公开(公告)号：US20190012364A1

公开(公告)日：2019-01-10

申请号：US16130109

申请日：2018-09-13

Applicant: NEC Corporation

Inventor： Takuya ARAKI

IPC: G06F17/30

Abstract: An information processing device includes: a feature quantity obtaining unit which obtains a feature quantity of an object to be extracted, which is extracted from a retrieval target, and specific information to be specified an appearing location of it; a feature quantity holding unit which, when storing the feature quantity in a feature quantity table, adds new identification information to the feature quantity and holds the feature quantity in the feature quantity table when a similar feature quantity in which a similarity with the feature quantity is no less than a threshold is not included in the feature quantity table, and outputs identification information of the similar feature quantity as identification information of the feature quantity when the similar feature quantity is included in the feature quantity table; and a retrieval table holding unit which holds the specific information associated with the added identification information or the outputted identification information.

公开(公告)号：US20170308578A1

公开(公告)日：2017-10-26

申请号：US15507317

申请日：2014-09-09

Applicant: NEC CORPORATION

Inventor： Ting CHEN ,  Takuya ARAKI

IPC: G06F17/30

CPC classification number: G06F16/2456 ,  G06F16/284

Abstract: One-to-one join is widely used in machine learning and business intelligent applications. Disclosed herein is an efficient method for one-to-one join to reduce memory usage and thus disk I/O accesses with limited memory. Disclosed method outputs and removes a pair of tuples immediately when they are matched to each other to allow join results to be generated without reading entire tables. Meanwhile, disclosed method increases matching rate for in memory blocks through the predication of data distribution patterns based on both statistics and history block matching information.