Title :
A geometric algorithm finding set of linear decision boundaries
Author :
Takada, Yoshihiro ; Zhuang, Xinhua ; Wakita, Hisashi J.
Author_Institution :
Dept. of Inf. & Comput. Sci., Osaka Univ., Japan
fDate :
7/1/1994 12:00:00 AM
Abstract :
Proposes a geometric algorithm for finding all linear decision boundaries each of which correctly separates two nonoverlapping classes of pattern vectors. When the given T patterns are linearly separable, the algorithm terminates in T iterations, even though the computational time for each iteration tends to increase. When the given T patterns are linearly inseparable, the algorithm is able to detect it in at most T iterations
Keywords :
computational complexity; computational geometry; iterative methods; optimisation; pattern recognition; computational time; geometric algorithm finding set; iterations; linear decision boundaries; nonoverlapping classes of pattern vectors; Array signal processing; Associative memory; Computer networks; Distributed processing; Information processing; Neural networks; Physics computing; Signal processing; Signal processing algorithms; Vectors;
Journal_Title :
Signal Processing, IEEE Transactions on
