Meets Wednesdays 5pm-6pm, Allen Center 203 (Fall 2006)
Organizers: Sumit Basu
(Microsoft Research) and
Tanzeem Choudhury
(Intel Research Seattle)
The Markovia seminar surveys a variety of topics in machine learning via reading, presenting, and discussing papers from top conferences and journals. Our focus and level of formality vary from quarter to quarter: this time around we will be discussing some recent hotspots in the community, including learning to rank, new methods for unsupervised learning, and more. We will have a fairly informal reading-group setting this fall, where we will discuss the papers as a group; in other (for-credit) quarters we have a more formal structure with members volunteering for papers and presenting them to the rest of the group. We welcome all students and faculty to Markovia and extend a special invitation for new graduate students. Participants should ideally have a good background in probability; familiarity with the basic notions of machine learning (supervised vs. unsupervised, generative vs. discriminative, classification vs. inference, etc.) will be helpful. NOTE: this quarter 590mv is not for offered for credit, but will be again in future quarters.
If you are interested in attending, feel free to drop in; if you'd like to receive mailings about the next week's paper, future course announcements, etc., please email us so we can put you on the mailing list.
| Date | Topic | Paper(s) |
| October 18, 2006 | Learning to Rank | Burges et al., "Learning to Rank using Gradient Descent." Proceedings of the Int'l. Conf. on Machine Learning 2005 (ICML '05). |
| October 25, 2006 | Matrix Factorization | N. Srebro, J. Rennie, and T. Jaakkola. "Maximum Margin Matrix Factorization." NIPS 2004. |
| November 1, 2006 | Log-Linear Classifiers | N. Smith and J. Eisner. "Contrastive Estimation: Training Log-Linear Models on Unlabeled Data." Proceedings of the Assoc. of Comp. Linguistics 2005 (ACL'05). |
| November 8, 2006 | Risk Bounds | P. Bartlett, M. Jordan, and J. McAuliffe. "Convexity, Classification, and Risk Bounds." University of California, Berkeley, ECE Department Tech Report #638. pp. 1-25. |
| November 15, 2006 | GP Regression | Y. Shen, A. Ng, and M. Seeger. "Fast Gaussian Process Regression using KD-Trees." NIPS 2005. |
| November 22, 2006 - NO CLASS | Thanksgiving Week | |
| November 29, 2006 | TBA | Cancelled due to snowstorm |
| December 6, 2006 - NO CLASS | [NIPS in Vancouver] | |
| Dec. 13, 2006 (Finals Week) | Nearest Neighbor Methods | J. Goldstein, J.C. Platt, and C.J.C. Burges. "Redundant Bit Vectors for Quickly Searching High-Dimensional Regions." In Deterministic and Statistical Methods in Machine Learning, Springer Lecture Notes in AI, pp. 137-158 (2005). |