Hebbian learning

<artificial intelligence> The most common way to train a neural network; a kind of unsupervised learning; named after canadian neuropsychologist, Donald O. Hebb.

The algorithm is based on Hebb's Postulate, which states that where one cell's firing repeatedly contributes to the firing of another cell, the magnitude of this contribution will tend to increase gradually with time.

This means that what may start as little more than a coincidental relationship between the firing of two nearby neurons becomes strongly causal.

Despite limitations with Hebbian learning, e.g., the inability to learn certain patterns, variations such as Signal Hebbian Learning and Differential Hebbian Learning are still used.


< Previous Terms Terms Containing Hebbian learning Next Terms >
heat sink
heat slug
heavy metal
heavy wizardry
Hebbian learning
Helen Keller mode
hello packet
hello, sailor!