Hebbian learning is the principle that when one neuron repeatedly takes part in firing another, the connection between them strengthens. It is usually summarized in the slogan cells that fire together wire together. The rule offers a local, biologically plausible account of how experience could leave a physical trace in the brain: no global supervisor is needed, since each synapse adjusts itself based only on the activity of the two neurons it joins.
The idea comes from Donald Hebb’s 1949 book “The Organization of Behavior”, where he proposed that coincident activity drives lasting growth in synaptic efficiency and introduced the notion of the cell assembly, a group of neurons that learn to activate together as a unit. Decades later, the discovery of long-term potentiation gave Hebb’s postulate concrete physiological support, showing that synapses really do strengthen with correlated activity.
In artificial systems, Hebbian and Hebb-derived rules are among the simplest ways to make a network learn, and they underlie unsupervised methods that find structure in data without labeled targets. They also appear in Hopfield networks, where Hebbian storage lets a network settle into remembered patterns. While backpropagation became the dominant training method for deep networks, interest in local, Hebb-like rules persists among researchers seeking learning that is more biologically plausible or more efficient on neuromorphic hardware.
For a general reader, Hebbian learning is worth knowing as the original bridge between a fact about biology and a recipe for machine learning, and as a reminder that the deepest idea in the field, learning by adjusting connections, predates modern computers.