About Me
I am a third year graduate student working under Padhraic Smyth at the University of California, Irvine. My group, the Datalab, focuses on probabilistic modeling, specifically using graphical models. Generally, I am always interested in learning about novel methods for efficiently and/or elegantly modeling noisy data.
Current Work
I am currently working with Andrew Robertson on modeling rainfall through the use of nonhomogeneous Hidden Markov Models. Work done so far includes validating the predictive power of basic models and updating Sergey Kirshner's MVNHMM package to allow HMM inference to be done in parallel.
Publications
"Downscaling projections of Indian monsoon rainfall using
a nonhomogenous hidden Markov model."
A. Greene, A. Robertson, S. Triglia, P. Smyth.
QJRMS Spring 2011
"Distributed Gibbs Sampling for Latent Variable Models"
A. Asuncion, D. Newman, I. Porteous, S. Triglia, P. Smyth, M. Welling.
Scaling Up Machine Learning, Cambridge University Press, 2010.
Previous Work
Data Analytics
In the summer of 2008, I worked on several projects as an intern in the Data Analytics group at LinkedIn. One of my tasks involved helping create an algorithm for selecting news articles of interest to a particular company. I also collected semantic data on companies from dbpedia.org and used it to verify and improve existing LinkedIn records. Finally, I created an interactive visualization of individual users' social networks using Prefuse.
Online PDF Management System
Worked on a four-man team on a project for Bluebeam Software to design, implement and test an online PDF management system. Our project emphasized user-interface design, database interfacing and filesystem crawling. The final result was a remotely accesible, searchable database of PDFs which could be tagged, rated, and finally downloaded onto your local machine.
Automatic Circuit Recognition
Worked with Christine Alvarado to improve circuit diagram recognizing on tablet PCs. I developed a hill climbing algorithm which leveraged knowledge of past drawings of logic gates to improve recognition of future gates.
Autonomous Robotic Navigation
In Summer 2007, I took part in an REU at UCSC where I created software improvements for the autonomous navigation of an ex-DARPA Grand Challenge vehicle. The Overbot navigated using LIDAR and radar sensors along a predetermined set of waypoints, navigatings obstacles along the way. The project included work with splines, basic control theory, and a lot of test driving.