Artificial networks offer a new perspective on how learning mechanisms emerge
MIT researchers studied neural networks that can construct aspects of their own learning machinery.
Conventional machine-learning models often depend on update rules designed in advance. The researchers asked whether a system could develop parts of the mechanism that guides its learning.
The reported work examines artificial networks that can build learning-related machinery rather than merely adjust connections under a completely fixed procedure.
Such models may help researchers test ideas about adaptation and plasticity, while also suggesting ways to design learning systems.
An artificial neural network is a mathematical model. These results do not show that a biological brain learns using precisely the same mechanism.
Source and verification
This article is an original editorial synthesis of the research or institutional report cited below; it is not a reproduction of the source.
MIT News · Original publication ↗