How to use this Workbook

Copyright © Simon Dennis and Devin McAuley, 1998. Connectionist Models of Cognition is a workbook designed to introduce neural networks to undergraduate and postgraduate students either within the context of a course or by a program of self study. The first two chapters provide an introduction to neural networks and to the BrainWave simulator that is embedded throughout the textbook. Chapters four to eight cover a set of neural architectures that have been instrumental in the development of the field of artificial neural networks and that illustrate the key concepts necessary for understanding the area. The remaining chapters cover specific connectionist models of cognitive phenomena such as letter perception and recognition memory.

The approach is very hands on. The simulator is embedded throughout the text - as living figures. Learning will be facilitated greatly if students complete the exercises as they go rather than returning to them at a later date (or not at all). Learning by doing is a key part of the philosophy of the text.

The BrainWave simulator was developed in conjunction with the ic230 Introduction to Neural Networks class at the University of Queensland. It employs a highly graphical, direct manipulation interface - much like a drawing program - making it very easy to use. Care has been taken to ensure that all relevant information about a given network is visible (and that irrelevant information is not visible) so that students are not required to search for obscure menu options to see what is happening. The simulator is written in Java meaning that it can be run directly from web browsers. Because all of the course materials are online, updates and extensions (including new architectures) are available immediately so that you always have access to the most up-to-date versions.

Chapters four to eight each explain an architecture and highlight an important issue or concept that that architecture demonstrates. So, for instance, chapter six explains the Hopfield architecture and uses that architecture to highlight the idea of the state of a system descending on an energy surface. Descent on an energy surface, however, is a recurring theme in neural networks and is not just confined to the Hopfield architecture. Each chapter begins with a general introduction to the issue and architecture and moves very quickly into a hands on demonstration of it in operation - in keeping with the philosophy of the workbook. Then the architecture is described in detail and any interesting dynamical properties are explained. The chapters end with a demonstration of how the architecture embodies the issue or concept of interest.

The text is designed to provide an intuitive grasp of neural networks rather than an in depth mathematical treatment. All of the mathematics is accompanied by verbal description and sections which contain mathematics that can be skipped without distorting the basic understanding of an architecture are indicated in the Table of Contents with a "*". However, a deeper understanding requires mathematics and some background in linear algebra and calculus is an advantage. Even if you decide to omit the mathematics sections on a first reading we suggest that you return to them to improve your understanding.

We hope that you find Connectionist Models of Cognition an enjoyable and instructive resource. If you have any questions or comments about the book please don't hesitate to contact us at connectionist@connectionist.net.

Shalom,

Simon Dennis and Devin McAuley.

---

The Connectionist Models of Cognition homepage.