MINIMIZATION ALGORITHM (Adaptive)
| Collection | International Encyclopedia of Systems and Cybernetics |
|---|---|
| Year | 2004 |
| Vol. (num.) | 2(2) |
| ID | ◀ 2133 ▶ |
| Object type | Methodology or model |
An algorithm that schedules different hints for learning in a way that achieves a balance among them and with the training set.
Y.S. ABU MUSTAFA, who introduces the concept, writes: “The A.M.A. is adaptive in the sense that it constantly evaluates how well the machine is satisfying both the hints and the training examples, and continually modifies the adjustable parameters”
The term “minimization” “reflects that the algorithms is trying to minimize a quantitative measure of the error between the current actions of the machine and the behavior ultimately desired for it” (1995, p.73).
This concept could possibly be extended to the study of natural learning networks. Markets, for example, seem to search in an adaptive way for dynamic equilibrium.
The difficulty is to understand and interpret the behavior of participants in a self-organizing A.M.A.