Tutor: Dipl.-Inf. Kai Labusch

Time and place: Do 8:30 - 9:45, Room H1

Assigned to: Neuroinformatics V

Exercise 1: Artificial neurons for supervised and unsupervised learning

pdf-file

(to be handed in 2009-11-12)

Useful files for exercise 1:

• ueb11.mat: MATLAB data file
• learn_neuron.m: program frame for perceptron learning
• characteristics.m: MATLAB helper function
• plotPlane.m: MATLAB helper function
• show_classes.m: MATLAB helper function
• kmeansdata.mat: MATLAB data file
• kmeans.m: program frame for k-means clustering
• kmeans_visualize.m: MATLAB helper function

Exercise 2: Principle Component Analysis (PCA)

pdf-file

(to be handed in 2009-11-26)

Useful files for exercise 2:

• faces.mat: MATLAB data file
• pcaproj.m: MATLAB helper function
• dctcoeff.m: MATLAB helper function
• dispbasis.m: MATLAB helper function

 

Exercise 3: PCA and Entropy

pdf-Datei

(to be handed in 2009-12-10)

Useful files for exercise 3:

• gauss2d.mat: MATLAB data file
• pcaneuron.m: MATLAB program template
• dist2d.mat: MATLAB data file

Sample solution:

• rgb_ent.m: Entropy of RGB transformations
• pca_neuron_mus.m: Performing PCA by neuronal learning
• pcafaces_mus.m: Learning of face features by PCA

Exercise 4: Independent Component Analysis (ICA)

pdf-Datei

(to be handed in 2010-1-7)

Useful files for exercise 4:

• icadist2d.mat: MATLAB data file
• plotjoint.m : MATLAB visualization function
• audiodata.mat: MATLAB data file

Obtaining the ICA software package:

• Download the FastICA software package
• Extract the archive (use either tar or unzip)
• Add the FastICA directory to your MATLAB path (addpath command)

Sample solution:

• ica_by_hand.m: Performing ICA by hand
• cocktail_party.m: Demixing sound signals by ICA
• faces_ica.m: Learning of face features by ICA

Exercise 5: Statistical Learning Theory

pdf-Datei

(to be handed in 2010-1-21)

Sample solution:

• test_hoeffding.m: Testing Hoeffding's inequality
• growth_func.m: Computing the growth-function

Exercise 6: Support Vector Machine

pdf-Datei

(to be handed in 2010-2-4)

Useful files for exercise 6:

• data.mat: MATLAB data file

Sample solution:

• std_svm.m: Standard SVM approach (quadratic programming)
• doublemin.m: DoubleMinOver SVM approach

Literature

- A probabilistic approach to early computational vision, A. Hyvärinen, J. Hurri, P. Hoyer

Independent Component Analysis, A. Hyvärinen, J. Karhunen, E. Oja

The Nature of Statistical Learning Theory, V. Vapnik

Neural Network Learning: Theoretical Foundations, M. Anthony, P. Bartlett

An Introduction to Support Vector Machines and other kernel-based learning methods, N. Cristianini, J. Shawe-Taylor

Overview: Pattern Recognition and Machine Learning, C. Bishop