Mangalore!
Cycling around Mangalore, India

Matthew Whitehead

Hi! I'm an Assistant Professor at Colorado College in the Department of Mathematics and Computer Science. Here is my current CV.

If you're interested in computer science, then feel free to stop by my office (Tutt Science Center 206D) and chat about it.

If you're not interested in computer science, then feel free to stop by my office and we can chat about how to change that.


Current Class: Block 8: CP222 - Computer Science 2

Data structures!
Data structures and algorithms!

Learn about fundamental data structures that allow fast and efficient access to stored data. Also, learn about key algorithms for searching, sorting, and other related tasks.

Complete daily programming assignments to test your knowledge of programming and data structure use.




Current Research Focus: Deep Reinforcement Learning

My main research interests are in machine learning and artificial intelligence. I'm especially interested in applied ML: using machine learning to solve practical problems. I enjoy learning about new problems in all subject areas and figuring out how to use machine learning to solve those problems.

Reinforcement learning allows AI agents to learn from an enviroment. The environment provides feedback to the agent and then the agent can alter its actions to maximize its total reward over a longer time frame.

The reinforcement learning agent to the right is playing Tetris. The environmental feedback structure is very simple for the game: the agent gets positive feedback for each step of the game while the agent has not yet lost. Of course, losing means the agent doesn't get a positive reward (and will never get any future positive rewards either!).

The only input the agent receives is the state of the game, which is a snapshot image of the game before making the next move. In this case, the brain of the agent is a deep artificial neural network that uses convolutional layers to recognize visual patterns in the game images.

One funny/interesting thing about the Tetris agent is that it isn't very good until it gets close to losing. If you look near the bottom of the video, you can see the agent made a lot of bad moves early on. It is easier to make good moves when closer to losing because the chain of actions leading to a loss is shorter than near the beginning of the game.

I'm currently experimenting with deeper ANN models to see if they can solve Tetris and play as well as (re: better than) the best humans and the best hand-coded bots.




If you're interested in artificial intelligence, data mining, or any other cool CS topic, then come talk to me about it!

Some Papers


Personal Interests

When I'm not stuck at my computer I can usually be found bicycling, trail running, hiking, playing tennis/racquetball/badminton, playing guitar rather poorly, reading, playing video games, baking bread, playing board games (Scrabble, GrabScrab, and chess in particular), writing poetry and fiction of questionable artistic value, watching sports, or generally annoying my family Madhuja, Mia, Theo, and Helo.

Contact Information

Office: Tutt Science Center (TSC) 206D

Office Phone (email is preferred): 719-389-6536

Email:

Mailing Address (email is preferred):
Matthew Whitehead
Mathematics & Computer Science
Colorado College
Tutt Science Center
14 E. Cache la Poudre St.
Colorado Springs, CO 80903