CN109858545A - A kind of local core point clustering algorithm based on parallel neighbour naturally - Google Patents
A kind of local core point clustering algorithm based on parallel neighbour naturally Download PDFInfo
- Publication number
- CN109858545A CN109858545A CN201910080473.XA CN201910080473A CN109858545A CN 109858545 A CN109858545 A CN 109858545A CN 201910080473 A CN201910080473 A CN 201910080473A CN 109858545 A CN109858545 A CN 109858545A
- Authority
- CN
- China
- Prior art keywords
- core point
- local core
- point
- local
- 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.)
- Pending
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention discloses a kind of local core point clustering algorithm based on parallel neighbour naturally, including the following steps: the building of KD- tree S1: is carried out to data set using quick sort；S2: using parallel natural adjacent searching algorithm, the neighborhood information of each data object is obtained；S3: by calculating the density of each data object to obtain local core point；S4: the distance between local core point is calculated；S5: construction decision diagram realizes the cluster of local core point；S6: distributing non local core point, realizes the cluster of data set.The distance between the local core point based on shared nearest neighbor is defined through the invention, improves the efficiency of clustering algorithm.
Description
Technical field
It is the present invention relates to data mining technology field, in particular to a kind of based on parallel adjacent local core point cluster naturally
Algorithm.
Background technique
Clustering is that a kind of important method of data mining makes the purpose is to which data object to be divided into different classes
The object obtained in same class cluster is similar to each other, and the object in inhomogeneity cluster is different from each other.Clustering is by widely
Applied to fields such as big data, pattern-recognition, image processing and artificial intelligences.Therefore there is weight to the research of cluster algorithm
The meaning wanted.Existing clustering algorithm can substantially be divided into the method based on division, the method based on density, the side based on level
Method, the method based on grid and method based on model etc..
In recent years, the clustering algorithm based on center is increasingly becoming the hot spot of research.Method based on division, such as K-means
Algorithm and K-medoids algorithm etc., also referred to as based on the clustering algorithm at center.But K-means algorithm and K-medoids algorithm
Cluster result be easy to be influenced by initial cluster center, the cluster knot that different initial cluster centers may be differed greatly
Fruit.In order to avoid the selection at initial cluster center, Frey and Dueck propose AP algorithm, the algorithm by all data objects all
As potential cluster center, optimal cluster centre is then found by the information transmitting between data object.But AP algorithm
Cluster result be easy influenced by preference parameter, in order to solve this problem, K-AP algorithm is suggested.It is by drawing
Enter a constraint condition, is clustered using the intermediate result of K cluster.But AP algorithm and K-AP algorithm can not all identify aspheric
The cluster of shape.One adapts to and composes the AP clustering algorithm DAAP algorithm that dimension about subtracts based on density and is suggested, and asks for solving this
Topic, but due to needing to calculate the shortest path between all objects, time complexity with higher.
Rodriguez and Laio proposed a kind of clustering algorithm for quickly searching density peak DP in 2014 on " Science ".The calculation
Method thinks that there are cluster centre the neighbours compared with low-density to surround, and they with there is the distance between more highdensity object
It is relatively large.Based on this thought, data object is mapped to about density and δ distance by construction decision diagram by DP algorithm
In two-dimensional space (decision diagram), the cluster centre with greater density and δ distance is projected, to be readily available cluster
Center.DP algorithm does not need that constantly optimization object function is optimal to obtain as K-medoids algorithm as K-means algorithm
Cluster centre, but it can not identify complicated manifold cluster.
Summary of the invention
It can not be suitable for the deficiency of complicated manifold data set for existing DP (Density Peaks) algorithm, the present invention mentions
It is a kind of based on parallel naturally adjacent local core point clustering algorithm out, using local domain information redefine local core point it
Between distance, can preferably be used for 3D point cloud data skeletal extraction.
To achieve the goals above, the present invention the following technical schemes are provided:
A kind of local core point clustering algorithm based on parallel neighbour naturally, which is characterized in that including the following steps:
S1: the building of KD- tree is carried out to data set using quick sort；
S2: using parallel natural adjacent searching algorithm, the neighborhood information of each data object is obtained；
S3: by calculating the density of each data object to obtain local core point；
S4: the distance of the shared nearest neighbor between local core point is calculated；
S5: construction decision diagram realizes the cluster of local core point；
S6: distributing non local core point, realizes the cluster of data set.
Preferably, quick sort described in the S1 is to be looked into using the position and needs of the nominal value after quicksort
The median location looked for is compared, to judge that median point is located at the left side or the right of nominal value, then recursively goes to inquire, directly
Until the position where the position of nominal value is median point.
Preferably, in the S2, the parallel natural adjacent searching algorithm is using collector node and two kinds of search node
The calculate node of type carries out parallel computation, the neighborhood information of each data object for obtaining data set to KD- tree.
Preferably, the collector node, for judging whether nature neighbour's searching algorithm algorithm stops, the sample point of inverse neighbour
Number is 0, then algorithm stops；Described search node, for searching for data neighborhood information in KD- tree and sending data neighborhood letter
Breath.
Preferably, in the S3, comprising the following steps:
S3-1: the density Den of each data object in data set, calculation formula are calculated are as follows:
In formula (1), the density of Den (p) expression data object p, inverse neighbour's quantity of nb (p) expression data object p, p,
q∈D,NNλ(p) be p λ nearest-neighbors set, q is the data object in the λ nearest-neighbors set of p, and dist (p, q) is p and q
Euclidean distance；
S3-2: choosing the corresponding data object of MaxDen (p) value in data object local neighborhood is local core point.
Preferably, in the S4, comprising the following steps:
S4-1: local core neighborhood of a point is defined:
In formula (2), NLORE (p) is the neighborhood of local core point p, and MLORE (p) is the collection of the member of local core point p
It closes, NNλ(q) be local core point q λ nearest-neighbors set；
S4-2: the shared nearest neighbor of two local core points is defined:
SLORE (p, q)=NLORE (p) ∩ NLORE (q) (3)
In formula (3), SLORE (p, q) indicates that the shared nearest neighbor of local core point p and q, NLORE (p) are local core point
The neighborhood of p, NLORE (q) are the neighborhood of local core point q；
S4-3: the shared nearest neighbor distance between two local core points is calculated:
In formula (4), SD (p, q) is the shared nearest neighbor distance of local core point p and q, and d (p, q) is two local cores
Euclidean distance between point p, q, Den (o) are the density of data object o, maxd be between the core point of any two part away from
From maximum value；| SLORE (p, q) | indicate the shared nearest neighbor number of local core point p and q.
Preferably, in the S5, comprising the following steps:
S5-1: the density p of local core point is obtained:
The density of local core point p is denoted as ρ (p)=Den (p)；
S5-2: the δ distance of local core point p is obtained:
In formula (5), δ (p) indicates the δ distance of local core point p, and LORE indicates local core point set, SD (p, q) table
Show the distance of the shared nearest neighbor of local core point p and q, SD (p, o) is the distance of the shared nearest neighbor of local core point p and o, max ρ
Indicate the density maxima of local core point；
S5-3: constructing two-dimentional decision diagram, and local core point is selected to be clustered as cluster centre；
It is vertical seat by abscissa, δ distance of ρ according to the density p and δ distance of core point local in local core point set
It is marked on local core point and constructs two-dimentional decision diagram, select ρ > α and the local core point of δ distance > β is formed for cluster centre
Cluster, α, β are preset threshold, and then assigning to remaining local core click and sweep with its distance is belonging to the local core point of minSD
Cluster in, to complete the cluster of local core point.
Preferably, in the S6, the non local core point is divided into cluster belonging to its corresponding local core point
In, to carry out the cluster of all data objects in data set.
In conclusion by adopting the above-described technical solution, compared with prior art, the present invention at least has beneficial below
Effect:
The present invention is using the parallel naturally adjacent Distributive Characters for obtaining data set, to obtain the local core of data set
Point, and the distance between the local core point based on shared nearest neighbor is defined again.And parallel naturally adjacent searching algorithm, Jin Jinkao
Consider the cluster on local core point, therefore improves the efficiency of algorithm；The distance between local core point redefined is abundant
The local neighborhood information of data object is utilized, enables new algorithm effectively to handle complicated manifold cluster, so as to answer
For in 3D point cloud data skeletal extraction.
Detailed description of the invention:
Fig. 1 is according to a kind of based on parallel adjacent local core point clustering algorithm stream naturally of exemplary embodiment of the present
Journey schematic diagram.
Fig. 2 is the quick sorting algorithm flow diagram according to exemplary embodiment of the present.
Fig. 3 is the parallel natural adjacent searching algorithm schematic diagram according to exemplary embodiment of the present.
Specific embodiment
Below with reference to embodiment and specific embodiment, the present invention is described in further detail.But this should not be understood
It is all that this is belonged to based on the technology that the content of present invention is realized for the scope of the above subject matter of the present invention is limited to the following embodiments
The range of invention.
Fig. 1 is according to a kind of based on parallel adjacent local core point clustering algorithm stream naturally of exemplary embodiment of the present
Journey schematic diagram.Specifically includes the following steps:
Step S1: the building of KD- tree is carried out to data set using quick sort.
In the present embodiment, data set to be clustered be D, the present invention using quick sort to the data in data set D into
Row sequence is to search the intermediate value of data set, to construct the KD- tree (k-Dimensional tree) of data set.
Referring to Fig. 2, particularly using the position Location of the nominal value mark after quicksort with require to look up
Median location median is carried out, and the numerical value on the nominal value left side is both less than nominal value, and the numerical value on the right of nominal value is both greater than nominal value,
To judge that median point should be in the Left or right of nominal value, then recursively go to inquire, until the position of nominal value
Until the position where median point.Since the value in the median point left side is both less than intermediate value, the value on the right of intermediate value is all big
In intermediate value, therefore the result can satisfy the demand of KD tree foundation, and not need to sort to overall data, therefore greatly reduce
Establish the time overhead of KD tree.Such as after first time quicksort nominal value mark position Location at the 4th point V1,
And median location median is in the 7th point, then second quicksort should since the 4th point V1, and so on, until nominal
The position Location of value mark is overlapped at V2 with the location of the position median of intermediate value.
Step S2: using parallel natural adjacent searching algorithm, the neighborhood information of each data object is obtained.
In the present embodiment, data set to be clustered is D.Be defined first to data lumped parameter: p, q ∈ D search for number
According to the k nearest neighbor (k successively takes 1,2,3 ..., n) for concentrating each object, when data set be 0 against neighbour's number data object
When constant, referred to as natural stability state；When data set D reaches natural stability state, the k value adaptively obtained is known as nature spy
Value indicative λ, then for p and q, if the λ neighbour (λ nearest-neighbors) of p contains q, and the λ neighbour of q contains p, then p and
Q natural neighbours each other；P is the k arest neighbors of q, then q is the inverse neighbour of p.
Referring to Fig. 3, in the present embodiment, the present invention is calculated using parallel natural adjacent searching algorithm, i.e., using collection
Node and the two kinds of calculate node of search node carry out parallel computation, for obtain data set each data object from
So neighbour's information, such as the inverse neighbour number nb (p) of naturally adjacent eigenvalue λ and any data object p.For example, institute in data set D
There is data object to be built into KD- tree, and n search node is set, it is real that the data object in KD- tree is divided into n region
It now corresponds, i.e. the 1st search node manages the 1st region, and the 2nd search node manages the 2nd region, the n-th search node management n-th
Region, and all search nodes are managed by collector node.
Each search node is used in corresponding range searching data neighborhood information and sends data neighborhood information.Search for number
It is the search thread executed always according to neighborhood information, the thread is continual to carry out the data object being responsible in region
Search inquires kth neighbour, then by neighbour's information preservation into the corresponding chained list of each Searching point.Send data neighborhood information
It is a communication thread, each search node is received respectively by the Fin and threshold k of collector node transmission；Stop if Fin=1
It searches for and exits；If Fin=0, look into whether the k value that each search node judgement searches is greater than threshold k, as k < K, table
Show that search thread has searched for the k nearest neighbor of data object in corresponding region not yet, at this time should search thread continue to execute；
If k > K, the k nearest neighbor that search data object is completed, the then neighbour of each data object obtained search node are indicated
Information is sent to collector node.
Collector node is as main node, there are two its task is main: first is that the process of naturally adjacent searching algorithm into
Row monitoring, by the neighborhood information of each data object currently saved, judges whether algorithm should terminate, i.e., no inverse neighbour
Data object number (nb (p)=0) do not change, then algorithm terminate；It is responsible for second is that collecting all search nodes
The kth neighbor information of data object in region, and these information are summarized, count the inverse neighbours of each data object with
Inverse neighbours' quantity.
S3: by calculating the density of each data object to obtain local core point.
In the present embodiment, define local neighborhood: to a data object p, its local neighborhood be its nb (p)-neighbour (i.e.
Neighbour's neighborhood of p), it is denoted as LN (p)=NNnb(p)(p)。
S3-1: the density Den of each data object in data set is calculated, calculation formula such as following formula:
In formula (1), the density of Den (p) expression data object p, inverse neighbour's quantity of nb (p) expression data object p, p,
q∈D,NNλ(p) be p λ nearest-neighbors set, q is the data object in the λ nearest-neighbors set of p, and dist (p, q) is p and q
Euclidean distance.
S3-2: select in the local neighborhood of data object p the corresponding data object of MaxDen (p) value as data object p
Representative point, be denoted as Rep (p), select Rep (p) to represent data object a little as local core point, local core point set is denoted as
LORE, i.e. LORE={ p ∈ D | Rep (p)=p } represent point further according to represent that delivery rules update each data object as phase
The set of the local core point answered, the member of local core point p is denoted as MLORE (p), then MLORE (p)=q ∈ D | Rep (q)=
P } delivery rules are represented described in refers to if Rep (p)=q and Rep (q)=r, Rep (p)=r.
S4: the distance between local core point is calculated.
S4-1: local core neighborhood of a point is defined.
In the present embodiment, note NLORE (p) is the neighborhood of local core point p, is defined as follows:
S4-2: the shared nearest neighbor of two local core points is defined.
Remember that SLORE (p, q) is the shared nearest neighbor of local core point p and q.It is defined as follows:
SLORE (p, q)=NLORE (p) ∩ NLORE (q) (3)
S4-3: the shared nearest neighbor distance between two local core points is calculated.
Remember that SD (p, q) is the shared nearest neighbor distance of local core point p and q.It is defined as follows:
(4) in, d (p, q) is the Euclidean distance between two local core point p, q, and Den (o) is the close of data object o
Degree, maxd is the maximum value of the distance between any two part core point.
S5: construction decision diagram realizes the cluster of local core point.
S5-1: obtaining the density of local core point, and the density of each part core point p is denoted as ρ (p), calculates as follows: ρ
(p)=Den (p).
S5-2: the δ distance of each local core point is obtained.
In the present embodiment, such as the δ distance of local core point p is denoted as δ (p), calculates as follows:
In formula (5), δ (p) indicates the δ distance of local core point p, and LORE indicates local core point set, SD (p, q) table
Show the distance of the shared nearest neighbor of local core point p and q, SD (p, o) is the distance of the shared nearest neighbor of local core point p and o, max ρ
Indicate the density maxima of local core point；
S5-3: according to the density p and δ distance of core point local in local core point set, it is by abscissa, δ distance of ρ
Ordinate constructs two-dimentional decision diagram on local core point, and selecting the local core point of ρ > α and δ distance > β is cluster centre shape
Cluster, α, β are preset threshold；And part remaining in local core point set core click and sweep is assigned into more high density and is with it
In cluster belonging to the local core point of minSD distance (shared nearest neighbor distance), to realize the cluster of local core point.Such as this
In embodiment, part core point p is selected to form the first cluster as cluster centre, local core point q is as cluster centre formation the
Two clusters, local core point o cannot function as cluster centre and SD (p, o) < SD (q, o), then part core point o is divided into the first cluster.
S6: distributing non local core point, realizes the cluster of data set data object.
The maximum data object of density is local core point, remaining data object in the local neighborhood of each data object
For non local core point.
In the present embodiment, it is corresponding that core point non local in the local neighborhood of each data object is directly divided into its
In cluster belonging to local core point, to obtain the cluster of all data objects in data set D.For example, the part of data object p
The local core point of neighborhood is Rep (p), and Rep (p) is cluster centre and is divided into the first cluster, then by the office of data object p
Non local core point is divided into the first cluster where local core point Rep (p) entirely in portion's neighborhood；And so on, to realize number
According to the cluster of all data objects in collection D.
Claims (8)
1. a kind of based on parallel adjacent local core point clustering algorithm naturally, which is characterized in that including the following steps:
S1: the building of KD- tree is carried out to data set using quick sort；
S2: using parallel natural adjacent searching algorithm, the neighborhood information of each data object is obtained；
S3: by calculating the density of each data object to obtain local core point；
S4: the distance of the shared nearest neighbor between local core point is calculated；
S5: construction decision diagram realizes the cluster of local core point；
S6: distributing non local core point, realizes the cluster of all data objects in data set.
2. as described in claim 1 a kind of based on parallel adjacent local core point clustering algorithm naturally, which is characterized in that described
Quick sort described in S1 is to be compared using the position of the nominal value after quicksort with the median location required to look up
Compared with to judge that median point is located at the left side or the right of nominal value, then recursively going to inquire, in the position of nominal value is
Until position where value point.
3. as described in claim 1 a kind of based on parallel adjacent local core point clustering algorithm naturally, which is characterized in that described
In S2, the parallel natural adjacent searching algorithm is using collector node and the two kinds of calculate node of search node to KD-
Tree carries out parallel computation, the neighborhood information of each data object for obtaining data set.
4. as claimed in claim 3 a kind of based on parallel adjacent local core point clustering algorithm naturally, which is characterized in that described
Collector node, for judging whether nature neighbour's searching algorithm algorithm stops, the sample point number of inverse neighbour is 0, then algorithm stops；
Described search node, for searching for data neighborhood information in KD- tree and sending data neighborhood information.
5. as described in claim 1 a kind of based on parallel adjacent local core point clustering algorithm naturally, which is characterized in that described
In S3, comprising the following steps:
S3-1: the density Den of each data object in data set, calculation formula are calculated are as follows:
In formula (1), Den (p) indicates the density of data object p, and nb (p) indicates inverse neighbour's quantity of data object p, p, q ∈
D,NNλ(p) be p λ nearest-neighbors set, q is the data object in the λ nearest-neighbors set of p, and dist (p, q) is p and q
Euclidean distance；
S3-2: choosing the corresponding data object of MaxDen value in data object local neighborhood is local core point.
6. as described in claim 1 a kind of based on parallel adjacent local core point clustering algorithm naturally, which is characterized in that described
In S4, comprising the following steps:
S4-1: local core neighborhood of a point is defined:
In formula (2), NLORE (p) indicates that the neighborhood of local core point p, MLORE (p) indicate the collection of the member of local core point p
It closes, NNλ(q) be local core point q λ nearest-neighbors set；
S4-2: the shared nearest neighbor of two local core points is defined:
SLORE (p, q)=NLORE (p) ∩ NLORE (q) (3)
In formula (3), SLORE (p, q) indicates that the shared nearest neighbor of local core point p and q, NLORE (p) are local core point p
Neighborhood, NLORE (q) are the neighborhood of local core point q；
S4-3: the shared nearest neighbor distance between two local core points is calculated:
In formula (4), SD (p, q) is the shared nearest neighbor distance of local core point p and q, and d (p, q) is two local core point p, q
Between Euclidean distance, Den (o) is the density of data object o, maxd be the distance between any two part core point most
Big value；| SLORE (p, q) | indicate the shared nearest neighbor number of local core point p and q.
7. as described in claim 1 a kind of based on parallel adjacent local core point clustering algorithm naturally, which is characterized in that described
In S5, comprising the following steps:
S5-1: the density p of local core point is obtained:
The density of local core point p is denoted as ρ (p)=Den (p)；
S5-2: the δ distance of local core point is obtained:
In formula (5), δ (p) indicates the δ distance of local core point p, and LORE indicates local core point set, SD (p, q) expression office
The distance of the shared nearest neighbor of portion core point p and q, SD (p, o) indicate the distance of the shared nearest neighbor of local core point p and o, max ρ table
Show the density maxima of local core point；
S5-3: two-dimentional decision diagram is constructed, the local core point for meeting condition is selected to be clustered as cluster centre；
It is that ordinate exists by abscissa, δ distance of ρ according to the density p and δ distance of core point local in local core point set
Two-dimentional decision diagram is constructed on local core point, and the local core point of ρ > α and δ distance > β is selected to form cluster, α, β for cluster centre
For preset threshold, remaining local core click and sweep is then assigned into more high density and the local core with its distance for minSD value
In cluster belonging to point, to complete the cluster of local core point.
8. as described in claim 1 a kind of based on parallel adjacent local core point clustering algorithm naturally, which is characterized in that described
In S6, the non local core point is divided into cluster belonging to its corresponding local core point, realizes the cluster of data set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910080473.XA CN109858545A (en) | 2019-01-28 | 2019-01-28 | A kind of local core point clustering algorithm based on parallel neighbour naturally |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910080473.XA CN109858545A (en) | 2019-01-28 | 2019-01-28 | A kind of local core point clustering algorithm based on parallel neighbour naturally |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109858545A true CN109858545A (en) | 2019-06-07 |
Family
ID=66896482
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910080473.XA Pending CN109858545A (en) | 2019-01-28 | 2019-01-28 | A kind of local core point clustering algorithm based on parallel neighbour naturally |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109858545A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111291276A (en) * | 2020-01-13 | 2020-06-16 | 武汉大学 | Clustering method based on local direction centrality measurement |
-
2019
- 2019-01-28 CN CN201910080473.XA patent/CN109858545A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111291276A (en) * | 2020-01-13 | 2020-06-16 | 武汉大学 | Clustering method based on local direction centrality measurement |
| CN111291276B (en) * | 2020-01-13 | 2023-05-19 | 武汉大学 | Clustering method based on local direction centrality measurement |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Özyer et al. | Integrating multi-objective genetic algorithm based clustering and data partitioning for skyline computation | |
| CN107992887A (en) | Classifier generation method, sorting technique, device, electronic equipment and storage medium | |
| CN106131862B (en) | Optimization covering method based on multi-objective Evolutionary Algorithm in a kind of wireless sensor network | |
| CN110334391A (en) | A kind of various dimensions constraint wind power plant collection electric line automatic planning | |
| CN110070121A (en) | A kind of quick approximate k nearest neighbor method based on tree strategy with balance K mean cluster | |
| CN106326637A (en) | Link predicting method based on local effective path degree | |
| CN106095920B (en) | Distributed index method towards extensive High dimensional space data | |
| CN108268603A (en) | A kind of community discovery method based on core member's identification | |
| CN109710774B (en) | Graph data partitioning and distributed storage method combining balance strategy | |
| CN105574541A (en) | Compactness sorting based network community discovery method | |
| CN103888541A (en) | Method and system for discovering cells fused with topology potential and spectral clustering | |
| CN102156756A (en) | Method for finding optimal path in road network based on graph embedding | |
| CN108573274A (en) | A kind of selective clustering ensemble method based on data stability | |
| CN109948695A (en) | A kind of power grid fragility node automatic identifying method based on neighbour's propagation clustering algorithm | |
| CN107545272A (en) | A kind of spatial network clustering objects method under MapReduce frameworks | |
| CN104731811B (en) | A kind of clustering information evolution analysis method towards extensive dynamic short text | |
| CN104361135A (en) | Image search method | |
| CN109066674A (en) | Reconstruction method of power distribution network based on the full Smoothing Newton Method of Constraints | |
| CN111724273A (en) | Method for automatically planning current collection circuit of offshore wind farm by adopting large-capacity wind turbine generator | |
| CN109858545A (en) | A kind of local core point clustering algorithm based on parallel neighbour naturally | |
| CN105138607B (en) | A kind of KNN querying methods based on combination grain distributed memory grid index | |
| CN103164487B (en) | A kind of data clustering method based on density and geological information | |
| CN108959652A (en) | Overlapping community discovery method based on density peaks with community's degree of membership | |
| Luo et al. | A survey of nearest-better clustering in swarm and evolutionary computation | |
| CN109033746A (en) | A kind of protein complex recognizing method based on knot vector |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190607 |
|
| RJ01 | Rejection of invention patent application after publication |