Dansk - English
Short version - Full version
Artificial Neural Network (Spring 2008) |
|||
| Course code : | EANN-U01 | ||
| ECTS Credits : | 7,5 | Status : | Optional |
| Revised : | 05/02 2008 | Written : | 11/05 2004 |
| Placement : | 5-7 semester | Hours per week : | 4 |
| Length : | 1 semester | Teaching Language : | Danish if no English students are present |
| Objective : | • To give the student a basic knowledge of Artificial Neural Networks (ANNs), i.e. the principles of neural networks, learning algorithms, and their testing and usage • To make the student able to analyse a given classification problem and argue for the selection of the appropriate neural network for its solution • To let the student gain experience in designing neural networks such as multilayer perceptrons, partially recurrent nets, and RBF nets.. • To give the student a basic understanding of advanced neural network technology, and thus to make the student able to follow the field and study its literature. |
||
| Principal Content : | Introduction to the principles of biological nervous systems with emphasis on the human brain and its neurons. Artificial Neural Networks (ANNs) and their use:: • Adaline, perceptron, multilayer perceptrons • Linear models, estimation, regression, least square, LMS algorithm • Pattern recognition, classifiers, supervised and non supervised learning • Training algorithms, back propagation, adaptive systems • Non-linear models, perceptron, MLP, training, classification, error and stop criteria • Function approximation with MLP, Radial Basis Function networks, vector support machines • Hebbian learning, Oja´s rule, anti-Hebbian learning, Associative Memory, Winner-take-all networks, • Self Organising Maps and their use • Use of ANNs in digital signal processing, with emphasis on, adaptive filters • Static versus Dynamic systems, time-delay neural network • Training and using recurrent networks, feedback parameters, Hopfield networks, • Introduction to pulsed neural network and their electronic equivalents |
||
| Teaching method : | The curriculum is based on an interactive book – overview lessons will allow (groups of) the students to study the interactive sessions and perform the related simulation exercises. Mat Lab examples and code will be used. | ||
| Required prequisites : | Math, analogue and digital design equivalent to 4.th semester level like DSM3/DSM4 | ||
| Recommended prerequisites : | Basic knowledge of statistics is a benefit |
||
| Relations : | The principles of neural networks are used in many other fields such as Decision analysis, Control-theory, Signal and Image-processing, Pattern recognition, Noise reduction, Robot-control, Sonar-analysis, Radar tracking systems, Planning, and Microwave simulation. | ||
| Type of examination : | Look under remarks | ||
| External examiner : | Internal | ||
| Marking : | Passed/Not passed | ||
| Remarks : | Internal, written examination. 7-point grading scale The grading is based on assignment reports and a written exam. To pass the course the students must be present at min. 80% of the lessons and each student must hand in reports on 2 mandatory practical assignments, which must be accepted. Evaluation criterias: 12: For an excellent performance. 7: For a good performance. 02: For an adequate performance. |
||
| Teaching material : | Neural and Adaptive Systems, Fundamental through simulations by Jose C. Principe, Neil R. Euliano, W. Curt Lefebvre, ISBN 0-471-35167-9 from 1999 (A cd containing the electronics version of the book is included with the book together with version 3 of the simulation tool Neurosolutions). (newest edition) | ||
| Responsible teacher : | |||