Difference between revisions of "Syllabus for Machine Learning with Large Datasets 10-605 in Fall 2016"

From Cohen Courses
Jump to navigationJump to search
 
(73 intermediate revisions by 5 users not shown)
Line 1: Line 1:
 
This is the syllabus for [[Machine Learning with Large Datasets 10-605 in Fall 2016]].   
 
This is the syllabus for [[Machine Learning with Large Datasets 10-605 in Fall 2016]].   
 +
 +
----
  
 
Notes:  
 
Notes:  
 
* Homeworks, unless otherwise posted, will be due when the next HW comes out.
 
* Homeworks, unless otherwise posted, will be due when the next HW comes out.
 
* Lecture notes and/or slides will be (re)posted around the time of the lectures.
 
* Lecture notes and/or slides will be (re)posted around the time of the lectures.
 +
* Classes are cancelled for Oct 27
 +
* '''No classes will be held on Nov 24 (Thanksgiving)'''
 +
 +
----
  
−
''note: this is under construction''
+
Schedule for 805 projects:
 +
* 11:59pm Sun 10/2: [[Initial 805 project proposal]] due.
 +
* 11:59pm Sun 10/16: Final 805 project proposal due. 
 +
** This is a revised writeup that will address any comments William raises from the initial proposal.
 +
* 11:59pm Sun 11/13: [[Midterm 805 project report]] due.
 +
* '''1:30-2:50pm Tues 12/6: Project presentations''' (in class).  One presentation per group, 12minutes per presentation.  Please send your slide deck to William by '''10am 12/6''' (PDF is best).
 +
* 11:59pm Sun 12/11: [[Machine_Learning_with_Large_Datasets_10-605_in_Fall_2016#Project_Info|Final 805 project writeup]] due.
  
−
Schedule:
 
−
* Thus Sep 1. [[Class meeting for 10-605 Overview|Overview of course, cost of various operations, asymptotic analysis.]]
 
−
* Tues Sep 6. [[Class meeting for 10-605 Probability Review|Review of probabilities, joint distributions and naive Bayes]]
 
−
* Thus Sep 8.  [[Class meeting for 10-605 Streaming Naive Bayes|Streaming algorithms and Naive Bayes; The stream-and-sort design pattern; Naive Bayes for large feature sets.]]
 
−
** HW1 out: streaming naive Bayes in Java. [https://s3.amazonaws.com/vincy/10605-15Fall/HW1_StreamingNB.pdf PDF Handout]
 
−
* Tues Sep 13. [[Class meeting for 10-605 Phrase Finding|Phrase Finding]]
 
−
* Thus Sep 15. [[Class meeting for 10-605 Phrases_with_Stream_and_Sort|Implementing Phrase Finding and Large-Data Testing for Naive Bayes with Stream-and-Sort]].
 
−
** Lecture also discusses: map-reduce abstractions/dataflow
 
−
* Tues Sep 20. [[Class_meeting_for_10-605_Hadoop_Overview|Hadoop Overview]]
 
−
** HW2 out: naive Bayes training on Hadoop in Java. [https://drive.google.com/file/d/0BzQQ-spWKjhUd0NXSTB6TW82LWM/view PDF Handout]
 
−
* Thus Sep 22 - Thus Sep 24. [[Class_meeting_for_10-605_Rocchio_and_Hadoop_Workflows|Hadoop Workflow Languages and Rocchio and TFIDF]]
 
−
** Lecture also discusses: hadoop streaming, mrjob, cascading, pipes, scaling, hive, pig, spark, flink
 
  
 
----
 
----
  
−
* Tues Sep 27.  [[Class meeting for 10-605 Similarity Joins|Fast KNN and similarity joins]]
+
Schedule for lectures and 605 assignments:
−
** HW3 out: Naive Bays in GuineaPig. [https://drive.google.com/file/d/0B-p8_eIVeEHFM1JOSGFWNFFJcU0/view PDF Handout]
 
−
* Thus Sep 29. [[Class meeting for 10-605 SGD and Hash Kernels|Scalable SGD and Hash Kernels]]
 
−
** For 805 students: an initial project proposal is due '''via email to wcohen+805@gmail.com'''. You will get feedback on it from the instructors, and it will also be posted to the class - mainly for 605 students that are interested in collaborating, but also for general interest.  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.
 
−
* Tues Oct 4. '''No class - Rosh Hashana.'''
 
−
* Thus Oct 6. [[Class meeting for 10-605 Parallel Perceptrons 1|Parallel Perceptrons 1]].
 
−
* Tues Oct 11. [[Class meeting for 10-605 Parallel Perceptrons 2|Parallel Perceptrons 2]].
 
−
* Thus Oct 13. [[Class meeting for 10-605 Advanced topics for SGD|More on parallel and streaming ML]]: Adaptive gradients, AllReduce, and Parameter Servers
 
−
** HW4 out: streaming logistic regression classifier [http://curtis.ml.cmu.edu/w/courses/images/8/86/Sgd_fall15.pdf PDF Handout]
 
  
−
stopped here
+
* Tues Aug 30, 2016 [[Class meeting for 10-605 in Fall 2016 Overview|Overview]].  Grading policies and etc, History of Big Data, Complexity theory and cost of important operations
−
 
+
* Thurs Sep 1, 2016 [[Class meeting for 10-605 in Fall 2016 Probability Review|Probability Review]].  Counting for big data and density estimation, streaming Naive Bayes, Rocchio and TFIDF
−
* Thus Oct 15. [[Class meeting for 10-605 SGD for MF|Matrix Factorization and SGD]]
+
** '''Start work on''' Assignment 1a: Streaming NB. [http://www.cs.cmu.edu/~wcohen/10-605/assignments/2016-fall/hashtable-nb.pdf Writeup].
−
** For 805 students: the final project proposal is due.
+
* Tues Sep 6, 2016 [[Class meeting for 10-605  in Fall 2016 Streaming Naive Bayes|Streaming Naive Bayes]].  Notes on scalable naive bayes, Local counting in stream and sort
−
* Tues Oct 20. Exam review tips ([http://www.cs.cmu.edu/~wcohen/10-605/midterm-review.pptx ppt], [http://www.cs.cmu.edu/~wcohen/10-605/midterm-review.pdf pdf]) and guest lecture from '''Mark Torrance of RocketFuel'''
+
* Thurs Sep 8, 2016 [[Class meeting for 10-605  in Fall 2016 Hadoop Overview|Hadoop Overview]].  Intro to Hadoop, Hadoop Streaming
−
* Thus Oct 22. ''midterm exam''
+
** '''Start work on'''  Assignment 1b: Streaming NB on Hadoop. Draft at https://autolab.andrew.cmu.edu/courses/10605-f16/assessments/hw1bhadoopnaivebayes/writeup
−
** [http://www.cs.cmu.edu/~wcohen/10-605/practice-questions/f2015-midterm.pdf practice questions for midterm - v1].  This document also references the relevant questions from two previous review sheets:
+
* Tues Sep 13, 2016 [[Class meeting for 10-605 Workflows For Hadoop|Workflows For Hadoop 1]].  Scalable classification, Scalable Rocchio and TFIDF, Abstracts for map-reduce algorithms, Joins in Hadoop, TFIDF in Pig, Guinea Pig intro, TFIDF in Guinea Pig
 +
* Thurs Sep 15, 2016 [[Class meeting for 10-605 Workflows For Hadoop|Workflows For Hadoop 2]]. Similarity joins, Similarity joins with TFIDF, Parallel simjoins
 +
** '''Start work on''' Assignment 2: Naive bayes testing in Guinea Pig, draft at https://autolab.andrew.cmu.edu/courses/10605-f16/assessments/hw2nbwithguineapig/writeup
 +
* Tues Sep 20, 2016 [[Class meeting for 10-605 Workflows For Hadoop|Workflows For Hadoop 3]].  PageRank in Pig, K-means in Pig, Spark, Systems built on top of Hadoop
 +
* Thurs Sep 22, 2016 [[Class meeting for 10-605 Phrase Finding|Phrase Finding]].  Phrase-finding in Pig, Other work with phrases
 +
* Tues Sep 27, 2016 [[Class meeting for 10-605 SGD and Hash Kernels|SGD and Hash Kernels]].  Learning as optimization, Logistic regression with SGD, Regularized SGD, Hash kernels for logistic regression
 +
* Thurs Sep 29, 2016 [[Class meeting for 10-605 Parallel Perceptrons|Parallel Perceptrons 1]]. Also wrapup for SGD, debugging ML algorithms
 +
** '''Start work on''' Assignment 3: scalable SGD at https://autolab.andrew.cmu.edu/courses/10605-f16/assessments/hw3sgd/writeup
 +
* Tues Oct 4, 2016 [[Class meeting for 10-605 Parallel Perceptrons|Parallel Perceptrons 2]]. 
 +
* Thurs Oct 6, 2016 [[Class meeting for 10-605 Parallel Perceptrons|Parallel Perceptrons 3]].  Structured perceptrons, Interative parameter mixing paper
 +
* Tues Oct 11, 2016 [[Class meeting for 10-605 SGD for MF|SGD for MF]].  Matrix factorization, Matrix factorization with SGD, distributed matrix factorization with SGD
 +
* Thurs Oct 13, 2016 [[Class meeting for 10-605 Midterm review|Midterm review]].
 +
** [http://www.cs.cmu.edu/~wcohen/10-605/practice-questions/f2015-midterm.pdf practice questions for midterm from 2015].  This document also references the relevant questions from two previous review sheets:
 
*** [http://www.cs.cmu.edu/~wcohen/10-605/practice-questions/s2014-final.pdf practice questions from final, 2014]
 
*** [http://www.cs.cmu.edu/~wcohen/10-605/practice-questions/s2014-final.pdf practice questions from final, 2014]
 
*** [http://www.cs.cmu.edu/~wcohen/10-605/practice-questions/s2015-final.pdf practice questions for final, 2015]
 
*** [http://www.cs.cmu.edu/~wcohen/10-605/practice-questions/s2015-final.pdf practice questions for final, 2015]
−
*** [http://www.cs.cmu.edu/~wcohen/10-605/midterm-review.pdf Some review tips - modified from last year's exam review session]
+
*** [http://www.cs.cmu.edu/~wcohen/10-605/midterm-review.pdf Some review tips - modified from last year's exam review session]  
−
 
+
** '''Last assignment due'''
−
----
+
* Tues Oct 18, 2016 [[Class meeting for 10-605 Midterm|Midterm]]. 
−
 
+
* Thurs Oct 20, 2016 [[Class meeting for 10-605 Subsampling a Graph|Subsampling a Graph]].  Sampling a graph, Local partitioning
−
* Tues Oct 27. [[Class meeting for 10-605 Randomized|Randomized Algorithms 1]]
+
** '''Start work on''' Assignment 4: Subsampling a Graph with Approximate PageRank, draft at https://autolab.andrew.cmu.edu/courses/10605-f16/assessments/hw4approximatepagerank/writeup
−
* Thus Oct 29. [[Class meeting for 10-605 Randomized|Randomized Algorithms 2]]
+
* Tues Oct 25, 2016 [[Class meeting for 10-605 Deep Learning|Deep Learning 1]].  Deep learning intro, Deep learning and GPUs, Expressiveness of MLPs, Exploding and vanishing gradients, Modern deep learning models
−
** HW5 out: dSGD for modeling text ([https://drive.google.com/file/d/0BzQQ-spWKjhUYUM1LUVZakx0ZlE/view])
+
* Thurs Oct 27, 2016. '''No class.'''
−
* Tues Nov 3. Finish up with randomized algorithms.
+
* Tues Nov 1, 2016 [[Class meeting for 10-605 Deep Learning|Deep Learning 2]].  Reverse-mode differentiation, Recursive ANNs, Word2vec
−
* Thus Nov 5. [[Class meeting for 10-605 Subsample A Graph|Scalable PageRank]]  
+
* Thurs Nov 3, 2016 [[Class meeting for 10-605 Randomized Algorithms|Randomized Algorithms 1]].  Bloom filters, The countmin sketch
−
* Tues Nov 10. [[Class_meeting_for_10-605_SSL_on_Graphs|SSL on Graphs]]
+
** '''Start work on''' Assignment 5: Autodiff with IPM.  This is a new assignment for Fall 2016. View writeup at https://github.com/KarandeepJohar/10605-f16-hw5/blob/master/automatic-reverse-mode.pdf
−
* Thus Nov 12. [[Class meeting for 10-605 LDA 1|Sparse sampling and parallelization for LDA]]
+
* Tues Nov 8, 2016 [[Class meeting for 10-605 Randomized Algorithms|Randomized Algorithms 2]]. Locality sensitive hashing
−
** HW6 out: approximate pagerank for sampling a graph ([https://goo.gl/ThtRc6])
+
* Thurs Nov 10, 2016 [[Class meeting for 10-605 Graph Architectures for ML|Graph Architectures for ML]]. Graph-based ML architectures, Pregel, Signal-collect, GraphLab, PowerGraph, GraphChi, GraphX
−
* Tues Nov 17.  ''Guest lecture, Chris Dyer.'' [http://demo.clab.cs.cmu.edu/cdyer/bigdata-cuda.pdf Learning with GPUs].
+
* Tues Nov 15, 2016 [[Class meeting for 10-605 SSL on Graphs|SSL on Graphs]]. Semi-supervised learning intro, Multirank-walk SSL method, Harmonic fields, Modified Adsorption SSL method, MAD with countmin sketches
−
* Thus Nov 19. ''Guest lecture: Aurick Qiao'', parameter servers [http://curtis.ml.cmu.edu/w/courses/images/8/85/Aurick_release.pptx ppt slides].
+
* Thurs Nov 17, 2016 [[Class meeting for 10-605 Unsupervised Learning On Graphs|Unsupervised Learning On Graphs]].  Spectral clustering, Power iteration clustering, Label propagation for clustering non-graph data, Label propagation for SSL on non-graph data
−
* Tues Nov 24.  [[Class meeting for 10-605 2013 LDA 2|Speeding up LDA-like models: All-reduce and other tricks]]
+
** '''Start work on''' Assignment 6:  Phrase-finding with Spark. Writeup at https://autolab.andrew.cmu.edu/courses/10605-f16/assessments/hw6phrasefindingwithspark/writeup
−
** HW7 out: LDA with a param server ([http://curtis.ml.cmu.edu/w/courses/images/1/16/Hw7-lda-ps.pdf PDF handout])
+
* Tues Nov 22, 2016 [[Class meeting for 10-605 LDA|LDA 1]].  DGMs for naive Bayes, Gibbs sampling for LDA
−
* Thus Nov 26. ''Happy Thanksgiving!''
+
* Tues Nov 29, 2016 [[Class meeting for 10-605 Parameter Servers|Parameter Servers]].
−
 
+
** '''Start work on''' Assignment 7: LDA with a Parameter Server, Writeup at https://autolab.andrew.cmu.edu/courses/10605-f16/assessments/hw7lda/attachments/677
−
----
+
* Thurs Dec 1, 2016 [[Class meeting for 10-605 LDA|LDA 2]].  Parallelizing LDA, Fast sampling for LDA, DGMs for graphs
−
 
+
* Tues Dec 6, 2016 [[Class meeting for 10-605 Project Reports|Project Reports]].
−
* Tues Dec 1, Thus Dec 3.  [[Class meeting for 10-605 GraphLab|Graph models for large-scale ML]]
+
** '''Last assignment due'''
−
* Tues Dec 8.  Review and project presentations (15 min each):
+
* Thurs Dec 8, 2016 [[Class meeting for 10-605 Final Exam|Final Exam]].  Note that we've posted:
−
** Schedule:
+
** [http://www.cs.cmu.edu/~wcohen/10-605/practice-questions/s2014-final.pdf practice questions from final, 2014]
−
*** Bhuwan Dingra/Yun Fu
+
** [http://www.cs.cmu.edu/~wcohen/10-605/practice-questions/s2015-final.pdf practice questions for final, 2015]
−
*** Rohit Girdhar
+
** Comments:
−
*** Siddha Ganju/Sravya Popuri/Srikant Avasarala
+
*** Most of the exam (approximately 80%) covers material from after the midterm.
−
*** Jingkun Gao/Yiming Gu
+
*** You may bring in '''two''' 8 1/2 by 11 sheets of paper with notes.
−
** HW7 due
 
−
* Thus Dec 10. In-class final exam.
 
−
* Tues Dec 15. Writeup for 10-805 projects are due (at 11:59pm).
 
−
 
 
−
== Topics covered in previous years but not in 2015 ==
 
−
 
 
−
* [[Class meeting for 10-605 Scalable FOL|Scalable First-order logics]]
 
−
* [[Class meeting for 10-605 PIG|Workflows in PIG]]
 
−
* [[Class meeting for 10-605 Phase Finding|Phrase Finding]]
 
−
* [[Class meeting for 10-605 Parallel Similarity Joins|Scalable Similarity Joins]]
 
−
* [[Class meeting for 10-605 Similarity Joins|Fast KNN and similarity joins]]
 
−
* [[Class meeting for 10-605 Rocchio and On-line Learning|Messages, records and workflows; Rocchio]]
 
−
* [http://www.cs.cmu.edu/~wcohen/10-605/schimmy.pptx Scalable pagerank - The Schimmy Pattern]
 
−
* [[Class meeting for 10-605 Spectral Clustering|Scalable spectral clustering techniques.]]
 

Latest revision as of 11:54, 11 August 2017

This is the syllabus for Machine Learning with Large Datasets 10-605 in Fall 2016.

Notes:

  • Homeworks, unless otherwise posted, will be due when the next HW comes out.
  • Lecture notes and/or slides will be (re)posted around the time of the lectures.
  • Classes are cancelled for Oct 27
  • No classes will be held on Nov 24 (Thanksgiving)

Schedule for 805 projects:

  • 11:59pm Sun 10/2: Initial 805 project proposal due.
  • 11:59pm Sun 10/16: Final 805 project proposal due.
    • This is a revised writeup that will address any comments William raises from the initial proposal.
  • 11:59pm Sun 11/13: Midterm 805 project report due.
  • 1:30-2:50pm Tues 12/6: Project presentations (in class). One presentation per group, 12minutes per presentation. Please send your slide deck to William by 10am 12/6 (PDF is best).
  • 11:59pm Sun 12/11: Final 805 project writeup due.


Schedule for lectures and 605 assignments:

Retrieved from "http://curtis.ml.cmu.edu/w/courses/index.php?title=Syllabus_for_Machine_Learning_with_Large_Datasets_10-605_in_Fall_2016&oldid=18893"