Pattern Recognition
Weekly outline
-
Textbook & References
- Pattern Recognition, by S. Theodoridis and K. Koutroumbas , second edition available on moodle,
- Book Website: http://cgi.di.uoa.gr/~stpatrec/welcome3d.html
- MATLAB-based examples accompanying the book: An Introduction to Pattern Recognition, by S. Theodoridis and K. Koutroumbas
- Richard O. Duda, Peter E. Hart, David G. Stock: Pattern Classification, 2nd edition, John Wiley & Sons, New York, 2001, EUR 114.45
- Trevor Hastie, Robert Tobshirani, Jerome Friedman: The Elements of Statistical Learning – Data Mining, Inference, and Prediction, 2nd edition, Springer, New York, 2009, EUR 71.05 http://www-stat.stanford.edu/~tibs/ElemStatLearn/
- Christopher M. Bishop: Pattern Recognition and Machine Learning, Springer, New York, 2006, EUR 75.05
- Used Material from Pattern Recognition Course offered by riedrich-Alexander University of Erlangen-Nuremberg. in 2012/2013 by Joachim Hornegger and Hahn Steidl
- More will be added as we need them.
Grading Criteria
- Week 7:
- Assignments (Part of the Project) 10%
- Midterm 20%
- Week 12:
- Assignments (Part of the Project) 10%
- Project presented in Week 12: 20%
- Final 40%
- Seminar presented in Week 15: 20%
- Final Exam in Week 16: 20%
- Pattern Recognition, by S. Theodoridis and K. Koutroumbas , second edition available on moodle,
-
6th of October Holiday - no class
-
Week 1: Introduction
ILOs:
- Introduction to pattern recognition applications, approaches
- Appreciate data visualisation techniques as methods for data mining and decision boundary shape estimation
- Classifiers' Performance Measures
-
Week 2: Probability & Bayes Decision Rule
ILOs:
- Revision to Probability Preliminaries
- Introduction to Bayes Classification
- Matlab functions to generate random Gaussian distributions, test the bayes decision rule
-
Week 3: Bayes Classification
ILOs:
- Practice Bayes Decision Rule
- Solve posterior probability using logistic regression
- Matlab Exercises
-
Week 4: Naive Bayes & MLE
ILOs:
- Understand the theory of Naive Bayes and when its useful
- Understand how parameter estimation works, and the derivation of Maximum Likelihood Estimation equations.
-
Week 5: Gaussian Mixture Models and Expectation-Maximisation Algorithm
ILOs:
- Understand the GMM datasets
- Understand how the EM algorithm works for semi-parametric datasets to estimate the density and the parameters iteratively
-
Week 6: Non-Parametric Methods
ILOs:
- Understand that datasets don't fit known distributions all the time and can not be described by a set of parameters
- Use Density Estimation Methods methods to describe the datasets
- Differentiate Parzan Windows (Kernel) method and K-Nearest Neighbours method
-
Week 7: Linear Classification Methods
ILOs:
- Understand how linear classification works and their accuracy measures
- Understand Linear (Fisher) Discriminant Analysis approach
- Understand Perceptron Algorithm and its variations
-
Week 8: Linear Classification: SVM
ILOs:
- Understand higher dimensions mapping, and Kernel trick using dot product
- Understand Support Vectors Machines and how they work
-
Week 9: Non-Linear Methods 1: Decision Trees
ILOs:
- Understand how to build decision trees for non-metric datasets
- Understand how to calculate the Information and Uncertainty in Every Feature, to build the most optimal tree
- Research methods to pruning and approximating tree building algorithms
-
Week 10: Midterm Exam
The Exam document, dataset, and submission link will be available here Tuesday the 15th of December, 2015 @ 6 p.m.
Please sign up and enrol in the course, and make sure you can submit. The deadline is Sunday the 20th of December, 2015 @ 8 a.m. in the morning. Please submit different answers and different approaches to qualify for a grade. Good Luck.
To know your ID, look it up in the attached document. Contact me on manal.helal@gmail.com to confirm your username and enrol you in the course.
-
Week 11: Non-Linear Classification: Neural Networks
ILOs:
- Understand how multilayer perceptrons can solve non-linear classification problems
- Understand the different models of Building a Neural Networks Classifier
-
Week 12: Non-Linear Classification: Non-Linear SVM
ILOs:
- Familiarisation with the different kernel functions, particularly RBF.
- Understand the Non-linear SVM Formulation.