Christopher DuBois

I am currently a first year PhD student in the Department of Statistics at University of California - Irvine, working with Padhraic Smyth and the DataLab.

My current research focuses on probabilistic models for relational events, e.g. communication or interaction within a social network.

Testing the Theory of Structural Holes For Sunbelt Social Networks Conference 2009 I presented an application of Carter Butts' relational events model to the Enron data set. The model is interesting because it allows one to test competing network formation mechanisms (participation shifts, preferential attachment, baseline activity levels, etc) and see which models best explain the history of network events. Here, I've parameterized the model using metrics related to structural holes theory in an effort to detect related behaviors in the dynamics of the Enron social network. [For more details, check the poster]

As an extension of the UCI Machine Learning Data Set Repository, I am creating a repository for network data that includes data sets from sociology, biology, the web and more. For more information, check it out: UCI Network Data Repository

I am coordinating an informal seminar series for those who want to learn more about the possibilities of the R system for statistical computing. People will be presenting tips on coding, debugging, data visualization, Sweave, parallel computing with R, and writing C extensions. For more, check out the website: UC Irvine R Seminar.

[resume]

I have recently won a National Defense Science and Engineering Graduate Fellowship.

Predicting the Release of Kinetics of Matrix Tablets In the winter of 2007 I participated in the Workshop on Applications of Mathematics to Biomedicine at the University of Otago in Dunedin, New Zealand. I worked with a small, multidisciplinary team to solve problems in pharmaceutical research, such as predicting drug release kinetics of common tablets. This work may be published as a working paper in coming months. [abstract]

HealthOne Staffing: I worked on some backend web programming (.Net, C#) for HealthOne Staffing, a travel nurse agency. In addition, I created a JavaScript widget that uses JavaScript, JSON, and Yahoo! Pipes that allows bloggers to display the company's available jobs and earn referral bonuses for nurses who end up taking an assignment. (August 2007 - August 2008)

Netflix Prize: Netflix lets its users rate movies and then predicts another movie that user may enjoy. They offered $1 million to anyone who can improve their prediction algorithm's accuracy (based on RMSE) by 10%. I've looked into hierarchical Bayesian networks and tried to implement an incremental SVD algorithm based on another competitor's work. (November, 2006)

Facebooking: I'm interested in the rise of social network websites (Facebook, etc) and their effects on my generation. What can we learn about one's "real" friendship network from the structure of their online network? I wrote a small Java app called FazeBooka using the Facebook API to output the user's friendship network in GraphML format. The goal was to visualize the data with JUNG. (March, 2007)

Tumor modeling (funding by HHMI): My primary undergraduate research, under Dr. Ami Radunskaya, focused on producing a cellular automata model of melanoma growth. The model correctly showed necrotic core development while considering only microenvironmental conditions (e.g. pH, glucose concentrations, and so on). We presented this work to a group of oncologists at UCLA. [more info coming] [preprint coming] [poster]

Image Analysis of Tumor Vasculature (Undergraduate Thesis): The blood vessels that grow near tumors are markedly irregular and "tortuous" compared to blood vessels in normal tissue. Under Dr. Vin de Silva, I used computational topology methods to compare scanning electron microscope images (ie. super high magnification) of vasculature in tumor tissue and normal tissue. [thesis] [presentation] [software]

Laser Tweezers (funding by SURP): So it turns out that tractor beams are possible - on a really small scale. Small styrofoam beads (~10 microns in diameter) can become "trapped" by the focal point of an infrared beam; when the beam moves, the bead moves. (Biologists use this tool all the time, eg. to measure the force of a sperm's flagellum.) We studied the trajectories of these beads by analyzing movies of the laser tweezer in action. [paper] [video]

Chaos and the Magnetron: The magnetron is a mesmerizing toy. I modeled the magnets as point dipoles and used a Lagrangian formulation for the fields, then numerically solved the equations of motion; the visualization reveals some of the mesmerizing feature of the real system. [paper]

Perry G. Schiro, Christopher L. DuBois, and Alfred S. Kwok. Large capture-range of a single-beam gradient optical trap. Optics Express, 11 (25) p.3485. 2003. [online]

Travel: I completed a six month trip around the world. For more on that, there's a blog and tons of photos. (December 2007 - June 2008)

Bicycle racing: During February-August 2007, I raced for an amateur cycling team in Spain. See the blog for more details.

chris [dot] dubois [at] gmail [dot] com
duboisc [at] ics [dot] uci [dot] edu