K VICINITY METHOD ASSOCIATION MEMORY WHICH CAN BE RECONFIGURED

• Main Authors: YAMAZAKI SHOGO, MATTHEW HANSJUERGEN
• Format: Patent
• Language: English; Japanese
• Published: 01.09.2016
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; INFORMATION STORAGE; PHYSICS; STATIC STORES

PROBLEM TO BE SOLVED: To provide a k vicinity method association memory which can expand a dimension and a number of reference data.SOLUTION: A k vicinity method association memory (100) which can be reconfigured comprises: a clock count type association memory (10) which can be reconfigured; and a k vicinity clustering circuit (30). The association memory (10) comprises: plural element circuits (70); and plural switch circuits (50) for optionally connecting the plural element circuits. Until some k pieces of match signals out of plural match signals output from the association memory (10) become an active state, the k vicinity clustering circuit (30) selects class data corresponding to each and at least one of the k pieces of match signals, which become the active state, from plural pieces of class data indicating classes of plural pieces of reference data, for every time that at least one of the plural match signals become the active state, and then determines a class in which a number of data becomes maximum, when the all k pieces of class data being selected, are classified for every class.SELECTED DRAWING: Figure 8 【課題】参照データの次元数および個数の拡縮が可能なk近傍法連想メモリを提供する。【解決手段】再構成可能なk近傍法連想メモリ(100)は、再構成可能なクロックカウント式連想メモリ(10)と、k近傍クラスタリング回路(30)とを備える。連想メモリ(10)は、複数のエレメント回路(70)と、それらを任意に接続する複数のスイッチ回路(50)を含む。k近傍クラスタリング回路(30)は、連想メモリ(10)から出力される複数のマッチ信号のうちいずれかk個のマッチ信号がアクティブになるまでの間、複数のマッチ信号の少なくとも一つがアクティブになるごとに、複数の参照データのクラスを表す複数のクラスデータから当該アクティブになった少なくとも一つのk個のマッチ信号のそれぞれに対応するクラスデータを選択し、当該選択した全部でk個のクラスデータをクラス別に分類した場合においてデータ数が最大となるクラスを判定する。【選択図】図８