Hebbian theory

Hebbian theory explains learning as changes in connection strength driven by correlated neural activity. Its best-known rule strengthens a synapse when its presynaptic and postsynaptic neurons are active together.

Hebbian learning rule: A learning rule that changes synaptic strength according to correlated activity in connected neurons. It states the central proposal in operational terms: correlated pre- and postsynaptic activity strengthens a connection.
Donald Hebb: A Canadian psychologist and neuroscientist whose work connected neural organization with learning and behavior. His 1949 book proposed the neural principle that bears his name.
Neuron: An electrically excitable cell that receives, processes, and transmits information through neural signals. Hebbian theory describes learning through changes in connections between these cells.
Associative memory: Memory in which one item or cue evokes another because they have been linked through experience. Hebbian strengthening offers a neural account of how co-occurring cues become associated.
Anti-Hebbian learning: A learning rule that weakens connections when pre- and postsynaptic activity are correlated. It reverses the sign of the central Hebbian association rule.
Long-term potentiation: A persistent increase in synaptic strength following particular patterns of neuronal activity. It is a major experimental example of lasting synaptic strengthening associated with Hebbian learning.
The Organization of Behavior: Donald Hebb’s 1949 book proposing that learning organizes groups of neurons through changes in their connections. It introduced the influential account from which Hebbian theory developed.
Synapse: A specialized junction through which one neuron communicates with another cell. The connection whose strength changes is usually a synapse between neurons.
Hopfield network: A recurrent artificial neural network that stores patterns as stable states and retrieves them from partial cues. Its learned connections use a Hebbian-style rule to store patterns.
Perceptron learning rule: A supervised learning rule that adjusts weights according to classification errors. Unlike basic Hebbian learning, it uses an explicit target and error signal.

Linked from 9 pages