SYNAPTIC WEIGHTS
Charles François (2004). SYNAPTIC WEIGHTS, International Encyclopedia of Systems and Cybernetics, 2(2): 3299.
| Collection | International Encyclopedia of Systems and Cybernetics |
|---|---|
| Year | 2004 |
| Vol. (num.) | 2(2) |
| ID | ◀ 3299 ▶ |
| Object type | Discipline oriented, Methodology or model |
Balancing coefficients applied to the inputs of artificial neurons in a network.
This expression is derived from the biological “threshold automat” model of brain neurons (W.S. Mac CULLOCH and W. PITTS, 1943). The weighted inputs are subsumed and the result triggers the element action or inhibits it.
Neural networks “learn” through the progressive determination of the synaptic weights which correspond to repeated inputs of the same types. Values of the coefficients become progressively stabilized through feedbacks from the outputs, which themselves tend to stabilize.
See also
'figure below'Parallel Distributed Processing