NETWORK (Learning)
| Collection | International Encyclopedia of Systems and Cybernetics |
|---|---|
| Year | 2004 |
| Vol. (num.) | 2(2) |
| ID | ◀ 2267 ▶ |
| Object type | Human sciences, Methodology or model |
G. PASK states that learning networks (which are self-organizing systems — see above Network, Evolutionary) need “…input filters which must be able to detect arbitrarily chosen attributes”.(1961, p.86). In the case of the brain, for instance, those arbitrarily chosen attributes are selected through the sense organs: we are able to register red or blue light, but not infra-red or ultra-violet one. Moreover, we must be able to pick up distinctions as for example differentiate red and blue, or in natural languages, specific sounds: a, i, u. (Characteristically, these written signs represent different sounds in different languages).
PASK observes moreover: “In our social environment we learn, not new attributes, but the yen to regard certain values of the attribute as identical” (Ibid).
Thus, the English sound 'a' is different from the French 'a' and, even from the Australian or North-American one. What one hears or sees is what one has been culturally trained to hear or see, within the basic genetic determinism of human speech or sight sensorial structures.
The same must be true — even if in a very different setting — for electronic networks constructed to recognize words or sounds.