Welcome
Hello and welcome to my home page! I am a Computer Science Ph.D. candidate at the MAPLE laboratory in the University of Maryland, Baltimore County (UMBC). My advisor is Dr. Marie desJardins. My research interests span a large range of Artifical Intelligence and Machine Learning. Below you can find past and current projects that I have worked on. Publications from these projects can be found in the publications section of the website. I also am an instructor at UMBC and links to course webpages that I teach can be found in the classes section of my website.
Research Projects
Hierarchical Skill Learning for High Level Planning
This is my dissertation topic. The goal is to create an agent architecture and set of algorithms, called Skill Bootstrapping, that allows an agent to start from a set of low-level primitive actions, and learn reactive skill hierarchies. Complex goals can be solved by using heuristic search planning that is abstracted to the level on which learned skills operate. Currently, I have developed an architecture that describes how Skill Bootstrapping will work, and it was presented at a ICML 2009 workshop for Abstraction in Reinforcement Learning, the ICAPS 2009 Doctorial Consortium and the AAAI 2010 Doctoral Consortium.
Game Theory Tournament Website
This work is focused on creating a centralized website for game theory tournament playing that will provide a centralized source of agents and games for researchers and educators to interact with. The aim of this project is to create a website that allows users to create different game types, hold custom tournaments, and submit agents that they, and others, can easily access to experiment with. A paper on this work has been published in FLAIRS-23. You can download the software from here. We plan to have a centralized site up soon.
Confidence-based Feature Acquisition
This research project focuses on developing novel machine learning classification algorithms when the features describing instances in a data set are not all initially available and may come with some measurable cost. The goal is to create an algorithm that can determine what features any given instance needs to acquire at both training and run time, such that accurate classification may be made with a minimum cost. This work has been published at SDM 2010.
Interactive Visual Clustering
Research project designed for identifying and automating a user's interactive clustering of data with a graphical interface. Users are presented with a graphical display of data that they can cluster by dragging data items to different areas of the screen. The program infers the user's intent and moves the remaining data items for them. I began working on this project in 2006 and the first paper resulting from it was accepted at the Intelligent User Interfaces Conference in January 2007. Additional research resulted in a book chapter published in 2008.