Why this site?

During my final year at Bowdoin, I’ll be working on an Honors Project: a year-long independent research project that acts as the culmination of my academic studies. I’m doing my project under the Computer Science department, advised by Professor Eric Chown.

The honors project can be the culmination of a student’s academic experience at Bowdoin and offers an unparalleled chance for intellectual and personal development.

Pretty cool.

I’m writing this the summer before my senior year, when the project is being fleshed out but work on it has not begun (clearly, since this is the first entry in the lab notebook). The project brief is, informally, to investigate the viability of synthetic training data for machine-learning models. More specifically, this means:

  1. Determining the best way to generate synthetic training data
  2. Testing different amounts and types of synthetic training data for different scenarios in which machine learning is used
  3. Coming to a broader conclusion about when and how synthetic data should be employed.

As this is an academic project, I need some way of recording progress and keeping track of resources. Most projects in the sciences have lab notebooks associated with them. This is a project in the computer sciences, so it has a computer lab notebook. Pretty fancy!

Why this project?

I’ve been part of the Bowdoin RoboCup team throughout my Bowdoin tenure. We compete in international robotic soccer tournaments; the robots play completely autonomously using input from cameras.

Ball detection is a huge part of RoboCup success, and since the SPL changed the ball from a bright orange one to a black-and-white one, it has been one of our weak points. We’ve investigated rewriting our ball detection algorithm to use machine learning, but ran into the challenge of not having enough training data. As we played around with synthetic training data, we realized there was a lot more research to be done, research that would easily translate into something more broad.

What to expect?

I’ll be posting journal entries most days I work on the project, long-form posts when I am able to come to longer conclusions, and resources I use (books, articles, videos, etc.) to help me in my learning. This will be the place to keep up with my project, and will be a place where I, when writing the final paper, will reference to figure out what I did all year.

Seems pretty ambitious!