Title :
Training dynamics of systems in a variable environment
Author :
Friesen, Donald K. ; Clingman, W.H.
Author_Institution :
Dept. of Comput. Sci., Texas A&M Univ., College Station, TX, USA
Abstract :
The authors describe an approach to bounded training time and robustness in learning systems such as neural networks that is based on the topological properties of an associated flow. The method uses data transformations to create a modified problem for which the desired topological properties hold. A simple example, the two-layer perceptron, is used to illustrate the concepts considered here
Keywords :
feedforward neural nets; learning (artificial intelligence); topology; associated flow; bounded training time; data transformations; neural networks; robustness; topological properties; training dynamics; two-layer perceptron; variable environment; Computer science; Control systems; Flow production systems; Neural networks; Noise robustness; Orbits; State-space methods; Terminology; Testing; Topology;
Conference_Titel :
Systems, Man and Cybernetics, 1992., IEEE International Conference on
Conference_Location :
Chicago, IL
Print_ISBN :
0-7803-0720-8
DOI :
10.1109/ICSMC.1992.271600
