Learning

• So what. These nets do these simple addition and multiplication tables. Why not just use the equations.
• The key to this (and just about all connectionist work) is that there are algorithms to learn the connection strengths.
• It seems that the brain has only one type of learning (aside from cell death and growth, and synaptic death and growth).
• It's called Hebbian learning (after Don Hebb who is popularly thought to have originated the idea).
• If a synapse connects two neurons, and both neurons tend to fire together than the weight is increased.
• At the physical level, that seems to be the only way you learn. (Brain physiology is not entirely known, but it's the only mechanism that is currently, widely accepted. Variable binding)
• Hebbian learning is used in connectionist systems, but other algorithms are also used for different systems.