Project suggestions for CS 4033/5033 Fall 2008

The following list of project suggestions is intended to get your project ideas started. Please feel free to come up with your own project ideas as well! Also, feel free to take one of these projects and modify it. Examine the previous year's suggestions for even more project ideas.

Precipitation type classification

• Ever wonder how a meteorologist classifies what type of precipitaton is falling in a winter storm without standing outside at every location? As you know from Oklahoma, precipitation type can vary widely within a very local area.
• Current radars do not provide much data for automatically determining the precipitation type of what is observed but new polarmetric radars are being tested and will have enhanced information available.
• Norman has the only polarmetric radar in the nation right now and the AMS AI committee has a contest on classifying polarmetric radar data based on hundreds of volunteer ground observations. The current automated approach has nearly no skill - try your hand at beating the approach and winning the contest!

Automatically detecting heatbursts and drylines (and fronts?)

• Living in Oklahoma provides us with some of the strangest and fastest weather changes in the country. In addition, Oklahoma is home to the densest observing network of ground-based weather stations in the country. In this project, you would combine information from the Oklahoma Mesonet with stored radar data to understand and automatically detect various aspects of severe weather. Ideas include but are not limited to detecting drylines, heatbursts, and fronts.

Understanding the dynamics of severe storms

• I have over 6 TB of simulated supercell thunderstorms available for study. I have several research projects using this data already and am quite open to ideas. Suitable machine learning project ideas include learning bayesian networks to predict storm dynamics (likely a dynamic bayesian network), learning decision trees to predict a particular aspect of storm classification, learning a probability model of how storms move, etc. Talk to me for more ideas if you are interested!

Viral networking (or other social networking projects)

• Imgine developing a new product at a startup company. You believe the product is great but you need to find the best way to market it. It is particularly well suited to a particular population and you need to find the best people to give your limited number of demo products to. In this project, you could analyze your social network (e.g. a facebook network?) to find the best people to market the product to.

Facebook app

• I'm sure you can think of fun applications for facebok that use machine learning. Think about your social networks and what you might be able to do with them!

Adpative personal robots

• Imagine having your own personal robot to help around the house, do laundry, cook dinner, etc. Or imagine another task where a robot could help you such as an office assistant. This is a possible future and it requires students to understand issues of how to create an adaptive robot.
• For this project, I'd like to see how well the robots from the Institute for Personal Robots in Education work as an educational platform for machine learning (and maybe artificial intelligence). One project idea is to have a team of robots playing capture the flag but many other ideas are possible. I will purchase the robots (to be owned by OU).

Bions

• The bions represent an exciting collaboration between art and science. To see pictures and movies, I highly suggest the following page.
• Projects with the bions would involve coding on the actual bions. The processors can run c code and you have a limit of 8K of code with only 1K of memory! An interesting limitation for an ML algorithm and one that will require your creativity to work with!
• Potential projects include: learning to change the song so as to attract the viewers and to retain them once they arrive, modeling neurons, hebbian learning, and many more.

Blimp

• Projects with the blimp are similar to projects with the helicopter except that the blimp is less dangerous (no rotating blades!). All control would still need to be performed off-board. The blimp has a PTZ camera attached.

This page written by amcgovern [at] ou.edu.

Last updated: May 23, 2016 1:59 PM