Title :
Perceptron learning in the domain of graphs
Author :
Jain, Brijnesh J. ; Wysotzki, Fritz
Author_Institution :
Dept. of Electr. Eng. & Comput. Sci., Technische Univ. Berlin, Germany
Abstract :
We develop a new mathematical framework, which embeds weighted graphs into quasi metric spaces. This concept establishes a theoretical basis to apply neural learning machines for structured data. To exemplarily illustrate the applicability of metric graph spaces, we propose and analyze a perceptron learning algorithm for graphs in its primal and dual form.
Keywords :
graph theory; graphs; learning (artificial intelligence); perceptrons; graphs domain; metric graph spaces; neural learning machines; perceptron learning algorithm; quasi metric spaces; structured data; weighted graphs; Algorithm design and analysis; Classification algorithms; Computer science; Electronic mail; Extraterrestrial measurements; Linear discriminant analysis; Machine learning; Neural networks; Training data; Vectors;
Conference_Titel :
Neural Networks, 2003. Proceedings of the International Joint Conference on
Print_ISBN :
0-7803-7898-9
DOI :
10.1109/IJCNN.2003.1223713
