Title :
Binary classification of real sequences by discrete-time systems
Author :
Kaliski, M.E. ; Johnson, T.L.
Author_Institution :
Northeastern University, Boston, Massachusetts
Abstract :
This paper considers a novel approach to coding or classifying sequences of real numbers through the use of (generally nonlinear) finite-dimensional discrete-time systems. This approach involves a finite-dimensional discrete-time system (which we call a real acceptor) in cascade with a threshold type device (which we call a discriminator). The proposed classification scheme and the exact nature of the classification problem are described, along with two examples illustrating its applicability. Suggested approaches for further research are given.
Keywords :
Automata; Erbium; Pattern recognition;
Conference_Titel :
Decision and Control including the 17th Symposium on Adaptive Processes, 1978 IEEE Conference on
Conference_Location :
San Diego, CA, USA
DOI :
10.1109/CDC.1978.268030