Connectionist models are based on networks of simple processing units, called nodes or neurons, which are linked by weighted connections.
Learning usually occurs by adjusting these connection weights to transform inputs into desired outputs.
Typical examples are artificial neural networks, which are trained using algorithms such as backpropagation.
Connectionist models are characterized by parallelism, distribution, and robustness to disturbances, and are well suited for pattern recognition and cognitive tasks.
They contrast with symbolic models, as they encode knowledge not explicitly, but distributed across the connections.