Connectionist Architectures
Based (often very loosely) on constraints from neural processing in the brain.
- Large numbers of “neuron-like” processing elements.
- Large number of weighted connections between elements. Knowledge is encoded as the weights.
- Highly parallel, distributed control.
- Emphasize learning internal representations rather than having them hand programmed.
To result in systems that are:
- Fast and resistant to damage.
- Able to generalize from its inputs.
- Able to learn efficiently for large-scale masses.