Fyfe C. Artificial Neural Networks

Department of Computing and Information Systems, The University of Paisley, 1996, -136 p.This course is an introduction to the subject of Artificial Neural Networks and Genetic Algorithms, two very new subjects forming part of Distributed Artificial Intelligence.
As you leaf through these notes you will notice that they are full of mathematical equations. The reason is simple: these subjects are inherently mathematical. However the course and assessments are such that it will be possible for you to pass if you do not touch the equations. However if you wish to gain a good pass you must attempt to master the equations. In addition, by doing so, you will gain a deeper insight into the operation of these two exciting technologies.
The aims of the course are that you should be able to
1. identify tasks which can be solved by these methods
2. identify and use the appropriate specific neural network or genetic algorithm for specific tasks
3. implement the method
Since these are broad aims, we begin each Chapter with a statement of objectives. We make explicit reference to these objectives in the Exercises at the end of each Chapter but expect you to be aware of these objectives when you are in the laboratory.
The laboratory work is designed to both teach you the contents of the course and assess your understanding of the algorithms. You will have one two-hour laboratory session each week. The first few laboratories are using a simulator which you can run by simply pointing and clicking the mouse. This is to allow you to concentrate on the high level features of the Artificial Neural Network you are using and not get involved in the details which you have to be aware of when you are programming a network. However in subsequent weeks you will have to program your own nets. This is an essential and assessable part of the course.
You will have two lectures per week which will be mainly devoted to me telling you about the charms of ANNs and GAs. There will however be a 5-10 minute slot in each lecture for daft questions. The "Daft Question" slot is important so please use it to ask any question (about the course) which comes to mind. All questions will be treated as equally daft.
You will have one tutorial per week. All tutorials are activity based - you will be directed to a piece of work each week and expected to report on the outcome of your (group's) deliberations during that tutorial.Associative Memory
Simple Supervised Learning
The Multilayer Perceptron: backprop
Unsupervised learning
Genetic Algorithms
Recurrent Networks
Faster Supervised Learning