Hello, I received a Computer Science PhD at George Mason University in May 2018. My field of research is in Artificial Intelligence.
For my PhD I focused on a very narrow subfield of AI called multiagent systems or MAS for short. Abstractly, this field is focused on understanding, describing, engineering, and motivating the paradigm where autonomous agents coexist. Unlike in distributed artificial intelligence, where there are rules and message passing that are dictated by the architect in order for the problem to be solved, a MAS must solve the problem without an architect. Each agent must make its own decisions based on its own limited input and knowledge. This makes coordination difficult because agents can accidentally step on each others toes, so to speak, without realizing it. The goal of the system designer then is to create coordination mechanisms in order to achieve some objective. The main interest of MAS to me is the fact that it is such a rich problem that exists in real life. My dissertation focused on dynamic multiagent task allocation using a mechanism I developed based on the concept of bounty hunting. A number of my papers can be found on my google scholar page, my advisor's page (look for papers with Drew Wicke), and on AAAI's website.
My current work (2016-present) has focused on machine learning and time series analysis. I have developed TSAT (Time Series Analysis Tool) which is based on GrammarViz, but with a number of enhancements and additional algorithms for time series classification. I have also enhanced Shadow a network simulator improving its parsing and creating a new type of packet generation based on a Poisson process. I have also implemented the Extreme Value Machine (EVM) in order to begin to enhance time series classification in open world settings. In my spare time I like to stay current with deep learning libraries and other machine learning tools. For example, here is a jupyter notebook (hosted using google's Colaboratory) where I study a wifi network localization dataset and use a number of techniques to produce better results than the paper that studied this data.
I have done collaborative research in deep reinforcement learning with Ermo Wei which has been published in AAAI Spring Symposium (Multiagent Soft Q-Learning and Hierarchical Approaches to Reinforcement Learning in Parameterized Action Space).
Additionally, I have done research in cloud robotics where we applied a bounty hunting paradigm to geographically adjacent servers (youtube demo).
In 2014 and 2015 I was involved in GMU's Robocup team, the RoboPatriots. We competed in the kid sized humanoid team and in 2014 we made it to the second round with robots who we taught to play soccer the day before. We were in an article by Fast Company. Some of the highlights of RoboCup 2014 are in our qualification video for RoboCup 2015. In RoboCup 2015 the rules changed and the ball size increased and instead of the field being carpet like it was in 2015 it was astro-turf. We did not perform as well in 2015 due to these changes, but we developed some new techniques for walking and continued to be the only team that trained their robots how to play soccer using machine learning the day before the competition. You can see our robots in RoboCup 2015 on youtube. I've also enjoyed mentoring middle schoolers at both Kramer Middle School and Fred Lynn Middle School for the Botball autonomous robotics competition in 2014 and 2015.
While obtaining my Masters in Computer Science I led a number of group projects in machine learning. In one project we examined facebook friend data (social network analysis) in order to predict and recommend (using unsupervised learning) possible friends we called Discovering Friendships. In another project we used tweets to predict the location of the user based on the content of their tweets (also on GitHub). We focused on a number of unsupervised learning algorithms and Natural Language Processing techniques. Many more of my projects can be found on my GitHub.