CS6780 - Advanced Machine Learning
Prof. Thorsten Joachims
Cornell University, Department of Computer Science
Shortcuts: [Slides] [Piazza] [CMS]
Time and Place
First lecture: January 22, 2015
Last lecture: May 5, 2015
Tuesdays/Thursdays, 2:55pm - 4:10pm
- Gates 114 (Cornell Ithaca)
- Ursa room (Cornell NYCTech)
Exam: April 16
Project Report: May 11 and May 15
This course gives a graduate-level introduction to machine learning and statistical pattern recognition and in-depth coverage of new and advanced methods in machine learning, as well as their underlying theory. It emphasizes approaches with practical relevance and discusses a number of recent applications of machine learning, such as data mining, computer vision, robotics, text and web data processing. An open research project is a major part of the course.
In particular, the course will cover the following topics:
- Supervised Batch Learning: model, decision theoretic foundation, model selection, model assessment, empirical risk minimization
- Decision Trees : TDIDT, attribute selection, pruning and overfitting
- Statistical Learning Theory : generalization error bounds, VC dimension
- Large-Margin Methods: linear Rules, margin, Perceptron, SVMs
- Kernels: duality, non-linear rules, non-vectorial data
- Probabilistic Models: generative vs. discriminative, maximum likelihood, Bayesian inference
- Sequence Prediction : hidden Markov model, Viterbi
- Structured Output Prediction : undirected graphical models, structural SVMs, conditional random fields
- Latent Variable Models: k-means clustering, mixture of Gaussians, expectation-maximization algorithm, matrix factorization, embeddings
- Online Learning : experts, bandits, online convex optimization
- Other topics: neural nets, ensemble methods, sparsity, ...
The prerequisites for the class are: programming skills (at the level of CS 2110) and basic knowledge of linear algebra (at the level of MATH 2940) and probability theory (at the level of MATH 4710) and multivariable calculus (at the level of MATH 1920).
Enrollment is limited to PhD students. Students who have already taken CS 4780/CS 5780 should not take this class.
- 01/22: Introduction [slides] [slides 6up]
- What is learning?
- What is machine learning used for?
- Overview of course, course policies, and contact info.
- 01/27: Supervised Batch Learning [slides] [slides 6up]
- Supervised learning for binary classification, multi-class classification, regression, and stuctured output prediction
- Instance space, target function, training examples
- Hypothesis space, consistency, and version space
- Classifying with a decision tree
- Representational power of decision trees
- TDIDT decision tree learning algorithm
- 01/29: Empirical Risk Minimization [slides] [slides 6up]
- Independently identically distributed (iid) data
- Risk, empirical risk, and Empirical Risk Minimization (ERM)
- Bayes decision rule and Bayes risk
- Model selection and overfitting
- Model assessment and machine learning experiments
- 02/10: Statistical Learning Theory: Generalization Error Bounds [slides] [slides 6up]
- Generalization error bound for finite H and zero error
- Sample complexity
- Probably Approximately Correct (PAC) learning
- Generalization error bound for finite H and non-zero error
- Theoretical characterization of overfitting
- 02/12: Linear Classifiers and Perceptrons [slides] [slides 6up]
- Linear classification rules
- Batch Perceptron learning algorithm
- Online Perceptron learning algorithm
- Margin of linear classifier
- Perceptron mistake bound
- 02/19: Statistical Learning Theory: Error Bounds and VC-Dimension [slides] [slides 6up]
- Vapnik-Chervonenkis Dimension
- VC dimension of hyperplanes
- Symmetrization and error bound
- 02/24: Support Vector Machines: Optimal Hyperplanes and Soft Margin [slides] [slides 6up]
- Optimal hyperplanes and margins
- Hard-margin Support Vector Machine
- Soft-margin Support Vector Machine
- Primal optimization problems
- 02/26: Support Vector Machines: Duality and Leave-one-out [slides] [slides 6up]
- Dual Perceptron
- Dual SVM optimization problem
- Connection between primal and dual
- Bounds on the leave-one-out error of an SVM
- 03/03: Support Vector Machines: Kernels [slides] [slides 6up]
- Input space and feature space
- Kernels as a method for non-linear learning
- Kernels for learning with non-vectorial data
- 03/10: Regularized Linear Models (slides] [slides 6up]
- Conditional maximum likelihood
- Maximum a posteriori estimates
- Logistic regression
- Ridge regression
- 03/12: Generative Models for Classification [slides] [slides 6up]
- Multivariate naive Bayes classifier
- Linear discriminant analysis
- Multinomial naive Bayes classifier
- MLE vs Bayesian
- 03/17: Structured Output Prediction: Generative Models [slides] [slides 6up]
- Hidden Markov model (HMM)
- Estimation of HMM parameters
- Viterbi algorithm
- 03/19: Structured Output Prediction: Discriminative Learning [slides] [slides 6up]
- Structural SVMs
- Conditional Random Fields
- 03/24: Online Learning: Expert Setting [slides] [slides 6up]
- Expert Online Learning Model
- Halving Algorithm and Optimal Mistake Bound
- Weighted Majority Algorithm
- Exponentiated Gradient Algorithm
- 03/24: Online Learning: Bandit Setting [slides] [slides 6up]
- Adversarial Bandit Model
- EXP3 Algorithm
- Stochastic Bandit Model
- UCB1 Algorithm
- 04/07: ERM Optimization 1 (Yoram Singer) [slides]
- ERM Review
- Convexity, Smoothness, Lipschitzness
- Gradient Descent Algorithm
- 04/09: ERM Optimization 2 (Yoram Singer)
- Convergence rate of GD
- Faster convergence for special cases
- Stochastic Gradient Algorithm
- 04/14: Unsupervised Learning: k-Means and Mixtures of Gaussians [slides] [slides 6up]
- Flat clustering
- k-Means algorithms
- Mixture of gaussians model
- EM-algorithm for mixture of gaussians
- 05/05: Wrap-up [slides] [slides 6up]
- Batch learning: ERM vs. Conditional vs. Generative
- Online learning: Experts and Bandits
- Unsupervised learning: Latent variable models
- What we did not cover and other courses
Staff and Office Hours
Please use the CS6780 Piazza Forum for questions and discussions. The forum is monitored by all the TA's and the prof -- you will get the best response time. And all the TA's will know the question you asked and the answers you received.
- Prof. Thorsten Joachims (homepage) [Office hours: Friday, 2:30-3:30pm (Gates 418)]
- Daniel Sedra, TA (homepage) [Office hours: Monday 9:30-10:30am (Gates G19), Friday 9:30-10:30am (Gates G13)]
- Adith Swaminathan, TA (homepage) [Office hours: Monday 5-6pm (Gates G11), Thursday 5:30-6:30pm (Gates G17)]
Assignments and Exams
Homework assignments can be downloaded from CMS.
All assignments are due at the beginning of class on the due date. Assignments turned in late will be charged a 1 percentage point reduction of the cumulated final homework grade for each period of 24 hours for which the assignment is late. However, every student has a budget of 5 late days (i.e. 24 hour periods after the time the assignment was due) throughout the semester for which there is no late penalty. So, if you have perfect scores of 100 on all 4 homeworks and a total of 8 late days, your final homework score will be 97. No assignment will be accepted after the solution was made public, which is typically 3-5 days after the time it was due. You can submit late assignments in class or following the policy written on the homework handout.
Graded homework assignments and prelims can be picked up in Gates 216 (opening hours Monday - Friday 12:00pm - 4:00pm).
Regrade requests can be submitted within 7 days after the grades have been made available on CMS. Regrade requests have to be submitted in writing and in hardcopy using this form (or similar). They can be submitted in class or to Gates 216.
This is a 4-credit course. Grades will be determined based on a written exam, a research project, homework assignments, and class participation.
- 35%: Exam
- 35%: Research Project
- 20%: Homework (~4 assignments)
- 10%: Class Participation (e.g., lecture, piazza)
To eliminate outlier grades for homeworks, the lowest grade is replaced by the second lowest grade when grades are cumulated at the end of the semester.
All assignment, exam, and final grades (including + and - of that grade) are roughly on the following scale: A=92-100; B=82-88; C=72-78; D=60-68; F= below 60.
For the research project, we will use peer review analogous to how scientific papers are reviewed for conferences and journals. This means that you will read and comment on other students work, which provides input for the TA's and the professor to determine the project grades. The quality of your reviewing also becomes a component of your own grade.
Students taking the class S/U do all work, except for the project, and need to receive at least a grade equivalent to a D to pass the course.
The main textbooks for the class is:
- Kevin Murphy, "Machine Learning - a Probabilistic Perspective", MIT Press, 2012. (online via Cornell Library)
We will also read original research papers and other sources from the following list:
- Tom Mitchell, "Machine Learning", McGraw Hill, 1997.
- Cristianini, Shawe-Taylor, "Introduction to Support Vector Machines", Cambridge University Press, 2000. (online via Cornell Library)
- Schoelkopf, Smola, "Learning with Kernels", MIT Press, 2001. (online)
- Bishop, "Pattern Recognition and Machine Learning", Springer, 2006.
- Ethem Alpaydin, "Introduction to Machine Learning", MIT Press, 2004.
- Devroye, Gyoerfi, Lugosi, "A Probabilistic Theory of Pattern Recognition", Springer, 1997.
- Duda, Hart, Stork, "Pattern Classification", Wiley, 2000.
- Hastie, Tibshirani, Friedman, "The Elements of Statistical Learning", Springer, 2001.
- Joachims, "Learning to Classify Text using Support Vector Machines", Kluwer, 2002.
- Leeds Tutorial on HMMs (online)
- Manning, Schuetze, "Foundations of Statistical Natural Language Processing", MIT Press, 1999. (online via Cornell Library)
- Manning, Raghavan, Schuetze, "Introduction to Information Retrieval", Cambridge, 2008. (online)
- Vapnik, "Statistical Learning Theory", Wiley, 1998.
This course follows the Cornell University Code of Academic Integrity. Each student in this course is expected to abide by the Cornell University Code of Academic Integrity. Any work submitted by a student in this course for academic credit will be the student's own work. Collaborations are allowed only if explicitly permitted. Violations of the rules (e.g. cheating, copying, non-approved collaborations) will not be tolerated.