COMP-644A: Pattern Recognition - Course Outline

Course: COMP-644A
Title: Pattern Recognition (4 credits, 3 hours (2 lectures) per week)
Time & Place: MW 11:35-12:55, McConnell Engineering 103
Instructor: Godfried Toussaint , Room 307, McConnell
Phone: External - 398-7077, Internal - 5911
Office hours: Tues & Thurs, 16:00-17:00.
Teaching Assistant: Erin McLeish, Office: McConnell Engineering 109, phone: 398-5485, email:, Office hours:  Monday, Wednesdays 3 - 4pm.
Course prerequisites: Performance assesment and workload: Course Material: Course Contents:
  • Introduction to pattern recognition via character recognition: grids, connectivity and contour tracing.
  • Smoothing: regularization, local averaging, median filtering and polygonal approximation.
  • Differentiation: image enhancement, the Laplacian operator and unsharp masking.
  • Moments of area and perimeter for shape measurement.
  • Medial axis transforms: skeletonization algorithms and medial axis algorithms.
  • Topological feature extraction: convex hulls and convex deficiencies.
  • Processing line drawings: Freeman chain coding and geometric probability.
  • Detecting structure in noisy pictures: Hough transforms, proximity graphs, relative neighborhood graphs, sphere-of-influence graphs, alpha-hulls, crusts and beta-skeletons.
  • Neural networks: non-parametric learning and error-correction rules.
  • Bayesian decision theory: discrete and continuous case, Gaussian density functions, Mahalanobis distance.
  • Feature selection: independence of measurements, redundancy and synergism, information theory and feature evaluation criteria and feature selection search strategies.
  • Measuring string similarity.
  • Estimation of parameters: maximum likelihood and Bayesian estimation.
  • Estimation of misclassification: generalization, substitution, leave-one-out and bootstrap.
  • Nearest neighbor decision rules: editing, condensing and efficient nearest neighbor search.
  • Cluster analysis and unsupervised learning: decision-directed learning, graph-theoretic methods, agglomerative and divisive methods.
  • Using context in pattern recognition: Markov methods and the Viterbi algorithm.
  • Support Vector Machines.
  • Music Information Retrieval.
  • Reminder:
  • McGill University values academic integrity.  Therefore all students must understand the meaning and consequences of cheating, plagiarism and other academic offences under the Code of Student Conduct and Disciplinary Procedures (see for more information).