Selection of features into the object's own space based on the measure of its compactness

Abstract

 Z ( S d , ρ) ≠  . The compactness of the object S d ∊ K t on the set X ( d i t i d t X k S K O S Z S K         The object S i ∊ K i ,ρ)|/| K 3- t |,       3 , m i n , r t i j t j i j r S K g S K S S S S     3 -t ∩ Γ ( p ) is the nearest one. The set X ( u )  X ( n ), computed on E 0 \{ S i } as             m a x d d X k X n X u X k     is considered informative for the object S d ∊ K t , and the value of θ d ( X ( u )) is considered as a measure of its compactness. To implement the algorithm of step by step selection of features into an informative set, data preprocessing is performed. The purpose of preprocessing is to select the first pair ( x i , x j ) into an informative set based on the proposed criterion. The criterion is used to search for a cluster of data with a maximum density of descriptions of objects of one with S d class