|
Course Highlights
This two-day course provides a broad introduction to the basic concepts of artificial neural networks. Numbers of neural network architectures and their training algorithms are reviewed. Examples of neural networks architectures that are covered in this course are single layer perceptrons, multilayer perceptrons, radial basis function and Hopfield neural networks. Applications that are covered are object recognition, character recognition, handwritten word and digit recognition.
Who Must Attend
Engineers, researchers, scientists, managers from the manufacturing, government and defense sectors as well as graduate students who are interested in acquiring the basic technical knowledge in neural network that can be applied in disciplines, such as signal and image processing.
Prerequisites
Attended “MATLAB Fundamentals”, experience with basic computer operations, and having experience on image processing field. Attended “Applying Image Processing Techniques with Matlab and Simulink” is strongly recommended.
Course Benefits
Upon completion, you will be able to:
- understand the fundamental concepts of artificial neural networks techniques
- distinguish between the classical pattern recognition algorithms and the neural network techniques
- compare the relative merits of various neural networks, i.e single layer Perceptrons, multilayer Perceptrons, and radial basis function
- explain supervised and unsupervised training algorithms
- describe the typical applications of neural networks to signal and image processing problems
Course Outline
Neural Network: A Brief Introduction
Presents a brief history of neural network and its MATLAB terminology:
- What is Neural Network
- Neural Network Application
- Definition of Neural Network
- Biological prospective of Neural Networks
- Simple Neuron Model
- MATLAB representation of the simple neuron model.
- Architecture of neural network
- Data structures
Perceptrons
Learn about the simplest form of neural network and its applications:
- Introduction
- Linearly separable problems
- The perceptron neuron
- MATLAB representation of the perceptron neuron.
- The perceptron architecture
- Creating a perceptron
- The Common-line approach.
- Perceptron learning rule
- Training of perceptron
Linear Networks
Learn about a more advance form of neural network for solving linearly-separable problems:
- Introduction
- Linear Neuron
- MATLAB representation of the perception Neuran.
- The architecture of Linear Networks
- The Widrow-Hoff learning Algorithm
- Linear clarification
- Adaptive filtering
- Designing adaptive filtering
Backpropagation Networks
Learn about one of the most popular neural networks, renowned for its strength in solving highly complex and non-linear problems:
- Introduction
- Architecture of Feedforward BP network
- MATLAB representative of Feedforward BP network
- Transfer function of BP networks
- Learning algorithms for backpropagation networks
- Training of backpropagation network
- Batch gradient descent training
- Batch gradient descent with momentum
- Faster training
- Comparison of Training algorithm
- Improving generalization with early stopping
- Preprocessing and postpossessing
Radial Basis Networks
Learn about an alternative form of neural network to backpropagation networks:
- Introduction
- Radial basis Neuron Model
- Generalized Regression Networks
- Probabilistic Neural Networks
Applications
Apply Neural Network Techniques in Image Processing field:
- object recognition
- character recognition
- handwritten word and digit recognition
|
