Models of psychological and cognitive phenomena that are based on connectionist processing have recently been described. These include Norris's back propagation model, Schyns's model based on Kohonen nets and Hinton and Shallice's model that adds a recurrent layer to an MLP type network. This paper looks at some of the ways data are represented and at the metrics that are employed in these models. It investigates the techniques that are appropriate for different processing tasks in connectionist networks. The term "connectionist network" is an alternative to "neural network", which is the name more often used in the computer science literature.
|Name||UH Computer Science Technical Report|
|Publisher||University of Hertfordshire|