Syllabus for Machine Learning with Large Datasets 10-605 in Spring 2015
From Cohen Courses
Revision as of 14:50, 14 October 2015 by Wcohen
This is the syllabus for Machine Learning with Large Datasets 10-605 in Spring 2015.
- The assignments posted are drafts based on the assignments from 2014, and will be modified over the course of the semester - some may be changed substantially.
- Lecture notes and/or slides will be (re)posted around the time of the lectures.
- Tues Jan 13. Overview of course, cost of various operations, asymptotic analysis.
- Thus Jan 15. Review of probabilities, joint distributions and naive Bayes
- HW1A: streaming Naive Bayes 1 (with feature counts in memory). PDF Handout
- Tues Jan 20. Streaming algorithms and Naive Bayes; The stream-and-sort design pattern; Naive Bayes for large feature sets.
- HW1B: streaming Naive Bayes 2 (with feature counts on disk) with stream-and-sort. PDF Handout
- For 10/11-805 students: a one-paragraph summary of a recent research result you'd like to present is due. If you're planning/hoping to transfer from 605, but haven't yet transferred, then also submit this assignment. Email to wcohen+805 AT gmail.com with the subject "Presentation" and include, in addition to your summary:
- Your name and andrew id
- A link to the paper
- Your best guess as to what lectures should precede the presentation
- Due by 11:59:59pm EST Tuesday.
- Thus Jan 22. Messages, records and workflows; Phrase finding.
- Tues Jan 27. Hadoop and Map-Reduce
- Thus Jan 29. PIG and Other Workflow Systems for Hadoop
- Tues Feb 3. Rocchio and TFIDF
- Thus Feb 5. Fast KNN and similarity joins
- Tues Feb 10. Parallel Perceptrons 1.
- Thus Feb 12. Parallel Perceptrons 2.
- HW2 due: phrase finding with stream-and-sort
- Tues Feb 17. Scalable SGD and Hash Kernels
- HW3: Naive Bayes with Hadoop MapReduce. PDF Handouts: HW3.
- For 10/11-805 students: initial draft of project proposal is due. I will give you feedback on this, so please be clear about your proposal. I'm expecting approximately one page. You should discuss what dataset you plan to use, what results you hope to obtain, what baseline technique you will build on and/or compare to. Also include a section saying if you have a partner; and if you are willing to work with/mentor one or more 605 students, and if so, how you anticipate them contributing to the project.
- Thus Feb 19. Randomized Algorithms 1
- Tues Feb 24. Randomized Algorithms 2
- Thus Feb 26. Matrix Factorization and SGD
- Sun Mar 1.
- HW3 due: Naive Bayes with Hadoop MapReduce
- HW4: PDF wrteup
- Tues Mar 3. student presentations
- Thus Mar 5. student presentations
- Quiz: 
- Matt Gardner (mg1 at cs): Large-scale extensions of the path ranking algorithm 
- Jesse Dodge (jessed at andrew): large-scale lasso regularization 
- Ishan Misra (imisra at andrew): LSH for object detection 
- HW5: memory-efficient SGD PDF handout
- For 10/11-805 students: project proposal is due. This must contain a complete description of the data you will use.
- Sat Mar 7 (extended from Friday):
- HW4 due: Phrase-finding with Hadoop
- Tues Mar 10. no class - spring break.
- Thus Mar 12. no class - spring break.
- Tues Mar 17. Scalable PageRank PDF handout
- Thus Mar 19. Subsampling a graph with RWR
- HW5 due: memory-efficient SGD
- HW6: Subsampling and visualizing a graph. PDF handout
- Tues Mar 24.
- Thus Mar 26. Guest lecture: D. Sculley, Google, TBA
- Tues Mar 31. Sparse sampling and parallelization for LDA
April and May
- Wed April 1
- Thus Apr 2. Speeding up LDA-like models: All-reduce and other tricks
- Tues Apr 7. Guest lecture - Alex Beutel, SGD for Tensors
- Thus Apr 9. Guest lecture - Alex Smola, Scalable parameter servers
- If you don't like the MediaTech one, a Youtube video on is also available for Alex's talk.
- Mon Apr 13. Informal update due for students working on project teams due.
- Each student working on a project should send to email@example.com an update, between 1/2 page and 1 page long, saying what concrete tasks you've accomplished to date, how these tasks are part of the overall project (if you're not the only member), and what you plan to do between 4/13 and the presentation on 4/23.
- Additionally, each project lead (i.e., each 805 student that has any 10-605 student working with them) should add a list of who's working on their project, and one line indicating if they're making good progress so far.
- Tues Apr 14. SSL on Graphs
- Thus Apr 16. no class : carnival
- HW7 due
- HW8: Matrix factorization on parameter server
- Tues Apr 21. Graph models for large-scale ML
- Thus Apr 23. Presentation for 10/11-805 projects
- Tues Apr 28. Exam review session.
- Thus Apr 30. In-class exam.
- Tues May 5.
- For 10/11-805 students: project reports are due