Big Picture
W4240 is a master’s / advanced undergraduate level course in data mining, pattern recognition, and statistical machine learning. W6240 is the same but for PhD’s. Extra work will be required of the PhD students.
High Level Syllabus
This course will be divided into roughly six parts, five structured, the sixth, a grab bag which may include presentations from other faculty and students who specialize in particular techniques. Each section except the last will be accompanied by a programming project. The subject of the programming project is listed below the topic for each section.
- Introduction / Review
- Iterated Conditional Modes
- Graphical Models
- Belief propagation on discrete Bayes nets
- Expectation Maximization (EM)
- EM for linear regression
- EM for Gaussian mixture models (GMM)
- Variational Inference
- Variational GMM
- Sampling
- Bayesian Logistic Regression
- Latent Dirichlet Allocation (LDA)
- Miscellaneous Topics
- Regression
- Bayesian linear regression, Gaussian process regression
- Sequential Data
- Kalman and particle filtering
- K-means
- Neural nets
- Dimensionality reduction
- Bayesian nonparametrics
- Matrix factorization
- Regression
Miscellaneous Course Information
- Text
- Pattern Recognition and Machine Learning Christopher M. Bishop. Springer, 2006.
- Prerequisites
- Calculus; probability and statistics at the level of W4150, or W4105 and W4107 taken concurrently.
- Linear algebra.
- Introductory level programming (java, matlab, c)
- Grading
- Grades will be assigned on a curve, using the following percentages: 75% Homework, 25% Final Project. Active participation can move you up a half a grade. No participation can move you down half a grade.
- Computing:
- You will be required to use Matlab to complete your homework assignments. Software of your choice may be used to complete the final project. See the ACIS page for information on software and computing labs.