Hebbian learning

Hebbian learning

(artificial intelligence)The most common way to train aneural network; a kind of unsupervised learning; namedafter canadian neuropsychologist, Donald O. Hebb.

The algorithm is based on Hebb's Postulate, which statesthat where one cell's firing repeatedly contributes to thefiring of another cell, the magnitude of this contributionwill tend to increase gradually with time. This means thatwhat may start as little more than a coincidental relationshipbetween the firing of two nearby neurons becomes stronglycausal.

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

http://neuron-ai.tuke.sk/NCS/VOL1/P3_html/node14.html.