CN109446293A - A kind of parallel higher-dimension nearest Neighbor - Google Patents
A kind of parallel higher-dimension nearest Neighbor Download PDFInfo
- Publication number
- CN109446293A CN109446293A CN201811345660.8A CN201811345660A CN109446293A CN 109446293 A CN109446293 A CN 109446293A CN 201811345660 A CN201811345660 A CN 201811345660A CN 109446293 A CN109446293 A CN 109446293A
- Authority
- CN
- China
- Prior art keywords
- dimension
- tree
- desired positions
- coordinate value
- neighbour
- 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.)
- Granted
Links
Abstract
The invention discloses the querying methods of parallel higher-dimension neighbour a kind of.This method utilizes the sequential access method of B+ tree index, in parallel query environment, among existing desired positions approach application to NN Query.In spatial data, a B is established to each dimension i of all data objects+Tree index BDi-tree establishes the B including all coordinate scores of object with the coordinate value of the sequential access dimension, for all data objects+Tree index PID safeguards a B according to the creation of dimension Different Dynamic for the data object of all sequential access+Tree indexes BVj-tree to obtain the desired positions and its coordinate value of dimension j.B is utilized+Index technology and the NN Query technology based on desired positions threshold value are set, realizes parallel higher-dimension nearest Neighbor, user can retrieve k neighbour's object according to specified query object and given parameter k, and provide best performance.
Description
Technical field
The present invention relates to the indexes and inquiring technology in spatial database, look into more particularly to a kind of parallel higher-dimension neighbour
Inquiry method.
Background technique
Spatial database is the Database Systems of storage and management spatial data.In order to quickly and efficiently access magnanimity sky
Between data, experts and scholars propose a large amount of space index method, wherein there is R tree index, R* tree index, K-D-B tree to index, with
And the B+ tree index of single dimension.On this basis, various inquiries and its solution with their own characteristics, such as neighbour are more proposed
Inquiry, k NN Query, Continuous Nearest Neighbors Inquiry, Skyline inquiry etc..
With the fast development of computer, communication, internet and location technology, a large amount of data are in scientific algorithm, society
It can live and be constantly be generated with fields such as industrial productions.Based on these data, we can construct various complicated and multiplicity intelligence
It can processing system.It is just basic and important at one to go out qualified data according to certain specific binding characteristic quick-searching
The problem of namely database in inquiry problem.
In these inquiries, NN Query is particularly important, it constitutes the basis of various other NN Queries.NN Query
Most simple, most popular processing mode is to be indexed with R-tree to data object, finds neighbour using branch-bound method
Object, or indexed using B+ tree and establish the method for pivot point using subregion to retrieve neighbour's object.In big data of today
Generation, the data volume of processing are usually very big, it is desirable that are realized using parallel neighbour's processing technique.
Parallel method is the key point for improving NN Query efficiency.Common parallel near neighbor method includes: based on index
The parallel side of parallel method, the parallel method based on GPU, the parallel method based on MapReduce and position sensing Hash LSH
Method etc..The existing near neighbor method based on B+ tree realizes that this needs to cluster and turn using the method for cluster projection to the one-dimensional space
Change process of the point into the one-dimensional space in hyperspace.However, not usually being able to satisfy in current demand such pretreated
Journey, and actual efficiency can be by larger impact.
Summary of the invention
The purpose of the present invention is to provide a kind of parallel higher-dimension nearest Neighbors in higher dimensional space.
The step of the technical scheme adopted by the invention to solve the technical problem is as follows:
Step 1) establishes the B of an entitled BDi-tree according to the coordinate value of each dimension i to all data objects+
Tree index resettles the B+ tree index of an entitled PID；
Step 2) concurrently obtains it in the initial access of each dimension according to the coordinate value q [i] of query object q dimension i
Position, i.e., with the coordinate value position on the immediate BDi-tree of q [i] coordinate value.Here access position is according to their distance
Distance determines that its nearlyr Position Number of distance is smaller, and access order is carried out according to access Position Number ascending order；
Step 3) calculates most concurrently from the initial position of BDi-tree according to the far and near accessed node of nodal distance q [i]
Good position bpi, desired positions threshold TbWith k-th current of neighbour's object distance best_kdist；
BVj-tree index on step 4) concurrent maintenance dimension j, the keyword of the index are all for what is currently accessed
Coordinate value of the data object in jth dimension；
Step 5) compares the coordinate value that BVi-tree on BDi-tree and desired positions bpi is indexed on the Pos of current location, from
And desired positions pointer is moved, and determine whether to terminate inquiry；
Step 6) is less than desired positions threshold T until the preferably distance best_kdist of k-th of neighbour's object and qb, knot
Data object in fruit list is exactly the k higher-dimension neighbour obtained parallel.
Two kinds of B+ tree index node keywords and leaf node content are different in the step 1), in two kinds of situation
Consider:
1) for BDi-tree tree, keyword key is the fixed length word that the i-th dimension degree coordinate value of each object is converted
Symbol string, such as: the string that distance is 0.1101 is ' 00001101 ', and the content of leaf node is i-th dimension degree coordinate value and object identity；
2) for PID tree, keyword key is the identification strings of each data object regular length, such as:
' 00001000 ', and leaf node content is the coordinate value and object identity of all dimensions of data object.
The access position of i-th dimension in the step 2) refers to the coordinate value q [i] and data pair of the dimension query object
As the absolute value of the coordinate value o [i] of o | q [i]-o [i] | number corresponding to size, number is according to ascending order accessed node in B+ tree
Be sequentially generated, the nearlyr number of distance is smaller, such as: nearest node location number is 1, then the node location of next access is compiled
Number be 2 ..., and so on.
Desired positions bpi in the step 3), desired positions threshold TbWith k-th current of neighbour's object distance
The meaning of best_kdist is as follows:
1) the desired positions bpi of i-th dimension refers to location column of all data objects accessed in BDi-tree index
The maximum position of continuous position in table；
2) desired positions threshold TbRefer to that (each dimension i is most in the sum of coordinate value of each dimension desired positions bpi
Good position bpi may not be identical)；
3) k-th of neighbour's object distance best_kdist refer to k-th current of neighbour's object at a distance from query object q,
Including two kinds of situations:
A) when neighbour's number of objects of acquisition be less than k when, best_kdist be all acquisitions neighbour's object in distance q most
Big value；
B) when neighbour's number of objects of acquisition is equal to k, best_kdist is the distance between current k-th of neighbour and q.
BVj-tree in the step 4) is created to have accessed the jth dimensional coordinate values of data object by each dimension
B+Tree, keyword are the regular length character string that coordinate value is converted into, and leaf node is coordinate value and object identity.
The BVj-tree index on dimension j is concurrently safeguarded in the step 5), is divided into following five step:
1) BDi-tree next node position is concurrently moved on to, position indicator pointer Pos is allowed to be directed toward current location, obtains node
Coordinate value and data object tag；
2) the concurrently coordinate value of the data object p of the current location more current dimension i and the best position of BVi-tree index
Whether the node coordinate value for the next position set is equal；If differed, by PID indexed search data object p in residue
The coordinate value of other dimensions, then be sequentially inserted into BVj-tree (j ≠ i) index of remaining dimension；
3) desired positions bpi, desired positions threshold T are concurrently updatedb, preferably distance best_kdist and k it is close
Adjacent set of data objects；
4) when preferably distance best_kdist is less than desired positions threshold TbWhen, terminate NN Query；
5) when desired positions in the next node of the current location Pos on the BDi-tree in any dimension and BVi-tree
When the next node of bpi is equal, desired positions are concurrently moved to the next position, update desired positions threshold value, until terminating；
Meanwhile being less than desired positions threshold T in preferably distance best_kdistbWhen, terminate NN Query.
The invention has the advantages that:
The present invention takes full advantage of grinding for existing index technology in database, NN Query technology and desired positions algorithm
Study carefully and Realizing Achievement, provides a kind of method that neighbour is concurrently inquired in higher dimensional space based on B+ tree index, user can
According to dynamic select query object, efficient NN Query is executed.
Detailed description of the invention
Fig. 1 is the implementation steps flow chart of parallel higher-dimension nearest Neighbor provided by the invention；
Fig. 2 is parallel higher-dimension nearest Neighbor operation principle schematic diagram provided by the invention.
Specific embodiment
Below in conjunction with attached drawing, the technical characteristic and advantage above-mentioned and other to the present invention are clearly and completely described,
Obviously, described case study on implementation is only part case study on implementation of the invention, rather than whole case study on implementation.
Technical solution of the present invention is described further now in conjunction with attached drawing and specific implementation.
1, as shown in Figure 1, specific implementation process of the present invention and working principle are as follows:
Step 1) establishes the B of an entitled BDi-tree according to the coordinate value of each dimension i to all data objects+
Tree index resettles the B+ tree index of an entitled PID；
Step 2) concurrently obtains it in the initial access of each dimension according to the coordinate value q [i] of query object q dimension i
Position, i.e., with the coordinate value position on the immediate BDi-tree of q [i] coordinate value.Here access position is according to their distance
Distance determines that its nearlyr Position Number of distance is smaller, and access order is carried out according to access Position Number ascending order；
Step 3) calculates most concurrently from the initial position of BDi-tree according to the far and near accessed node of nodal distance q [i]
Good position bpi, desired positions threshold TbWith k-th current of neighbour's object distance best_kdist；
BVj-tree index on step 4) concurrent maintenance dimension j, the keyword of the index are all for what is currently accessed
Coordinate value of the data object in jth dimension；
Step 5) compares the coordinate value that BVi-tree on BDi-tree and desired positions bpi is indexed on the Pos of current location, from
And desired positions pointer is moved, and determine whether to terminate inquiry；
Step 6) is less than desired positions threshold T until the preferably distance best_kdist of k-th of neighbour's object and qb, knot
Data object in fruit list is exactly the k higher-dimension neighbour obtained parallel.
Need to establish static B+ tree (BDi-tree, PID tree) in the present invention for data object to realize to data object
The traversal of each dimension, and dynamic B+ tree (BVj-tree), as shown in Figure 2.It is each of data object in step 1)
Dimension establishes BDi-tree, and the PID tree of entire data set, both B+ tree index node keywords and leaf node content
It is different:
1) for BDi-tree tree, keyword key is the fixed length word that the i-th dimension degree coordinate value of each object is converted
Symbol string, such as: the string that distance is 0.1101 is ' 00001101 ', and the content of leaf node is i-th dimension degree coordinate value and object identity；
2) for PID tree, keyword key is the identification strings of each data object regular length, such as:
' 00001000 ', and leaf node content is the coordinate value and object identity of all dimensions of data object.
The access position of i-th dimension in step 2) refers to the coordinate value q's [i] and data object o of the dimension query object
The absolute value of coordinate value o [i] | q [i]-o [i] | number corresponding to size, number according in B+ tree ascending order accessed node it is suitable
Sequence generates, and the nearlyr number of distance is smaller, such as: nearest node location number is 1, then the node location number of next access is
2 ..., and so on.
Desired positions bpi in step 3), desired positions threshold TbWith k-th current of neighbour's object distance best_
The meaning of kdist is as follows:
1) the desired positions bpi of i-th dimension refers to location column of all data objects accessed in BDi-tree index
The maximum position of continuous position in table；
2) desired positions threshold TbRefer to that (each dimension i is most in the sum of coordinate value of each dimension desired positions bpi
Good position bpi may not be identical)；
3) k-th of neighbour's object distance best_kdist refer to k-th current of neighbour's object at a distance from query object q,
Including two kinds of situations:
A) when neighbour's number of objects of acquisition be less than k when, best_kdist be all acquisitions neighbour's object in distance q most
Big value；
B) when neighbour's number of objects of acquisition is equal to k, best_kdist is the distance between current k-th of neighbour and q.
BVj-tree in step 4) is has been accessed the B that the jth dimensional coordinate values of data object are created by each dimension+
Tree, keyword are the regular length character string that coordinate value is converted into, and leaf node is coordinate value and object identity.
The BVj-tree index on dimension j is concurrently safeguarded in step 5), such as the BVj-tree index maintenance module of Fig. 2,
It is divided into following five step:
1) BDi-tree next node position is concurrently moved on to, position indicator pointer Pos is allowed to be directed toward current location, obtains node
Coordinate value and data object tag；
2) the concurrently coordinate value of the data object p of the current location more current dimension i and the best position of BVi-tree index
Whether the node coordinate value for the next position set is equal.If differed, by PID indexed search data object p in residue
The coordinate value of other dimensions, then be sequentially inserted into BVj-tree (j ≠ i) index of remaining dimension；
3) desired positions bpi, desired positions threshold T are concurrently updatedb, preferably distance best_kdist and k it is close
Adjacent set of data objects；
4) when preferably distance best_kdist is less than desired positions threshold TbWhen, terminate NN Query；
5) when desired positions in the next node of the current location Pos on the BDi-tree in any dimension and BVi-tree
When the next node of bpi is equal, desired positions are concurrently moved to the next position, update desired positions threshold value, until terminating.
Meanwhile being less than desired positions threshold T in preferably distance best_kdistbWhen, terminate NN Query.
Particular embodiments described above has carried out further the purpose of the present invention, technical scheme and beneficial effects
It is described in detail, it should be understood that the above is only a specific embodiment of the present invention, the protection being not intended to limit the present invention
Range.It particularly points out, to those skilled in the art, all within the spirits and principles of the present invention, that is done any repairs
Change, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.
Claims (6)
1. a kind of parallel higher-dimension nearest Neighbor, it is characterised in that: the step of this method is as follows:
Step 1) establishes the B of an entitled BDi-tree according to the coordinate value of each dimension i to all data objects+Set rope
Draw, resettles the B+ tree index of an entitled PID；
Step 2) concurrently obtains it in the initial access position of each dimension according to the coordinate value q [i] of query object q dimension i
Set, i.e., with the coordinate value position on the immediate BDi-tree of q [i] coordinate value；Here access position is remote according to their distance
Close to determine, its nearlyr Position Number of distance is smaller, and access order is carried out according to access Position Number ascending order；
Step 3) calculates best position concurrently from the initial position of BDi-tree according to the far and near accessed node of nodal distance q [i]
Set bpi, desired positions threshold TbWith k-th current of neighbour's object distance best_kdist；
BVj-tree index on step 4) concurrent maintenance dimension j, the keyword of the index are all data currently accessed
Coordinate value of the object in jth dimension；
Step 5) compares the coordinate value that BVi-tree on BDi-tree and desired positions bpi is indexed on the Pos of current location, to move
Dynamic desired positions pointer, and determine whether to terminate inquiry；
Step 6) is less than desired positions threshold T until the preferably distance best_kdist of k-th of neighbour's object and qb, as a result arrange
Data object in table is exactly the k higher-dimension neighbour obtained parallel.
2. the parallel higher-dimension nearest Neighbor of one kind according to claim 1, it is characterised in that: in the step 1)
Two kinds of B+ tree index node keywords and leaf node content are different, and are considered in two kinds of situation:
1) for BDi-tree tree, keyword key is the regular length character that the i-th dimension degree coordinate value of each object is converted
String, such as: the string that distance is 0.1101 is ' 00001101 ', and the content of leaf node is i-th dimension degree coordinate value and object identity；
2) for PID tree, keyword key is the identification strings of each data object regular length, such as: ' 00001000 ',
And leaf node content is the coordinate value and object identity of all dimensions of data object.
3. the parallel higher-dimension nearest Neighbor of one kind according to claim 1, it is characterised in that: in the step 2)
I-th dimension access position refer to the dimension query object coordinate value q [i] and data object o coordinate value o [i] it is absolute
Value | q [i]-o [i] | number corresponding to size, number is sequentially generated according to ascending order accessed node in B+ tree, apart from nearlyr volume
It is number smaller.
4. the parallel higher-dimension nearest Neighbor of one kind according to claim 1, it is characterised in that: in the step 3)
Desired positions bpi, desired positions threshold TbIt is as follows with the meaning of k-th current of neighbour's object distance best_kdist:
1) the desired positions bpi of i-th dimension refers to all data objects accessed in the list of locations in BDi-tree index
The maximum position of continuous position；
2) desired positions threshold TbRefer to the sum of the coordinate value in each dimension desired positions bpi；
3) k-th of neighbour's object distance best_kdist refer to k-th current of neighbour's object at a distance from query object q, including
Two kinds of situations:
A) when neighbour's number of objects of acquisition is less than k, best_kdist is the maximum of distance q in neighbour's object of all acquisitions
Value；
B) when neighbour's number of objects of acquisition is equal to k, best_kdist is the distance between current k-th of neighbour and q.
5. the parallel higher-dimension nearest Neighbor of one kind according to claim 1, it is characterised in that: in the step 4)
BVj-tree to have accessed the B that the jth dimensional coordinate values of data object are created by each dimension+Tree, keyword are coordinate
The regular length character string that value is converted into, leaf node are coordinate value and object identity.
6. the parallel higher-dimension nearest Neighbor of one kind according to claim 1, it is characterised in that: in the step 5)
It concurrently safeguards the BVj-tree index on dimension j, is divided into following five step:
1) BDi-tree next node position is concurrently moved on to, allows position indicator pointer Pos to be directed toward current location, obtains the seat of node
Scale value and data object tag；
2) the concurrently coordinate value of the data object p of the current location more current dimension i and BVi-tree index desired positions
Whether the node coordinate value of next position is equal；If differed, by PID indexed search data object p remaining other
The coordinate value of dimension, then be sequentially inserted into BVj-tree (j ≠ i) index of remaining dimension；
3) desired positions bpi, desired positions threshold T are concurrently updatedb, preferably distance best_kdist and k neighbour's number
According to object set；
4) when preferably distance best_kdist is less than desired positions threshold TbWhen, terminate NN Query；
5) as desired positions bpi in the next node of the current location Pos on the BDi-tree in any dimension and BVi-tree
Next node it is equal when, concurrently move desired positions to the next position, update desired positions threshold value, until terminate；Together
When, it is less than desired positions threshold T in preferably distance best_kdistbWhen, terminate NN Query.
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 true CN109446293A (en) | 2019-03-08 |
| CN109446293B 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) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111966678A (en) * | 2020-07-06 | 2020-11-20 | 复旦大学 | Optimization method for effectively improving B + tree retrieval efficiency on GPU |
| 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 |
| CN113010525A (en) * | 2021-04-01 | 2021-06-22 | 东北大学 | Ocean space-time big data parallel KNN query processing method based on PID |
Citations (11)
| 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 |
| US20140090023A1 (en) * | 2012-09-27 | 2014-03-27 | Hong Kong Baptist University | Method and Apparatus for Authenticating Location-based Services without Compromising Location Privacy |
| CN106126571A (en) * | 2016-06-20 | 2016-11-16 | 山东理工大学 | The increment type k nearest Neighbor of n dimension point set |
| 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 |
| US20170161271A1 (en) * | 2015-12-04 | 2017-06-08 | Intel Corporation | Hybrid nearest neighbor search tree with hashing table |
| 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 |
| CN107169372A (en) * | 2017-05-10 | 2017-09-15 | 东南大学 | Privacy protection enquiring method based on Voronoi polygons Yu Hilbert curve encodings |
| 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 |
| 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 |
| 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 |
-
2018
- 2018-11-13 CN CN201811345660.8A patent/CN109446293B/en active Active
Patent Citations (11)
| 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 |
| US20140090023A1 (en) * | 2012-09-27 | 2014-03-27 | Hong Kong Baptist University | Method and Apparatus for Authenticating Location-based Services without Compromising Location Privacy |
| US20170161271A1 (en) * | 2015-12-04 | 2017-06-08 | Intel Corporation | Hybrid nearest neighbor search tree with hashing table |
| CN106126571A (en) * | 2016-06-20 | 2016-11-16 | 山东理工大学 | The increment type k nearest Neighbor of n dimension point set |
| 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 |
| CN107169372A (en) * | 2017-05-10 | 2017-09-15 | 东南大学 | Privacy protection enquiring method based on Voronoi polygons Yu Hilbert curve encodings |
| 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 |
| 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 |
| 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 |
|---|
| QI LIU 等: "GB-Tree: An efficient LBS location data indexing method", 《2014 THE THIRD INTERNATIONAL CONFERENCE ON AGRO-GEOINFORMATICS》 * |
| 宋竹 等: "微观交通仿真系统的近邻查询算法", 《计算机应用》 * |
Cited By (4)
| 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 |
| CN113010525A (en) * | 2021-04-01 | 2021-06-22 | 东北大学 | Ocean space-time big data parallel KNN query processing method based on PID |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109446293B (en) | 2021-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kraska et al. | The case for learned index structures | |
| Beckmann et al. | A revised R*-tree in comparison with related index structures | |
| CN110134714B (en) | Distributed computing framework cache index method suitable for big data iterative computation | |
| US9141666B2 (en) | Incremental maintenance of range-partitioned statistics for query optimization | |
| CN101404032B (en) | Video retrieval method and system based on contents | |
| KR20160145785A (en) | Flash optimized columnar data layout and data access algorithms for big data query engines | |
| CN107341178B (en) | Data retrieval method based on self-adaptive binary quantization Hash coding | |
| CN109166615B (en) | Medical CT image storage and retrieval method based on random forest hash | |
| CN104391908B (en) | Multiple key indexing means based on local sensitivity Hash on a kind of figure | |
| CN106874425B (en) | Storm-based real-time keyword approximate search algorithm | |
| CN106777343A (en) | increment distributed index system and method | |
| CN109446293A (en) | A kind of parallel higher-dimension nearest Neighbor | |
| Li et al. | Skyline index for time series data | |
| US11294816B2 (en) | Evaluating SQL expressions on dictionary encoded vectors | |
| CN108388666A (en) | A kind of database multi-list Connection inquiring optimization method based on glowworm swarm algorithm | |
| CN104899326A (en) | Image retrieval method based on binary multi-index Hash technology | |
| CN103020054A (en) | Fuzzy query method and system | |
| CN105740428B (en) | A kind of higher-dimension disk index structure and image search method based on B+ tree | |
| CN104933143A (en) | Method and device for acquiring recommended object | |
| Eghbali et al. | Online nearest neighbor search using hamming weight trees | |
| CN110334290B (en) | MF-Octree-based spatio-temporal data rapid retrieval method | |
| CN115794873A (en) | Authority control method based on full-text retrieval technology | |
| CN113792709B (en) | Rapid large-scale face recognition method and system | |
| Huang et al. | Pisa: An index for aggregating big time series data | |
| Yu et al. | Materialized view selection based on adaptive genetic algorithm and its implementation with Apache hive |
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 |