworks and/or final projects. I will ask for evidence of this if you haven't taken my class.
  2. You know how to write code independently. I unfortunately often don't have time to help you with the technical details. There are other people in my team that might be able to help out, but they might not be local and it might take time.
Another rule applies only for paid RA-ships:
  1. I'm not going to pay you as a master's student if I haven't worked with you before. An RA-ship is a significant financial commitment and I will only occasionally consider this if you have completed a very successful project with me or have been an outstanding TA in my class. I might make exceptions to this if I have external funds and need to deliver a project, but then you'll have to still meet the standard project rules. RA-ships imply that you're working on a project with a clear goal to write a paper.

Don't be discouraged!

These guidelines are here to help you understand what a project entails and what I expect from project students. That said, it can be a lot of fun and a very rewarding experience. If you read all of this and can imagine doing a project like that - please reach out - I'll be glad to work with you! Also, check out my guidelines of how a project should be structured.