Computer Science 246/446

Machine Learning

Spring 2019

Instructor: Dan Gildea office hours Tu/Th 1-2pm Wegmans Hall 3019

TAs:

Location: M/W 9-10:15am, Wegmans 1400

The course is full. If spaces become available, I will add people who hand in the first homework. If you would like to try to add, write "NEED TO ADD" on the top of your assignment, along with your name, program (eg, BS in CS) and year.

Prereqs: Probability, Linear Algebra, Vector Calculus.

Homeworks

Lecture notes

Required text: Christopher M. Bishop, Pattern Recognition and Machine Learning.

The following are useful references in addition to the reading material assigned for each class:

  • Stuart Russell and Peter Norvig, Artificial Intelligence, A Modern Approach.
  • Trevor Hastie, Robert Tibshirani, Jerome Friedman, The Elements of Statistical Learning: Data Mining, Inference, and Prediction.
  • Larry Wasserman, All of Statistics, 2004.

Syllabus

Onwe will coverwhich means that after class you will understandif before class you have read
1/16 Regression and Classification constrained optimization, perceptron bishop 1.2, 1.4, app E
1/23 Logistic Regression stochastic gradient descent bishop 4.3
1/28 Backpropagation DP for gradient descent bishop 5.1, 5.2, 5.3
1/30 Deep Learning drop-out Krizhevsky 2012
2/4 Support vectors max-margin bishop 3.1, 4.1
2/6 Support Vectors strong duality bishop 7.1
2/11 Support Vectors the kernel trick bishop 6.1, 6.2
2/13 Hidden Markov Models forward-backward bishop 13.2
2/18 Graphical Models bayes ball
2/20 Probabilistic Inference message passing bishop 8.4
2/25 Tree decomposition cyclic graphs koller and friedman
2/27 Tree decomposition cont'd vertex elimination
3/4 Review
3/6 Midterm
3/18 Midterm Solutions
3/20 Expectation Maximization L = Q + H + D bishop 9
3/25 Expectation Maximization mixture of gaussians bishop 9
3/27 EM for HMM minimum bayes risk
4/1 Sampling Markov Chain Monte Carlo bishop 11.2
4/3 Metropolis Hastings detailed balance bishop 11.2
4/8 Gibbs sampline annealing bishop 11.3
4/10 Learning Theory PAC Kearns and Vazirani
4/15 Learning Theory VC dimension
4/17 Optimization SGD convergence Ruder 2016
4/22 PCA Eigenvectors
4/25 Reinforcement Learning q-learning sutton ch 3, 4.3, 4.4, 6.1, 6.5, 7.2, 11.1
4/29 Something fun
5/1 Review come to class with questions!
Final Exam: Wed May 8, 8:30-11:30am, Douglass Ballroom.

Grading

  • Homeworks: 50%
  • Final exam: 30%
  • Midterm: 20%

gildea @ cs rochester edu
January 23, 2019