LEARNING ALGORITHM
| Collection | International Encyclopedia of Systems and Cybernetics |
|---|---|
| Year | 2004 |
| Vol. (num.) | 2(1) |
| ID | ◀ 1872 ▶ |
| Object type | Epistemology, ontology or semantics, Methodology or model |
A set of rules that does not rigidly determines the behavior of a system, but defines the basic characteristics and limits of all its potential behaviors.
The learning algorithm is implicit in any neural network which possesses a predetermined structure. Such a network organizes itself and constructs behavioral patterns by interaction with its environment. These patterns are conserved, or evolve within the limits of the organizational closure proper to the network.
The learning algorithm is a kind of second degree algorithm, that is able to produce numerous simpler and more specialized algorithms.
It allows an “economy” which, in G. BATESON's words “… consists precisely in not re-examining or rediscovering the premises of habit every time the habit is used” (1967, p.245).