NEURAL COMPUTATION
| Collection | International Encyclopedia of Systems and Cybernetics |
|---|---|
| Year | 2004 |
| Vol. (num.) | 2(2) |
| ID | ◀ 2277 ▶ |
| Object type | Human sciences, Epistemology, ontology or semantics, Methodology or model |
- “The construction of invariances by means of neural mechanisms” (St. KATZ, 1976, p.44).
St. KATZ adds: “When these invariances are assigned a name and given significance or meaning within some framework, they become what POWERS calls ”controlled quantities“ (Ibid).
In other words, the neural network constructs its algorithms, through its own activity. In psychology, POWERS field, neural computation may be redefined “… as a constructive activity occuring within the subject” (Ibid).
KATZ observes that the subject “… may attach ontological significance to the items it knows, (but) those who see the subject-environment relation within the framework of neural computation believe differently” (Ibid).
Neural computation can be considered as the root of connectionism. G. FERRY writes: “The connectionists…believe that systems can be developed, which, if connected together and endowed with sufficient complexity, will automatically process information and manipulate knowledge. Connectionists are strongly influenced by biological systems. These biological models give hints about how information is stored and then processed in the brain”. G. FERRY traces back this idea to D. HEBB views about the weighed flexibility of neurons connections with their neighbours and the reinforcement of pathways by continuous use (1986, p.36).
The basics of neural computation could introduce a new concept of Artificial Intelligence, not anymore relying on strict preprogramming, but based on progressive training through numerous examples of objects or processes, i.e. closer to natural intelligence.