Both work with (discrete or continuous) feature-based reps, and
support learning.
Differences:
- NNs distributed, BNs localized,
- BNs easier for humans to choose structure,
- BN semantics better understood,
- BNs handle multiple values and probabilities at the same time,
- NN inference is easy, BN inference is NP-hard,
- hard to compare speed of learning.
Up: LEARNING
Previous: Learning Problem