Neural Networks
Parametric description
| Lecturer | ||
|---|---|---|
| Andrzej Kordecki, Ph.D. | ||
| Course number and semester | ECTS | Classes per week |
| Neural Networks (1130-POW00-MSA-2012 Neural networks for classification and identification (1130-EMARO-MSA-1005) |
1130-POW00-MSA-2012: ECTS – 3 1130-EMARO-MSA-1005: ECTS – 5 |
1130-POW00-MSA-2012: W - 2 1130-EMARO-MSA-1005: W-2, C-1 |
Download
Prerequisites
- Basics of mathematical analysis, algebra and statistics from the previous years of studies,
- Basics of optimizing methods nonlinear functions,
- Basic of numerical methods and programming in Python languages.
Course Aims
- Fundamentals of neural network operation and training. Knowledge and skills necessary for independent problem solving using neural networks.
- Fundamentals of neural networks types and applications with an indication of advantages and limitations.
Lecture:
- Introduction: basic ideas, history, applications. Description of the neuron, activation functions, basic characteristics of neural networks. Neural network structures: feedforward and recurrent.
- Machine learning types: supervised, unsupervised and reinforced. Characteristics of input data and data augmentation in neural network learning. Application of neural networks in static and dynamic modeling.
- Loss function and evaluation measures of network results used in regression and classification. Perceptron learning rule. On-line and off-line training. Gradient optimization methods in neural network learning. The Widrow-Hoff rule. Back propagation algorithm and its characteristics.
- Characteristics of overfitting and underfitting in neural networks training. Validation methods. Generalization methods of neon networks: regularization, dropout, batch normlization.
- Neural networks with radial base functions. Support vector machine. Self-organizing maps, Kohonen networks. LSTM recurrent network
- Convolutional neural networks: description of ideas, structures, learning and their characteristics. Description of basic structures of neural networks in applications of computer vision
Exercise:
- Basics of Python and Matlabie programming,
- Tensorflow and Keras libraries in neural network problem formulation,
- Practical applications of neural networks.
Recommended reading
- S. Haykin, Neural Networks and Learning Machines, Third Edition, Prentice Hall, 2009
- I. Goodfellow, Y. Bengio, A. Courville, Deep Learning, MIT Press, 2016
- G. Dreyfus, Neural Networks: Methodology and Applications, Springer, New York 2005
- S. S. Haykin, Neural Networks and Learning Machines. Prentice Hall, 2009.