Title :
Combined and diffeo-quadratic Hebbian neural systems: initial concepts
Author_Institution :
Gen. Dynamics, San Diego, CA, USA
Abstract :
Summary form only given, as follows. Unsupervised learning laws could carry the greatest promise for the future of neural computing. These laws allow a system to adapt to its environment with the least amount of external information. Common unsupervised learning laws include signal Hebbian learning, differential Hebbian learning, and competitive learning. The author extends this list to include two more classes of autoassociative adapters-combined neural systems and diffeo-quadratic Hebbian neural systems. The learning and recall dynamics of these systems are extensions of existing systems. In addition to the mathematical characterization of these new systems, a preliminary stability analysis has been made.<>
Keywords :
learning systems; neural nets; autoassociative adapters; combined neural systems; competitive learning; diffeo-quadratic Hebbian neural systems; differential Hebbian learning; learning systems; neural computing; recall dynamics; signal Hebbian learning; stability analysis; unsupervised learning; Learning systems; Neural networks;
Conference_Titel :
Neural Networks, 1989. IJCNN., International Joint Conference on
Conference_Location :
Washington, DC, USA
DOI :
10.1109/IJCNN.1989.118345
