ARTIFICIAL NEURAL NETWORK
Course Objective:
To introduce the concept of artificial network as an alternative options for solving engineering problems.
- Working with data (4 hours)
- Data types; data, information and knowledge; concept of data mining; Dimension reduction of data matrix: Principal component analysis.
- Introducton of Artifical Neural Network (ANN)
(6 hours)
- Biological Analogy, Historical development; ANN terminology; network structure; basis functions; activation functions; advantages of ANN; application areas of ANN.
- Learning process & optimization techniques
(10 hours)
- supervised learning: Error correction learning, memory based learning
- unsupervised learning: Hebian learning, competitive learning
- learning with critic
- gradient descent and least mean square
- Derivative free optimization techniques: advantages of derivative free techniques; genetic algorithm: fundamental of GA and biological background.; GA operators & GA operation.
- Simulated annealing: theoretical background and algorithm.
- Supervised network (8 hours)
- McCullotch and Pitt Neuron; LTUs, simple perceptron and perceptorn learning. Limitation of simple percepron.
- ADDALINE network and delta rule
- Multilayer perceptron: Needs of multilayer network, generalized delta rule (error‐backpropagation), effect of momentum term and learning rate
- Error backpropagation learning of sigmoidal units; drawbacks of error‐backpropagation
- Unsupervised network (4 hours)
- competitive network: network structure & working;
- dissimilarity measures;
- Self Organizing Map and Kohonen learning;
- applications
- Special networks (4 hours)
- Radial basis function network: structure and working procedure, advantages
- LVQ network: structure and learning approach
- Hopefield network
- Autoassociative memory network: general structure and Purpose, Autocorrelator; Heterocorrelator
- Application of ANN in Electrical Engineering (8 hours)
- Fault diagnosis
- Control application
- Network planning
- Forecasting task.
- State estimation
- Unit commitment
Practical:
- Computer simulation of PCA.
- Computer simulation of perceptron network
- computer simulation of back propagation network
- A Shortterm case study demonstrating ANN application for a specific purpose.
References:
- Simon Hykin, "Neural networks A Comprehensive Foundation", second edition; Pearson Education.
|