Aaron Arvey

About Me

Research interests include inference in machine learning and bioinformatics. I am a PhD student at UCSD and working with Yoav Freund on a few very cool projects. I am currently focusing on developing a new non-convex boosting algorithm that will be available shortly in an open source package called JBoost. I am also starting work in probe creation for FISH experiments, functional proteomics, mass spectrometry analysis, and metagenomics.

My CV (PDF) describes some of my experiences.

Software

I've been part of a few large software projects. If you have any questions, don't hesitate to email me, or the appropriate mailing list. All projects are either posted on SourceForge or are in the process of being posted.

• JBoost is an implementation of several popular boosting algorithms using alternating decision tree. It is very effective at producing reliable, easy to use, and easy to visualize classifiers.
• RepeatMap is client/server pair of programs to determine very specific probes for in situ hybridizations and microarray experiments. You can use the client provided, set up a server of your own, or write your own client.

Refereed Journal Papers

Generalized Jensen-Shannon divergence measure for sequence analysis. Aaron Arvey, Alpan Raval, Rajeev Azad, and Jeffrey Lawrence. Submitted to The Journal of Theoretical Biology.

Refereed Conference Papers and Posters

Predicting Catalytic Residues Using Protein Sequence and Structure Features. Ron Alterovitz, Sriram Sankararaman, Aaron Arvey, Yoav Freund, and Kimmen Sjolander. Submitted to Intelligent Systems for Molecular Biology (ISMB) 2007.

Markovian Change Point Analysis to Detect Coding Regions of DNA. Aaron Arvey and Alpan Raval. Joint MAA and AMS Meeting Undergraduate Research Poster Session. San Antonio, Texas. January 2006. presentation (Outstanding Poster Award)

Online Markov Decision Processes for Learning Movement in Games. Aaron Arvey and Eric Aaron. Proceedings of the 6th International Conference on Computer Games: Artificial Intelligence and Mobile Systems. July 2005. Paper and videos

Empirical Bayesian Data Adjustment and Subpopulations. Aaron Arvey. Southern California Meeting of the American Statistics Association. November 2005. presentation (Best Undergraduate Poster)

Asymptotically Periodic Solutions of Linear ODE's. David Nichols and Aaron Arvey. Joint MAA and AMS Meeting Undergraduate Research Poster Session. Phoenix, Arizona. January 2004. Abstract, poster, and presentation (Outstanding Poster Award)

Technical Reports

These are fairly lengthy reports, some of which have proprietary information. Thus some are unfortunately not available on my site. However, feel free to contact me for more information and I will tell you what I can.

Hierarchical Dynamic Bayesian Networks and Level 3 Fusion. Aaron Arvey. Submitted to Toyon Research Corporation. September 2005.

DFR Lower Confidence Bounds and LE Subpopulations. Mike Reine, Aaron Arvey, and Janet Myhre. Submitted to Lockheed Martin and the Navy. November 2005.

Differential Code Coverage in the Context of the Wine Project. Aaron Arvey, Cal Pierog, Ed Kim, Evan Parry, Elizabeth Sweedyk, and Dan Kegel. Submitted to Google. May 2005.   Documents, information, patches, and scripts.

Presentations

Be sure to download the OpenOffice.org if you want to edit my presentations: .
OpenOffice exports to almost all formats (including PowerPoint) and is entirely open source.

• Introduction to Boosting and Alternating Decision Trees [PDF], [ODP],
• Maximizing the Normalized Margin [PDF], Handouts [PDF],
• RepeatMap: Highly Specific Hybridization Probes [PDF], [ODP],
• Intrusion Detection Systems
• Clustering Methodologies
• Information Theory
• Differential Equations
• Reliability Theory
• Machine Learning
• Markov Decision Processes

I'll add links to all of the actual presentations at some point in time...

Links

Here are some links to other parts of the site.

• Papers on Boosting
• Shell One-Liners

Here are some links which I use on a daily basis and don't want to have to deal with bookmarks.