Rensselaer
Polytechnic Institute
ECSE 35-6610 Pattern
Recognition - Fall 2004
Teusday & Friday 2:00 -
3:20 Carnegie 102 (?)
Instructor..................................................................... Professor George Nagy
(nagy@ecse.rpi.edu)
Office hours........................... Tuesday
& Thursday 11:00 - 1:00, or by appointment, JEC 6020, 276-6078
Prerequisites.................................................................................. Probability;
programming skill
Grading..................................................................................... 5
programming assignments 50%
Term
paper (due October 15) 10%
Midterm
examination (October 19) 15%
Final
(as scheduled by the Registrar) 25%
Text........................................................... Duda,
Hart and Stork, Pattern Classification, Wiley 2001
Topics
Linear
and nonlinear decision procedures
Parametric
and nonparametric classifiers
Adaptation
and non-supervised learning
Feature
extraction and selection
Parameter
estimation and performance evaluation (bias and variance)
Clustering,
mixture distributions, Expectation Maximization
Linguistic
and style context, Hidden Markov Methods
Applications
to Optical Character Recognition
All of the programming assignments will be based on a file of optically
scanned typeset characters,
which we will use to compare the performance of various classification
algorithms.
Course objective
On completion
of the course, students should be sufficiently familiar with the formal
theoretical structure, notation, and vocabulary of pattern recognition to be
able to pursue matters of interest in the current technical literature. They
will also understand some of the engineering aspects of a prototypical
application of pattern recognition, Optical Character Recognition.
Examinations will be open book, but Cheating (copying on examinations
or programs) will result in a grade of F for the course Discussing programming
assignments between students is encouraged., but each team must write a separate program, and a
separate report Assignments up to one week late will be given half marks
Assignments more than one week late will not be accepted. Special
considerations will, however, be given to special situations, such as the DQE:
they must be negotiated in advance Grades may be appealed initially to
Professor Nagy: if the difficulty cannot be resolved, procedures in the
Rensselaer Handbook must be followed.
35-661 Assignments
and Due Dates
All assignment due at the beginning of the class!
1 Moment feature
extraction from printed characters....................................... 9-17
2 Bayes classifier for binary and Gaussian distributions................................... 10-1
3 Term paper.................................................................................... 10-15
MIDTERM 10-19
4 Nearest-neighbors classifier................................................................. 10-29
5 Decision tree classifier....................................................................... 11-12
6 Clustering and unsupervised learning....................................................... 12-3
FINAL scheduled by the Registrar