Research and Publications

This page contains links to some of my research (primarily in information and computer science and mathematics, but with the occasional dash of social and political science, among other things).

These are some of the research groups of which I am a part. More information on each is below.

• The UCI Datalab is the research lab of my advisor, Professor Padhraic Smyth.
• UCI KDD Project
• The JUNG Framework

• Code

  You may download, use, modify, and redistribute code to your little heart's content, subject to the terms of the GNU Public License. (This does not apply to JUNG, which is distributed under the BSD license.) I also ask that you keep my name on it, a copy of the disclaimer, and add a record of any changes that you may have made. (No, you may not edit the disclaimers to remove the humor.)

• Papers/Writing

  You may download the papers for your personal use, but I insist that they remain unchanged (in particular, my name must remain on them, and I do retain copyright). If you find errors, let me know.

Format

Some papers on this site are in PostScript format. GSView is a decent free PS viewer, but if you absolutely must have a PDF copy instead, contact me and I may be able to arrange it for you.

Learning Predictive Models for Link Formation

EventRank: A Framework for Ranking Time-Varying Networks

JUNG (Java Universal Network/Graph) Framework

In February 2003, two other PhD students at UCI (Danyel Fisher and Scott White) and I created the JUNG Project. JUNG is an open-source Java software library, hosted on SourceForge, that provides a common language for the modelling, analysis, and visualization of data that can be represented as graphs or networks. For more information, or to download JUNG, visit the project's web site.

JUNG Framework Technical Report

The UCI KDD Project: Entity-Based Data Mining from Spacio-Temporal and Text-Based Data Streams

Web Log Data Mining and User Modeling

Neural Network-Based Strategies for the Iterated Prisoner's Dilemma

In early 2002, I investigated the design of (artificial) neural networks to play the iterated Prisoner's Dilemma (IPD), for a class on neural networks taught by Pierre Baldi. Out of this came the research paper noted above.

In the course of this project I wrote a collection of neural network and IPD-related investigative tools in Java. The code is currently shockingly undocumented (next mini-project: javadoc!), but I did take some care in the design. However, the usual disclaimer applies: if it does something horrid to your computer, that's your problem, not mine. (I don't expect that it should--it's fairly harmless, really--but I make no guarantees.)

Here are some sample description files that may be used to create various PD players (specifically, neural network-based strategies and periodic strategies).

Fuzzy Term Expansion and Document Reweighting

In the fall of 2001, I took a class in information retrieval from Wanda Pratt. In this context, I came up with an idea for doing query term expansion with a novel metric (based on a term association graph) in combination with a mechanism for reweighting documents based on the calculated relevance of each term (basic and expanded) to the query; this project resulted.

Here is the related work paper that I wrote for the project. Note that the name of the technique changed between this paper and the current version.

The allicaMet Fan Club: PIRATE

This is a project that I did with Joannie Humphreys (a fellow graduate student) in the spring of 2001. It is a prototype of a peer-to-peer system for song data exchange, whose features include mechanisms for payment, security, and reputation. It was built on the PROEM platform (written by Gerd Kortuem, a PhD student at the University of Oregon), which is an application framework for mobile ad-hoc networks.

(The name of the "research group" and project are somewhat, but not entirely, tongue-in-cheek...and by the way, that's pronounced "pee-rah-tay" if you don't happen to share our sense of humor, and "pie-rit" if you do.)

Distributed Models of Reputation Information

In 2000, I worked for Steve Fickas on a project whose purpose was to study the self-organization of people into groups to form tasks (club theory, in the language of political science). My part of it involved investigating the definition of reputation information, and its transmission. These are some working notes, in moderately rough form.

I feel that someone ought to continue this work, but it may not be me for quite some time.

Going Critical: An Investigation of Diameter-Critical Graphs

This is my Master's thesis on algorithmic graph theory. I did this work while at the University of British Columbia Mathematics Department, to satisfy the requirements of the UBC Institute of Applied Mathematics, under the supervision of Richard Anstee. It was completed on 30 April 1999.

(And someday, chapter 3 is actually going to get published. Really. I have Top Men working on it.)

Weak Context-Sensitive Independence and Approximate Inference on Belief Networks

This is my final project for a course that I took in 1999 at the University of British Columbia. The topic of the course was reasoning under uncertainty in the context of artificial/computational intelligence. At some point, I hope to develop this a bit further and publish it.

Traffic Signals and Weighted Star-Chromatic Numbers

This is the document that my group generated at the 1998 Pacific Institute for the Mathematical Sciences (PIms) Mathematical Modelling Workshop. (Yes, I know that would have been funnier if this were an algebra paper-- sorry). It's not terribly rigorous, but it does have some new results and interesting material.

traffic.c

This is the source code of the (portable, ANSI) C program that I wrote that implements two different greedy algorithms for calculating weighted chromatic numbers. (Yes, I said "chromatic", not "star-chromatic". Read the paper.)

Return to my home page

Last modified 30 August 2005