| Office: Gates Hillman Center (GHC) 8010 |
| Phone: 412-268-2627 |
| jkbradle (yes, without the y) at cs dot cmu dot edu |
I just finished my fifth year in the Ph.D. Program in Machine Learning in the Machine Learning Deparment at Carnegie Mellon University. My advisor is Carlos Guestrin, and I am part of the Select Lab. My interests are in machine learning and probabilistic models.
I am researching tractable methods for learning CRFs with arbitrary structures. We aim to use decomposable learning methods which do not require inference during learning, but which also come with strong theoretical guarantees for finite sample sizes. This work is ongoing but should be published before long.
I am also researching learning tractable (low-treewidth) structures for CRFs. Up to now, little work has been done on CRF structure learning, but our techniques permit efficient learning of tree structures and do very well at recovering ground-truth models from synthetic data. (project page under construction)
Many CRF parameter and structure learning sub-problems can be solved using parallel computing. My main interest in parallel learning is using Graphics Processing Units (GPUs), which can give large speedups for relatively little cost.
My first year in grad school, I was advised by Eric Xing and worked on Population Genetics, modeling ancestral populations of single species using latent variable models.
Before coming to grad school, I was an undergraduate at Princeton University, where I received a B.S.E. in Computer Science. At Princeton, my main research was with Robert E. Schapire. We researched boosting in the filtering framework, where the learner does not use a fixed training set but rather has access to an example oracle which can produce an unlimited number of examples from the target distribution. This setting is useful for modeling learning with datasets too large to fit into a computer, learning in memory-limited situations, or learning from an online source of examples (e.g. from a web crawler).
Joseph K. Bradley and Carlos Guestrin.
Sample Complexity of Composite Likelihood.
In
the 15th International Conference on Artificial Intelligence and Statistics (AISTATS), 2012.
Paper (PDF)
Joseph K. Bradley, Aapo Kyrola, Danny Bickson, and Carlos Guestrin.
Parallel Coordinate Descent for L1-Regularized Loss Minimization.
In
the 28th International Conference on Machine Learning (ICML), 2011.
Paper (PDF)
Talk slides (PPT)
Project page (with code, data, and supplementary material)
Joseph K. Bradley and Carlos Guestrin.
Learning Tree Conditional Random Fields.
In
the 27th International Conference on Machine Learning (ICML), 2010.
Paper (PDF)
Talk slides (PPT)
Code available for download (Note: This code is part of a larger lab codebase which we are preparing to release. The new release features many improvements but will not be completely compatible with this previous release.)
Joseph K. Bradley and Robert E. Schapire.
FilterBoost: Regression and Classification on Large Datasets.
In
Advances in Neural Information Processing Systems 20 (NIPS), 2008.
Paper (PDF) with Appendix
Slides (PPT) from oral at NIPS
I do competitive Latin, Standard and Smooth ballroom dancing.
It's awesome. You should do it too.
(Check out CMU's Ballroom Dance Club!)
From our Christmas 2011 piece:
I like traveling.
Before going to college, I was born and grew up in Birmingham, Alabama, where I spent less time in front of a computer. I like returning to my former state.