INFORMATION PROCESSING DEVICE, DATA DECOMPOSITION METHOD, AND DATA DECOMPOSITION PROGRAM

To provide an information processing device, a data decomposition method, and a data decomposition program, which can obtain approximate data that has a higher approximation degree with respect to predetermined data and is the product of basis data and coefficient data.SOLUTION: A data decomposition...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: ABE MITSURU, KADOWAKI MASASHI
Format: Patent
Sprache:eng ; jpn
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:To provide an information processing device, a data decomposition method, and a data decomposition program, which can obtain approximate data that has a higher approximation degree with respect to predetermined data and is the product of basis data and coefficient data.SOLUTION: A data decomposition method approximates a weight matrix by an approximate matrix comprising a combination of a basis matrix and a coefficient matrix. Accordingly, an information processing device selects a basis matrix candidate which is a candidate that can constitute the approximate matrix, obtains an approximate matrix candidate using the basis matrix candidate, calculates an evaluation metric that evaluates the approximate matrix candidate, generates a regression model representing a relationship between the evaluation metric and the basis matrix candidate that constitutes the approximate matrix candidate, repeats processing in an evaluation metric calculation process and a regression model generation process at least once using the selected basis matrix candidate, and calculates the coefficient matrix using the basis matrix candidate that constitutes the approximate matrix candidate whose evaluation metric has a desirable value when a predetermined termination condition is satisfied.SELECTED DRAWING: Figure 5 【課題】所定のデータに対してより近似度の高い基底データと係数データとの積である近似データを得ることができる、情報処理装置、データ分解方法及びデータ分解プログラムを提供する。【解決手段】データ分解方法は、重み行列を基底行列と係数行列との組合せで構成される近似状列で近似する。このため情報処理装置は、近似行列を構成しうる候補となる基底行列候補を選択し、基底行列候補を用いて近似行列候補を求め、近似行列候補を評価する評価指標を算出し、評価指標と近似行列候補を構成する基底行列候補との関係を表す回帰式を生成し、選択された基底行列候補を用いて、評価指標算出工程及び回帰式生成工程での処理を少なくとも1回以上繰り返し行う。所定の終了条件を満たした時点で、評価指標が望ましい値を持つ近似行列候補を構成する基底行列候補を用いて係数行列を算出する。【選択図】図5