Database management method and information processing system
AssigneeKioxia Corporation
InventorsShinichi KANNO, Toru WATABE, Takahiro KURITA
Life of this patent
Estimating the term…
Abstract
According to one embodiment, a database management method manages a plurality of clusters each having a representative vector. When registering a first D-dimensional vector in a vector database, the method identifies a cluster that has a representative vector closest to the first D-dimensional vector among the plurality of clusters, and calculates D differences obtained by subtracting D elements contained in the representative vector of the identified cluster from D elements contained in the first D-dimensional vector, respectively. The method stores an identifier of the identified cluster and the D differences in the vector database, as position information of the first D-dimensional vector.
Claims20
A database management method of managing a vector database, comprising:
- managing a plurality of clusters each having a representative vector, the representative vector indicating a representative point within a D-dimensional vector space and having D dimensions, D being an integer of two or more;
- and
- when registering a first D-dimensional vector in the vector database,
- identifying a cluster that has a representative vector closest to the first D-dimensional vector among the plurality of clusters, as a cluster to which the first D-dimensional vector should belong;
- calculating D differences obtained by subtracting D elements contained in the representative vector of the identified cluster from D elements contained in the first D-dimensional vector, respectively;
- and
- storing an identifier of the identified cluster and the D differences in the vector database, as position information indicating a position of the first D-dimensional vector within the D-dimensional vector space.
A database management method of managing a vector database, comprising:
- for each of R subvector spaces obtained by dividing a D-dimensional vector space, managing a plurality of clusters each having a representative vector, the representative vector indicating a representative point within the subvector space and having D/R dimensions, D being an integer greater than or equal to two, R being an integer greater than or equal to two and less than D;
- and
- when registering a first D-dimensional vector in the vector database,
- dividing the first D-dimensional vector into R subvectors, each of the R subvectors having D/R dimensions;
- for each of the R subvectors, identifying a cluster that has a representative vector closest to the subvector among the plurality of clusters of a subvector space corresponding to the subvector, as a cluster to which the subvector should belong;
- for each of the R subvectors, calculating D/R differences obtained by subtracting D/R elements contained in the representative vector of the identified cluster from D/R elements contained in the subvector, respectively;
- and
- for each of the R subvectors, storing an identifier of the identified cluster and the D/R differences in the vector database, as position information indicating a position of the subvector within the subvector space corresponding to the subvector.
An information processing system configured to manage a vector database, comprising:
- a processor configured to:
- manage a plurality of clusters each having a representative vector, the representative vector indicating a representative point within a D-dimensional vector space and having D dimensions, D being an integer of two or more;
- and
- when registering a first D-dimensional vector in the vector database,
- identify a cluster that has a representative vector closest to the first D-dimensional vector among the plurality of clusters, as a cluster to which the first D-dimensional vector should belong;
- calculate D differences obtained by subtracting D elements contained in the representative vector of the identified cluster from D elements contained in the first D-dimensional vector, respectively;
- and
- store an identifier of the identified cluster and the D differences in the vector database, as position information indicating a position of the first D-dimensional vector within the D-dimensional vector space.
Dependencies are inferred from the claim text.