Some Properties of Threshold Logic Unit Pattern Recognition Networks

作者: C.S. Weaver

DOI: 10.1109/T-C.1975.224209

关键词: Sample spacePartition (number theory)Convex hullLinear separabilityRegular polygonGeneral positionMathematicsPattern recognitionSet (abstract data type)Artificial intelligenceHyperplane

摘要: A condition for separation of two categories by a pattern-recognition network formed an array threshold logic units (TLU) and additional decision element is discussed. The implements set hyperplanes that partition the sample space, outputs vector elements form input to element. It shown this separable convex surface if, in each cells partition, hull first-category vectors cell does not intersect with second-category cell. When TLU can be zero another array, comparison shows capacity significantly increased. sufficient general position described. An algorithm suggested places so equal error rates occur on both sides. This property tends more finely regions overlap between categories.