What Is a Neural Network?
Models are used throughout the cognitive sciences primarily to illustrate and test theories and to generate new predictions. Scientists develop a model, use data to test the model and compare the outcomes to what would be expected based on real world phenomena. Neural networks, (also known as Connectionist Models or Parallel Distributed Processing - PDP) are one of two ways to computationally model aspects of cognition.
This type of computation differs greatly from that of the classical computer inspired by Turing. Rather than manipulation of abstract symbols based on set rules or algorithms, connectionist models are dynamical systems that rely on causal processes by which units excite and inhibit each other in order to produce unpredictable emergent behaviour.
The image above shows an illustration of a neural network model; each individual unit is connected by lines showing the other units that it excites or inhibits.