CN109446293B - Parallel high-dimensional neighbor query method - Google Patents
Parallel high-dimensional neighbor query method Download PDFInfo
- Publication number
- CN109446293B CN109446293B CN201811345660.8A CN201811345660A CN109446293B CN 109446293 B CN109446293 B CN 109446293B CN 201811345660 A CN201811345660 A CN 201811345660A CN 109446293 B CN109446293 B CN 109446293B
- Authority
- CN
- China
- Prior art keywords
- tree
- neighbor
- dimension
- query
- distance
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a parallel high-dimensional neighbor query method. The method utilizes a sequential access method of B + tree index, and applies the existing best position method to the neighbor query in a parallel query environment. In spatial data, a B is established for each dimension i of all data objects+The tree index BDi-tree sequentially accesses the coordinate values of the dimension and establishes a B including all coordinate scores of the objects for all data objects+A tree index PID for dynamically creating and maintaining a B for all sequentially accessed data objects according to different dimensions+The tree indexes BVj-tree to obtain the best location of the dimension j and its coordinate value. Utilize B+The tree index technology and the neighbor query technology based on the best position threshold value realize a parallel high-dimensional neighbor query method, and a user can retrieve k neighbor objects according to a specified query object and a given parameter k and provide the best performance.
Description
Technical Field
The invention relates to an indexing and query technology in a spatial database, in particular to a parallel high-dimensional neighbor query method.
Background
A spatial database is a database system that stores and manages spatial data. In order to access massive spatial data quickly and effectively, expert scholars propose a large number of spatial indexing methods, including R tree indexes, K-D-B tree indexes, and single-dimensional B + tree indexes. On the basis, various characteristic queries and solutions thereof are provided, such as neighbor query, k neighbor query, continuous neighbor query, Skyline query and the like.
With the rapid development of computer, communication, internet and positioning technologies, a large amount of data is continuously generated in the fields of scientific computing, social life, industrial production and the like. Based on these data, we can build a variety of complex and diverse intelligent processing systems. The fast retrieval of qualified data based on a specific constraint characteristic becomes a fundamental and important problem, namely the query problem in the database.
Of these queries, neighbor queries are particularly important, and form the basis for various other neighbor queries. The simplest and most popular processing method of neighbor query is to index a data object with an R-tree, find a neighbor object with a branch boundary method, or retrieve a neighbor object with a B + tree index and a method of creating pivot points with partitions. In today's big data era, the amount of data processed is often very large, requiring implementation with parallel neighbor processing techniques.
The parallel method is a key point for improving the neighbor query efficiency. Common parallel neighbor methods include: an index-based parallel method, a GPU-based parallel method, a MapReduce-based parallel method, a location-sensitive Hash LSH-based parallel method, and the like. The existing neighbor method based on the B + tree is realized by adopting a method of projecting clustering to a one-dimensional space, which needs the processes of clustering and converting points in a multi-dimensional space to the one-dimensional space. However, such a pretreatment process is often not satisfied in real-world demand, and the actual efficiency is greatly affected.
Disclosure of Invention
The invention aims to provide a parallel high-dimensional neighbor query method used in a high-dimensional space.
The technical scheme adopted for solving the technical problems comprises the following steps:
step 1) establishing a B named BDi-tree for all data objects according to the coordinate value of each dimension i+And (4) tree indexing, namely establishing a B + tree index named PID. The key of BDi-tree is the ith dimension coordinate value of each object and converts it into character string with fixed length, the content of leaf node is the ith dimension of data objectCoordinate values and object identifications. The key of the PID tree is an identification character string with fixed length corresponding to each data object, and the leaf node content of the PID tree is coordinate values and object identification of all dimensions of the data object;
and 2) obtaining the initial access position of the query object q in each dimension in parallel according to the coordinate value q [ i ] of the dimension i of the query object q, namely the coordinate value position on the BDi-tree closest to the coordinate value q [ i ]. Here, the access positions are determined according to the distance between the access positions, the position numbers of the access positions are smaller as the distance is shorter, and the access sequence is carried out according to the ascending order of the access position numbers;
step 3) parallelly from the initial position of BDi-tree according to the node distance q [ i]The best position bpi and the best position threshold value T are calculatedbThe distance best _ kdist from the current k-th neighbor object;
and 4) parallelly maintaining a B + tree BVj-tree in the dimension j, wherein keywords of the B + tree are character strings converted from coordinate values of all currently accessed data objects in the dimension j, and leaf nodes are the character strings converted from the coordinate values and object identifications. The coordinate value of the j-th dimension of the data object accessed each time is dynamically inserted into BVj-tree, which is used for determining the best position;
step 5) comparing the index BDi-tree on the Pos at the current position with the coordinate value of the BVi-tree on the bpi at the best position, thereby moving the pointer at the best position and determining whether to terminate the query;
step 6) until the best distance best _ kdist between the kth neighbor object and q is less than the best position threshold value TbThe data objects in the result list are k high-dimensional neighbors obtained in parallel.
In the step 1), the contents of the key words and the leaf nodes of the two B + tree index nodes are different, and the two conditions are considered:
1) for a BDi-tree, a keyword key of the BDi-tree is the ith dimension coordinate value of each object and is converted into a character string with a fixed length, and the content of leaf nodes is the ith dimension coordinate value and an object identifier;
2) for the PID tree, the key is an identification character string with fixed length corresponding to each data object, and the leaf node content is coordinate value and object identification of all dimensions of the data object.
The access position of the ith dimension in step 2) refers to a number corresponding to the size of an absolute value | q [ i ] -o [ i ] of the coordinate value q [ i ] of the dimension query object and the coordinate value o [ i ] of the data object o, the number is generated according to the sequence of ascending access nodes in the B + tree, and the closer the distance, the smaller the number is, for example: the nearest node location is numbered 1, the next visited node location is numbered 2, …, and so on.
The best position bpi in the step 3), the best position threshold value TbThe distance best _ kdist from the current k-th neighbor has the following meaning:
1) the best position bpi of the ith dimension refers to the maximum position of continuous positions of all accessed data objects in the position list in the BDi-tree index;
2) best position threshold value TbMeans the sum of the coordinate values of the best position bpi in each dimension (the best position bpi in each dimension i may not be the same);
3) the kth neighbor object distance best _ kdist refers to the distance between the current kth neighbor object and the query object q, and includes two cases:
a) when the number of the obtained neighbor objects is less than k, best _ kdist is the maximum value of the distances q in all the obtained neighbor objects;
b) when the obtained number of the neighbor objects is equal to k, best _ kdist is the distance between the current k-th neighbor and q.
The BVj-tree in the step 4) is B created by the j-th dimension coordinate value of each dimension accessed data object+The key word of the tree is a fixed-length character string converted from coordinate values, and the leaf nodes are the coordinate values and object identifications.
The step 4) of maintaining the BVj-tree index in the dimension j in parallel comprises the following five steps:
1) parallelly moving to the next node position of the BDi-tree, and enabling a position pointer Pos to point to the current position to obtain the coordinate value and the data object identifier of the node;
2) parallelly comparing whether the coordinate value of the data object p at the current position of the current dimension i is equal to the coordinate value of the node at the position next to the best position of the BVi-tree index; if not, searching coordinate values of the data object p in other remaining dimensions through the PID index, and then sequentially inserting the coordinate values into BVj-tree (j ≠ i) indexes of the remaining dimensions;
3) updating the best position bpi, the best position threshold value T in parallelbBest distance best _ kdist, and k sets of neighboring data objects;
4) when the best distance best _ kdist is less than the best position threshold value TbWhen the neighbor query is finished, the neighbor query is terminated;
5) when the next node of the current position Pos on the BDi-tree in any dimension is equal to the next node of the best position bpi on the BVi-tree, moving the best position to the next position in parallel, and updating the best position threshold value until the end; meanwhile, when the best distance best _ kdist is smaller than the best position threshold value TbThen the neighbor query is terminated.
The invention has the beneficial effects that:
the invention fully utilizes the research and realization results of the prior index technology, the neighbor query technology and the best position algorithm in the database, provides a method for querying neighbors in parallel in a high-dimensional space based on the B + tree index, and a user can select a query object according to the dynamic state and execute efficient neighbor query.
Drawings
FIG. 1 is a flow chart of the implementation steps of a parallel high-dimensional neighbor query method provided by the present invention;
fig. 2 is a schematic diagram of the working principle of the parallel high-dimensional neighbor query method provided by the present invention.
Detailed Description
The above and further features and advantages of the present invention will be described more fully hereinafter with reference to the accompanying drawings, in which some, but not all embodiments of the invention are shown.
The technical solution of the present invention will be further explained with reference to the accompanying drawings and specific embodiments.
1. As shown in fig. 1, the specific implementation process and the working principle of the present invention are as follows:
step 1) establishing a B named BDi-tree for all data objects according to the coordinate value of each dimension i+And (4) tree indexing, namely establishing a B + tree index named PID. The keyword key of the BDi-tree converts the ith dimension coordinate value of each object into a character string with fixed length, and the content of the leaf node is the ith dimension coordinate value and the object identification of the data object. The key of the PID tree is an identification character string with fixed length corresponding to each data object, and the leaf node content of the PID tree is coordinate values and object identification of all dimensions of the data object;
and 2) obtaining the initial access position of the query object q in each dimension in parallel according to the coordinate value q [ i ] of the dimension i of the query object q, namely the coordinate value position on the BDi-tree closest to the coordinate value q [ i ]. Here, the access positions are determined according to the distance between the access positions, the position numbers of the access positions are smaller as the distance is shorter, and the access sequence is carried out according to the ascending order of the access position numbers;
step 3) parallelly from the initial position of BDi-tree according to the node distance q [ i]The best position bpi and the best position threshold value T are calculatedbThe distance best _ kdist from the current k-th neighbor object;
and 4) parallelly maintaining a B + tree BVj-tree in the dimension j, wherein keywords of the B + tree are character strings converted from coordinate values of all currently accessed data objects in the dimension j, and leaf nodes are the character strings converted from the coordinate values and object identifications. The coordinate value of the j-th dimension of the data object accessed each time is dynamically inserted into BVj-tree, which is used for determining the best position;
step 5) comparing the index BDi-tree on the Pos at the current position with the coordinate value of the BVi-tree on the bpi at the best position, thereby moving the pointer at the best position and determining whether to terminate the query;
step 6) until the best distance best _ kdist between the kth neighbor object and q is less than the best position threshold value TbThe data objects in the result list are k high-dimensional neighbors obtained in parallel.
In the present invention, a static B + tree (BDi-tree, PID tree) needs to be established for the data object to implement traversal of each dimension of the data object, and a dynamic B + tree (BVj-tree) is also needed, as shown in fig. 2. Establishing a BDi-tree for each dimension of the data object in the step 1), and establishing a PID tree of the whole data set, wherein the key word and leaf node contents of the index nodes of the two B + trees are different:
1) for the BDi-tree, the key word is the ith dimension coordinate value of each object and is converted into a character string with fixed length, and the content of the leaf node is the ith dimension coordinate value and the object identifier.
2) For the PID tree, the key is the identification character string with fixed length corresponding to each data object, and the leaf node content is the coordinate value and object identification of all dimensions of the data object.
The access position of the ith dimension in step 2) refers to a number corresponding to the size of an absolute value | q [ i ] -o [ i ] of the coordinate value q [ i ] of the dimension query object and the coordinate value o [ i ] of the data object o, the number is generated according to the sequence of ascending access nodes in the B + tree, and the closer the distance, the smaller the number is, such as: the nearest node location is numbered 1, the next visited node location is numbered 2, …, and so on.
Best position bpi in step 3), best position threshold TbThe distance best _ kdist from the current k-th neighbor has the following meaning:
1) the best position bpi of the ith dimension refers to the maximum position of continuous positions of all accessed data objects in the position list in the BDi-tree index;
2) best position threshold value TbMeans the sum of the coordinate values of the best position bpi in each dimension (the best position bpi in each dimension i may not be the same);
3) the kth neighbor object distance best _ kdist refers to the distance between the current kth neighbor object and the query object q, and includes two cases:
a) when the number of the obtained neighbor objects is less than k, best _ kdist is the maximum value of the distances q in all the obtained neighbor objects;
b) when the obtained number of the neighbor objects is equal to k, best _ kdist is the distance between the current k-th neighbor and q.
The BVj-tree in the step 4) is visited by each dimensionB created by j-th dimension coordinate value of query data object+The key word of the tree is a fixed-length character string converted from coordinate values, and the leaf nodes are the coordinate values and object identifications.
Maintaining BVj-tree indexes on the dimension j in parallel in the step 4), such as a BVj-tree index maintenance module of fig. 2, which comprises the following five steps:
1) parallelly moving to the next node position of the BDi-tree, and enabling a position pointer Pos to point to the current position to obtain the coordinate value and the data object identifier of the node;
2) and parallelly comparing whether the coordinate value of the data object p at the current position of the current dimension i is equal to the coordinate value of the node at the position next to the best position of the BVi-tree index. If not, searching coordinate values of the data object p in other remaining dimensions through the PID index, and then sequentially inserting the coordinate values into BVj-tree (j ≠ i) indexes of the remaining dimensions;
3) updating the best position bpi, the best position threshold value T in parallelbBest distance best _ kdist, and k sets of neighboring data objects;
4) when the best distance best _ kdist is less than the best position threshold value TbWhen the neighbor query is finished, the neighbor query is terminated;
5) when the next node for the current position Pos on the BDi-tree in any dimension is equal to the next node for the best position bpi on the BVi-tree, the best position is moved to the next position and the best position threshold is updated in parallel until the end. Meanwhile, when the best distance best _ kdist is smaller than the best position threshold value TbThen the neighbor query is terminated.
The above-mentioned embodiments are provided to further explain the objects, technical solutions and advantages of the present invention in detail, and it should be understood that the above-mentioned embodiments are only examples of the present invention and are not intended to limit the scope of the present invention. It should be understood that any modifications, equivalents, improvements and the like, which come within the spirit and principle of the invention, may occur to those skilled in the art and are intended to be included within the scope of the invention.
Claims (4)
1. A parallel high-dimensional neighbor query method is characterized in that: the method comprises the following steps:
step 1) establishing a B named BDi-tree for all data objects according to the coordinate value of each dimension i+The method comprises the steps of tree indexing, establishing a B + tree index named PID, converting an ith dimension coordinate value of each object into a character string with fixed length by using a key of a BDi-tree, wherein the content of a leaf node of the BDi-tree is the ith dimension coordinate value and an object identifier of each data object, the key of the PID tree is an identifier character string with fixed length corresponding to each data object, and the content of the leaf node of the BDi-tree is the coordinate value and the object identifier of all dimensions of the data object;
step 2) according to the coordinate value q [ i ] of the dimension i of the query object q, obtaining the initial access position of the query object in each dimension in parallel, namely the coordinate value position on the BDi-tree closest to the coordinate value q [ i ]; here, the access positions are determined according to the distance between the access positions, the position numbers of the access positions are smaller as the distance is shorter, and the access sequence is carried out according to the ascending order of the access position numbers;
step 3) parallelly from the initial position of BDi-tree according to the node distance q [ i]The best position bpi and the best position threshold value T are calculatedbThe distance best _ kdist from the current k-th neighbor object;
step 4) maintaining a B + tree BVj-tree in the dimension j in parallel, wherein keywords of the B + tree are character strings converted from coordinate values of all currently accessed data objects in the j dimension, leaf nodes are the character strings converted from the coordinate values and object identifiers, the coordinate values of the data objects accessed each time in the j dimension are dynamically inserted into the BVj-tree, and the key values are used for determining the best position;
step 5) comparing the index BDi-tree on the Pos at the current position with the coordinate value of the BVi-tree on the bpi at the best position, thereby moving the pointer at the best position and determining whether to terminate the query;
step 6) until the best distance best _ kdist between the kth neighbor object and q is less than the best position threshold value TbThe data objects in the result list are k high-dimensional neighbors obtained in parallel.
2. The parallel high-dimensional neighbor query method according to claim 1, wherein: the access position of the ith dimension in the step 2) refers to a number corresponding to the absolute value | q [ i ] -o [ i ] of the coordinate value q [ i ] of the dimension query object and the coordinate value o [ i ] of the data object o, the number is generated according to the sequence of ascending access nodes in the B + tree, and the number is smaller when the distance is shorter.
3. The parallel high-dimensional neighbor query method according to claim 1, wherein: the best position bpi in the step 3), the best position threshold value TbThe distance best _ kdist from the current k-th neighbor has the following meaning:
1) the best position bpi of the ith dimension refers to the maximum position of continuous positions of all accessed data objects in the position list in the BDi-tree index;
2) best position threshold value TbMeans the sum of the coordinate values of the best position bpi in each dimension;
3) the kth neighbor object distance best _ kdist refers to the distance between the current kth neighbor object and the query object q, and includes two cases:
a) when the number of the obtained neighbor objects is less than k, best _ kdist is the maximum value of the distances q in all the obtained neighbor objects;
b) when the obtained number of the neighbor objects is equal to k, best _ kdist is the distance between the current k-th neighbor and q.
4. The parallel high-dimensional neighbor query method according to claim 1, wherein: the step 4) of maintaining the BVj-tree index in the dimension j in parallel comprises the following five steps:
1) parallelly moving to the next node position of the BDi-tree, and enabling a position pointer Pos to point to the current position to obtain the coordinate value and the data object identifier of the node;
2) parallelly comparing whether the coordinate value of the data object p at the current position of the current dimension i is equal to the coordinate value of the node at the position next to the best position of the BVi-tree index; if not, searching coordinate values of the data object p in other remaining dimensions through the PID index, and then sequentially inserting the coordinate values into BVj-tree (j ≠ i) indexes of the remaining dimensions;
3) updating the best position bpi, the best position threshold value T in parallelbBest distance best _ kdist, and k sets of neighboring data objects;
4) when the best distance best _ kdist is less than the best position threshold value TbWhen the neighbor query is finished, the neighbor query is terminated;
5) when the next node of the current position Pos on the BDi-tree in any dimension is equal to the next node of the best position bpi on the BVi-tree, moving the best position to the next position in parallel, and updating the best position threshold value until the end; meanwhile, when the best distance best _ kdist is smaller than the best position threshold value TbThen the neighbor query is terminated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811345660.8A CN109446293B (en) | 2018-11-13 | 2018-11-13 | Parallel high-dimensional neighbor query method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811345660.8A CN109446293B (en) | 2018-11-13 | 2018-11-13 | Parallel high-dimensional neighbor query method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109446293A CN109446293A (en) | 2019-03-08 |
| CN109446293B true CN109446293B (en) | 2021-12-10 |
Family
ID=65552212
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811345660.8A Active CN109446293B (en) | 2018-11-13 | 2018-11-13 | Parallel high-dimensional neighbor query method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109446293B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112860734A (en) * | 2019-11-27 | 2021-05-28 | 中国石油天然气集团有限公司 | Seismic data multi-dimensional range query method and device |
| CN111966678A (en) * | 2020-07-06 | 2020-11-20 | 复旦大学 | Optimization method for effectively improving B + tree retrieval efficiency on GPU |
| CN113010525B (en) * | 2021-04-01 | 2023-08-01 | 东北大学 | Ocean space-time big data parallel KNN query processing method based on PID |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1920822A (en) * | 2006-09-14 | 2007-02-28 | 浙江大学 | Interactive calligraphic character K approaching search method |
| CN106446227A (en) * | 2016-09-30 | 2017-02-22 | 南京航空航天大学 | Skyline checking processing mechanism for multi-preference ordered route with weighted Voronoi diagram index |
| CN107145796A (en) * | 2017-04-24 | 2017-09-08 | 公安海警学院 | Track data k anonymities method for secret protection under a kind of uncertain environment |
| CN108225333A (en) * | 2018-01-12 | 2018-06-29 | 中国电子科技集团公司第二十八研究所 | A kind of optimal path generation method for flight course planning |
| CN108415954A (en) * | 2018-02-06 | 2018-08-17 | 南京信息工程大学 | The uncertain monochromatic mutually K-NN search processing method of one kind |
| CN108763481A (en) * | 2018-05-29 | 2018-11-06 | 清华大学深圳研究生院 | A kind of picture geographic positioning and system based on extensive streetscape data |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9043927B2 (en) * | 2012-09-27 | 2015-05-26 | Neo Mechanic Limited | Method and apparatus for authenticating location-based services without compromising location privacy |
| US10210180B2 (en) * | 2015-12-04 | 2019-02-19 | Intel Corporation | Hybrid nearest neighbor search tree with hashing table |
| CN106126571B (en) * | 2016-06-20 | 2019-07-12 | 山东理工大学 | The increment type k nearest Neighbor of surface sampled data in kind |
| CN107169372B (en) * | 2017-05-10 | 2020-04-14 | 东南大学 | Privacy protection query method based on Voronoi polygon and Hilbert curve coding |
| CN108763292A (en) * | 2018-04-17 | 2018-11-06 | 上海交通大学 | Flexible polymer K-NN search A-sum methods on road network |
-
2018
- 2018-11-13 CN CN201811345660.8A patent/CN109446293B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1920822A (en) * | 2006-09-14 | 2007-02-28 | 浙江大学 | Interactive calligraphic character K approaching search method |
| CN106446227A (en) * | 2016-09-30 | 2017-02-22 | 南京航空航天大学 | Skyline checking processing mechanism for multi-preference ordered route with weighted Voronoi diagram index |
| CN107145796A (en) * | 2017-04-24 | 2017-09-08 | 公安海警学院 | Track data k anonymities method for secret protection under a kind of uncertain environment |
| CN108225333A (en) * | 2018-01-12 | 2018-06-29 | 中国电子科技集团公司第二十八研究所 | A kind of optimal path generation method for flight course planning |
| CN108415954A (en) * | 2018-02-06 | 2018-08-17 | 南京信息工程大学 | The uncertain monochromatic mutually K-NN search processing method of one kind |
| CN108763481A (en) * | 2018-05-29 | 2018-11-06 | 清华大学深圳研究生院 | A kind of picture geographic positioning and system based on extensive streetscape data |
Non-Patent Citations (2)
| Title |
|---|
| GB-Tree: An efficient LBS location data indexing method;Qi Liu 等;《2014 The Third International Conference on Agro-Geoinformatics》;20140814;1-5 * |
| 微观交通仿真***的近邻查询算法;宋竹 等;《计算机应用》;20150210;第35卷(第2期);572-577 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109446293A (en) | 2019-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | LISA: A learned index structure for spatial data | |
| Yao et al. | Computing trajectory similarity in linear time: A generic seed-guided neural metric learning approach | |
| CN110134714B (en) | Distributed computing framework cache index method suitable for big data iterative computation | |
| CN109446293B (en) | Parallel high-dimensional neighbor query method | |
| CN108846029B (en) | Information correlation analysis method based on knowledge graph | |
| CN104462582B (en) | A kind of web data similarity detection method based on structure and content secondary filtration | |
| CN109166615B (en) | Medical CT image storage and retrieval method based on random forest hash | |
| CN106874425B (en) | Storm-based real-time keyword approximate search algorithm | |
| CN104391908B (en) | Multiple key indexing means based on local sensitivity Hash on a kind of figure | |
| CN106897374B (en) | Personalized recommendation method based on track big data nearest neighbor query | |
| US20160364468A1 (en) | Database index for constructing large scale data level of details | |
| CN110334290B (en) | MF-Octree-based spatio-temporal data rapid retrieval method | |
| CN106777343A (en) | increment distributed index system and method | |
| Eghbali et al. | Online nearest neighbor search using hamming weight trees | |
| CN116244333A (en) | Database query performance prediction method and system based on cost factor calibration | |
| CN114153821A (en) | Electric quantity graph database construction and search method based on graph theory | |
| Wan et al. | Cd-tree: A clustering-based dynamic indexing and retrieval approach | |
| CN115794873A (en) | Authority control method based on full-text retrieval technology | |
| CN114417074A (en) | Rapid KNN (K nearest neighbor) retrieval method and system for high-dimensional measurement spatial data | |
| Yu et al. | Materialized view selection based on adaptive genetic algorithm and its implementation with Apache hive | |
| Rechkalov et al. | An approach to data mining inside PostgreSQL based on parallel implementation of UDFs | |
| Huang et al. | ACR-Tree: Constructing R-Trees Using Deep Reinforcement Learning | |
| CN112148830A (en) | Semantic data storage and retrieval method and device based on maximum area grid | |
| Chen et al. | DisITQ: A distributed iterative quantization hashing learning algorithm | |
| Han et al. | Dynamic skyline computation on massive data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |