Some basic principles of connectionist research are explained along with an account of a number of the techniques necessary for constructing connectionist models. The objective is to introduce the area to people with limited mathematical and computational backgrounds by reducing the examples to simple arithmetic. In this way, a solid basis will be provided for one of the learning algorithms that have been fundamental to the development of network learning: the Hebbian learning rule. After outlining the technique in detail, two examples are provided to make the ideas concrete. These are learning to associate visual features with words and learning case representations.
Connection science (also called Neural Network and Parallel Distributed Processing --PDP research) is a new information processing paradigm. Connectionist computing does not involve conventional computer programs in a yon Neumann architecture. Instead, it imitates the architecture and processes of the brain. The knowledge in connectionist systems is expressed as weighted connections in networks of very simple processing units. Furthermore, connectionist programs start from some initial configuration and then learn from the demands of their environment, rather than having programmers install the information that they think will be required. Much of the research effort in the field is directed towards finding new and more powerful learning algorithms like those used in the brain. This is an exciting new area which is bringing together many researchers from diverse disciplines such as computer science, physics, psychology, engineering, philosophy, linguistics, biology, artificial intelligence, and neuroscience. At last there is a common focus for these disciplines and, perhaps more importantly, a common language. Nowadays, it is quite usual to read in the literature of the various disciplines that connectionism represents a paradigm shift within cognitive science and within the individual disciplines themselves. But why should the 1980s be moving so rapidly towards the new paradigm when work on neural networks has been around since the 1940s? The answer lies in part in advances in