Syllabus for Structured Prediction 10-710 in Fall 2011

This is the syllabus for Structured Prediction 10-710 in Fall 2011. The slides available may be out of date (from the last version of the class) but will be updated after each lecture.

August/September

• Tues 8/30. Introduction/Overview (Noah)
• Thus 9/1. HMMs as structured-output prediction models. (William)
• Tues 9/6. HMMs to MEMMs. (William)
• Thus 9/8. CRFs. (William).
• Mon 9/12. Add a short project proposal to the wiki (even one sentence might be enough).
• Tues 9/13. Pseudo-likelihood inference. (William).
• Thus 9/15. Understanding "conditional" models. (Noah).
• Mon 9/19. Join or recruit a 2-3 person project team and add that information to your project proposal.
• Tues 9/20. Perceptrons and margin-based learning. (William).
• Thus 9/22. Ranking perceptrons and margin-based learning. (William).
• Mon 9/26. Identify the baseline method(s) and dataset you will use in your project.
• Tues 9/27. Stacking, Searn, and related "meta-learning" techniques. (William).
• Thus 9/29. Generalizing SVM to structures. (Noah; note: this is Rosh Hashanah)
• Fri 9/30. First set of four wiki assignments due. Brendan has posted some detailed guidelines. Some more information on this was given in beginning of the lecture on Sept 1.

October

• Tues 10/4. Making structured predictions with dynamic programming (Noah).
• Thus 10/6. Inference with ILP and other methods (Noah).
• Thus 10/6. Complete your project proposal by explaining what your "big idea" will be.
• Tues 10/11. Natural language parsing with CFGs (Noah).
• Thus 10/13. Dependency parsing. (Noah).
• Tues 10/18. EM and Learning edit distances 1. (William).
• Thus 10/20. EM and Learning edit distances 2. (William).
• Tues 10/25. Mid-term project reports - short presentations of task, dataset, and baseline.
• Thus 10/27. Weighted FSTs. (Noah).

November

• Tues 11/1. Unsupervised grammar induction. (Noah).
• Wed 11/2. Deadline for October round of writeups.
• Thus 11/3. Alignment and translation (guest lecture by Kevin Gimpel).
• Tues 11/8. Kernels for structured inputs. (William).
• Thus 11/10. Kernels for structured outputs. (Noah).
• Tues 11/15. Discriminative latent-variable models. (Noah).
• Thus 11/17. Information extraction (guest lecture by Brendan O'Connor).
• Tues 11/22. Regularizers derived from unannotated data. (William).
• Thus 11/24 -- Thanksgiving, no class
• Tues 11/29. Classification on graphs (guest lecture by Frank Lin, Ni Lao)
• 11/30. Deadline for November round of writeups.

December

• Thus 12/1. Bayesian approaches to structured models. (Noah).
• Tues 12/6. Project presentations 1
• Thus 12/8. Project presentations 2.
• Mon 12/12. Final projects due.