Subject description - A0M31ASN

Summary of Study | Summary of Branches | All Subject Groups | All Subjects | List of Roles | Explanatory Notes               Instructions
A0M31ASN Algorithms and Structures of Neurocomputers Extent of teaching:2+2c
Guarantors:Tučková J. Roles:V Language of
teaching:
CS
Teachers:Tučková J. Completion:Z,ZK
Responsible Department:13131 Credits:5 Semester:Z

Anotation:

Information about the basic principles and possibility of the application of the neural informative technology for the signal processing are the main topic. The lectures are devoted to the introduction into the artificial neural networks (NN) theory and applications, to the choice and the optimisation of the structures, the choice of the data, and to the solutions of the classification. The neural network applications at the speech and image processing are investigated in detail. Some neural network applications in the biomedical engineering and hardware realization of the SOM are described. The applications are o focused to EEG and ECG processing, also to possibilities of applications ANN at physiotherapy,

Study targets:

Students acquire real experiences with the Neural Network Toolboxu from MATLAB for their project applications with MLNN and SOM. Our goal is the possibility to the familiarize with the perspective topics currented in the foreign countries. Our aim is also help to students with diploma work topic choice. The signal processing by artificial neural networks (normal or pathological speech and emotion analyses, recognition and synthesis) will be study.

Content:

The study is the introduction to the theory and applications of the most expanded artificial neural networks paradigms. The attention is dedicated to the multilayer neural network (MLNN), to the several options of the BPG training algorithm and to standard self-organizing maps (SOM) and also to the new variant - to the supervised self-organizing maps (SSOM). Individual projets are solved by MATLAB and by free software from the Helsinki technical university.

Course outlines:

1. Neural networks - research history, biological and artificial NN, applications
for signal processing, neural models, activation functions.
2. Learning principles, Self-Organizing Maps (SOM), Kohonen's maps.
3. Supervised SOM, U-matrix, LVQ classifier.
4. Multilayer networks (feedforward and Elman networks) with Back-Propagation learning algorithm (BPG).
5. Basic BPG, modifications.
6. Deep neural networks.
7. Optimisation of the structure, Data Mining. neural network pruning, Input data choice.
8. Support vector machine learning.
9. ANN and prediction and classification.
10. ANN applications in speech processing and emotion analysis. Basic terms of phonetics, characteristics
of the spoken speech (mormal and pathological).
11. Speech synthesizer. Image recognition.
12. ANN applications in neurology and rehabilitation medicine and in selected medical branches.
13. Special structures (CNN, TDNN, Wavelet NN, fuzzy-neuron networks). Genetic algorithm.
14. ANN realizations. neurocomputers. The others ANN applications.

Exercises outline:

1. Introduction, MATLAB, NN-Toolbox fundamentals, information of the semester
projects.
2. ANN basic function, Perceptron, ADALINE, MADALINE, LMS algorithm.
3. Self-Organizing Maps, supervised SOM, U-matrix. SOM Toolbox.
4. Kohonen's maps, LVQ algorithms - NN Toolbox, MATLAB.
5. Multilayer neural networks. Assignment of the semester projects.
6. Modifications of the BPG algorithm.
7. Deep neural networks.
8. Speech Laboratory - experiments. Assignment of the semester projects.
9. Presentation of the semester project thesis - control.
10. Pruning - ANN optimisation. Semester projects - consultations.
11. Experiments with neural network parameters. Semester projects - consultations.
12. Experiments with SOM Toolbox . Semester projects - consultations.
13. Semester projects - consultations.
14. Semester projects - evaluation, credits.

Literature:

1. Kohonen,T.: Self-Organizing Maps. Berlin Heidelberg, 3rd Edition, Springer Series in Information Sciences, Springer-Verlag, 2001, ISBN 3-540-67921-9.
2. Handbook of Neural Network Signal Processing.The Electrical Engineering and Applied Signal Processing Series. Ed.: Yu Hen Hu, Jenq-Neng Hwang. CRC Press, USA,2002, ISBN 0-8493-2359-2.
3. Haykin, S.: Neural Networks. A Comprehensive Foundation. Macmillan College Publishing Company, Inc. USA, 1994. 2nd.ed. 1998, Prentice/Hall, Upper Saddle River, NJ.
4. Program library SOM Toolbox 2.0. www.cis.hut.fi/projects/somtoolbox/download

Requirements:

Basic knowledge of the speech and image processing, MATLAB, probability calculus and statistics applications. Active participation on the seminars, develop semester project. More on http://amber.feld.cvut.cz/SSC

Note:

The range of the teaching: 14 weeks

Webpage:

https://moodle.fel.cvut.cz/courses/A0M31ASN http://amber.feld.cvut.cz/ssc

Keywords:

Neural Networks Multilayer Neural Networks Self-Organizing, Kohonen Maps Applications Signal processing

Subject is included into these academic programs:

Program Branch Role Recommended semester
MPEK3_2016 Electronics V
MPBIO1 Biomedical Informatics V
MPBIO2 Biomedical Engineering V
MPIB Common courses V
MPKME1 Wireless Communication V
MPKME5 Systems of Communication V
MPKME4 Networks of Electronic Communication V
MPKME3 Electronics V
MPKME2 Multimedia Technology V
MPEEM1 Technological Systems V
MPEEM5 Economy and Management of Electrical Engineering V
MPEEM4 Economy and Management of Power Engineering V
MPEEM3 Electrical Power Engineering V
MPEEM2 Electrical Machines, Apparatus and Drives V
MPKYR4 Aerospace Systems V
MPKYR1 Robotics V
MPKYR3 Systems and Control V
MPKYR2 Sensors and Instrumentation V


Page updated 19.7.2019 17:53:04, semester: Z,L/2020-1, L/2018-9, Z,L/2019-20, Send comments about the content to the Administrators of the Academic Programs Proposal and Realization: I. Halaška (K336), J. Novák (K336)