(Also available via video. Contact Kristi Wheeler - kristi.wheeler at okstate.edu)
Instructor: Martin Hagan
Office: 311 E.S. Phone: (405) 744-7340
Email: mhagan at okstate.edu
Neural Network Design (2nd Ed)
Authors: M. Hagan, H. Demuth, M. Beale, O. De Jesús
This course will be an introduction to the basic neural network architectures and learning rules. Emphasis will be placed on the mathematical analysis of networks and learning rules, and on the application of neural networks to certain engineering problems in pattern recognition, signal processing and control systems. The course will incorporate necessary background material (such as linear algebra, optimization and stability), while including extensive coverage of performance learning, like the Widrow-Hoff rule and backpropagation. Several enhancements of backpropagation, such as the conjugate gradient and Levenberg-Marquardt variations, will be discussed. Simple building blocks will be used to explain associative and competitive networks, including feature maps, learning vector quantization, and adaptive resonance theory. Recurrent associative memory networks, such as the Hopfield network, will also be presented.
The course grade will be based on two examinations, homeworks and quizzes, and a term project. The project will involve the implementation and testing of a neural network learning algorithm.
ECEN 5713, or equivalent background in linear algebra.