CN106100921A - The dynamic streaming figure parallel samples method synchronized based on dot information - Google Patents
The dynamic streaming figure parallel samples method synchronized based on dot information Download PDFInfo
- Publication number
- CN106100921A CN106100921A CN201610406969.8A CN201610406969A CN106100921A CN 106100921 A CN106100921 A CN 106100921A CN 201610406969 A CN201610406969 A CN 201610406969A CN 106100921 A CN106100921 A CN 106100921A
- Authority
- CN
- China
- Prior art keywords
- point
- limit
- sampling
- sampler
- subgraph
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/02—Capturing of monitoring data
- H04L43/022—Capturing of monitoring data by sampling
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/12—Discovery or management of network topologies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/142—Network analysis or design using statistical or mathematical methods
Abstract
The present invention provides a kind of dynamic streaming figure parallel samples method synchronized based on dot information, particularly as follows: S1. streaming limit arrives sliding window, it is judged that window is the fullest, if discontented, perform S1, otherwise performs S2；S2. the limit in sliding window is sequentially randomly assigned to multiple sampler；The most samplers obtain subgraph of sampling to allocated limit parallel processing；S4. read the point set of sampler, remove the point repeated, refresh global point set；S5. the limit collection that global point is derived is updated；S6. sampling target point set quantity is adjusted；If not arriving the collection sampling subgraph time point set, updating sliding window, returning S1；Otherwise perform S8；S8. sampling subgraph is synthesized according to the sampling results of each sampler.The present invention is while quickly obtaining sampling subgraph, it is ensured that sampling subgraph is high with the characteristic similarity of artwork, solves the problem that dynamic streaming figure serial sampling algorithm processes time length, can not meet requirement of real-time.
Description
Technical field
The invention belongs to streaming diagram data sampling techniques field, more particularly, to a kind of based on moving that dot information synchronizes
State streaming figure parallel samples method.
Background technology
Figure has had become as the one ubiquitous and requisite data knot expressing object and relation in real world
Structure, various application include that social networks, bioscience network, WWW etc. can be modeled as figure.But, in reality
A lot of application programs constantly change, and the graph structure of modeling also can correspondingly change, the most over time
Passage, it may occur that point, limit increase operation or deletion action.The information occurrence dynamics over time ground on point and limit changes
Figure is exactly so-called Dynamic Graph.The Perfected process processing Dynamic Graph data is to use streaming mode, i.e. artwork is regarded as by one
The point of consecutive streaming arrival and limit gradually form.
Dynamic Graph is sampled using streaming sampling algorithm, existing part derivation limit sampling PIES
(Partially-Induced Edge Sampling, is shown in paper: " Space-efficient sampling from social
Activity streams ") algorithm, due to the accumulative iterative characteristic of Stream Processing, whole sampling process convection type limit is successively
Carrying out processing, namely serial, then well imagine, sampling process needs to consume the substantial amounts of time.Real-time is wanted
Seeking the scene that comparison is high, control network congestion time strictly according to the facts, the sampling algorithm of these serials can not meet far away requirement, improves sampling
Algorithm execution speed is highly desirable to, so needing more quickly sampling algorithm.
Summary of the invention
For defect and the urgent needs of prior art, the invention provides a kind of dynamic streaming synchronized based on dot information
Figure parallel samples method, it is intended that while quickly obtaining sampling subgraph, it is ensured that the subgraph characteristic phase with artwork of sampling
Seemingly spend height, solve the dynamic streaming figure serial sampling algorithm process time long, it is impossible to meet asking of the high application of requirement of real-time
Topic.
A kind of dynamic streaming figure parallel samples method synchronized based on dot information, comprises the following steps:
S1. (u, v) arrives sliding window, it is judged that sliding window is the fullest, if discontented, performs S1 streaming limit e=, no
Then perform S2；
S2. the limit in sliding window is sequentially randomly assigned to p sampler, until being assigned；
S3.p sampler, to allocated limit parallel processing, obtains Gsk=(Vsk, Esk,Edk)；Kth sampler is smoked
Sampling subgraph is Gsk=(Vsk,Esk,Edk), 1≤k≤p, wherein, Vsk={ vsk1,vsk2,…,vSkn,It is the some set of subgraph,
vski, 1≤i≤n ' is the point in sampling subgraph, and the size of sampling point set | Vsk|=n '=n/p, wherein n is sampling impact point
Quantity；Esk={ esk1,esk2,…,eskmBe subgraph limit set, e thereinski, 1≤i≤m is the limit in sampling subgraph；Edk
={ edk1,edk2,…,edktBe subgraph derivation limit set, wherein edki, 1≤i≤t is the foundation global point in sampling subgraph
Collection VsglobalThe derivation limit obtained；
S4. global point synchronizing information: be successively read the point set V of each samplersk, remove the point repeated, refresh global point
Set VSglobal；
S5. the limit collection that global point is derived is updated: each sampler utilizes global point information V after updatingsglobal, for1≤k≤p, if the end points of e is not at VsglobalIn, then delete e；
S6. sampling target point set quantity is adjusted: when | Vsglobal| < n, then increase the sampling impact point of sampler equably
Quantity n；If | Vsglobal| > n, then reduce sampling impact point quantity n equably；
If not arriving the collection sampling subgraph time point set, increasing according to sampling impact point quantity n and then increasing slip
Window, impact point quantity n reduces the principle renewal sliding window size then reducing sliding window, returns step S1；Otherwise perform
Step S8；
S8. sampling subgraph is synthesized according to the sampling results of each sampler: sampling sub-chart is shown as: Gsperiod=
(Vsperiod,Esperiod), wherein, Vsperiod=Vs1∪Vs2∪…∪VspFor the union of all sampler Point Sets, Esperiod=
Es1∪Es2∪…∪Esp∪Ed1∪Ed2∪…∪EdpFor limit collection in all samplers and the union of overall situation derivation limit collection；
S9. terminate.
Further, in described step S3 each sampler to carry out the step of parallel processing as follows:
A) (u, v) arrives certain sampler to streaming limit e=, and this sampler determines whether to produce to be replaced, if it occur that point
Replace, then perform b), otherwise perform f)；
Decision principle is:
If the some u ∈ V i. in streaming limitsk∪Vsglobal,v∈Vsk∪Vsglobal, V will not be causedskAdd new point, do not occur
Replace；
If the some u ∈ V ii. in streaming limitsk∪Vsglobal,Or v
∈Vsk∪Vsglobal, and the number of existing point | Vsk| < n ' does not replaces；Otherwise, point v or u needs to add V toskIn
And replace an existing point；
If the iii. point in streaming limitThe number of existing point | Vsk
| < n '-1 is not replaced；Otherwise, u and v is required for adding V toskIn and replace two existing points；
B) each sampler is each independent according to the subgraph G that samples in this samplersk, 1≤j≤p, in the degree of point special
Property, determine a replacement probability function fk(di),di∈Dj, wherein DjIt it is the set of the degree of all nodes in jth sampler；According to
This replacement probability function calculates some viThe probability being replacedObtain replacing Making by Probability Sets
WhereinIt is a viDegree, and haveWherein Dk={ dk1,dk2,…,dkn′Be distributed for a degree at set midpoint；
Wherein require to replace probability function fk(di) in action scope [1, dmax] interior monotone decreasing, wherein dmaxFor the highest in degree distributed collection
The number of degrees；
C) each sampler all uses the selection algorithm select (P) in genetic algorithm, and wherein P is calculated in being b)
Point replaces Making by Probability Sets, chooses the some r being replaced；
D) each sampler is according to replacing principle, it is judged that whether some r c) selected meets is replaced requirement, if meeting, turns
To e)；Otherwise go to c)；
Replacement principle is as follows:
The most above-mentioned a) in the case of, when selecting to be replaced, it is impossible to select the point in newly-increased limit, and follow-up
Isolated point can not delete the point in newly-increased limit when deleting；
The most above-mentioned a) in the case of, it is assumed that first add u, then add v；When first adding u, VskIn the most associated with it
Point, replace and do not limit；When adding v again, due to VskMiddle there is coupled some u, the point being replaced can not be for u；
E) each sampler is from VskMiddle deletion r, and from EskAnd EdkThe limit that middle deletion is associated with r；Again from VskMiddle deletion orphan
Vertical point；Isolated point erasure request can not delete first point in newly-increased point；
F) each sampler adds subgraph G newly-increased point and limitskIn；Wherein increase point with the principle on limit to subgraph is: as
Really u, v are all at point set VskIn, then by e=, (u v) adds limit collection EskIn；If in u, v one at point set Vsk, another
Global point set VsglobalIn, then e is added derivation limit collection EdkIn；If u, v are all at global point set VsglobalAnd do not exist
Point set VskIn, do not increase and a little the most do not increase limit in subgraph.
Further, probability function f is replacedk(di) it is inverse proportion function.
Further, selection algorithm select (P) is the ratio selection algorithm in genetic algorithm.
Further, the detailed process of described step S6 adjustment sampling target point set quantity is: when | Vsglobal| < n, impartial
Ground increase sampler the sampling i.e. n=n+ of impact point quantity (n-| Vsglobal|)/p；If | Vsglobal| > n, reduce equably and take out
The sample i.e. n=n-of impact point quantity (| Vsglobal|-n)/p。
Further, update sliding window size be (2n-| Vsglobal|)/2.
Dynamic streaming figure parallel samples algorithm (referred to as PaStS) that the present invention synchronizes based on dot information, is quickly obtaining
While sampling subgraph, it is ensured that sampling subgraph is high with the characteristic similarity of artwork, solves dynamic streaming figure serial sampling algorithm
The process time is long, it is impossible to the problem meeting the high application of requirement of real-time.Compared with prior art PIES algorithm, for
The Dynamic Graph of identical scale, the execution efficiency of parallel algorithm PaStS compares serial PIES can improve p to p2Times, wherein, p is also
The number of line sampling device.
Accompanying drawing explanation
Fig. 1 is the inventive method flow chart；
Fig. 2 processes the schematic flow sheet of single edge for the unitary sampling device that the embodiment of the present invention provides.
Detailed description of the invention
In order to make the purpose of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, right
The present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, and
It is not used in the restriction present invention.If additionally, technical characteristic involved in each embodiment of invention described below
The conflict of not constituting each other just can be mutually combined.
If the sampling subgraph of kth (1≤k≤p) individual sampler is Gsk=(Vsk,Esk,Edk), V thereinsk={ vsk1,
vsk2,…,vskn′It is the some set of subgraph, vski(1≤i≤n ') is the point in sampling subgraph, and the size of sampling point set | Vsk|
=n '=n/p, wherein, n is the number at object sample subgraph midpoint；Esk={ esk1,esk2,…,eskmBe subgraph limit set,
E thereinski(1≤i≤m) is the limit in sampling subgraph；Edk={ edk1,edk2,…,edktBe subgraph limit set, wherein eski
(1≤i≤t) is the foundation overall situation point set V in sampling subgraphsglobalThe derivation limit obtained.
As it is shown in figure 1, the invention provides a kind of dynamic streaming figure parallel samples method synchronized based on dot information, including
Following steps:
S1. streaming limit e=(u, v) arrives sliding window, it is judged that whether sliding window full, if be unsatisfactory for, then performs S1,
Otherwise perform S2；
The size of sliding window is relevant to the sum at object sample subgraph midpoint, and initial value is typically less than object sample
Can arbitrarily arrange in the range of the sum of figure midpoint, adjust the sum approaching object sample subgraph midpoint subsequently through circulation.
In a preferred manner, initial setting up sliding window size is n/2, and sliding window size is by last round of thereafter
The S7 step of sampling determines, n is sampling impact point quantity；
S2. the limit in window is sequentially randomly assigned to p sampler, until being assigned；
Parallel samples is entered in allocated limit by S3.p sampler, obtains Gsk=(Vsk,Esk,Edk)；Kth sampler is taken out
Sampling subgraph be Gsk=(Vsk,Esk,Edk), 1≤k≤p, wherein, Vsk={ vsk1,vsk2,…,vskn′It is the some set of subgraph,
vski, 1≤i≤n ' is the point in sampling subgraph, and the size of sampling point set | Vsk|=n '=n/p, wherein n is sampling impact point
Quantity；Esk={ esk1,esk2,…,eskmBe subgraph limit set, e thereinski, 1≤i≤m is the limit in sampling subgraph；Edk
={ edk1,edk2,…,edktBe subgraph derivation limit set, wherein edki, 1≤i≤t is the foundation global point in sampling subgraph
Collection VsglobalThe derivation limit obtained.Overall situation point set VsglobakRefer to after the time of program self definition all samplers
Sampling subgraph in VskThe point set of (1≤k≤p) carries out collecting the point set obtained afterwards, and derivation limit refers at a sampler k
In, end points V in sampling subgraphskIn, and another end points is at overall situation point set VsglobalIn.
S4. global point synchronizing information: be successively read the point set V of each samplersk(1≤k≤p), removes the point repeated, brush
New global point set Vsglobal；
S5. the limit collection that global point is derived is updated: each sampler utilizes global point information V after updatingsglobal, for1≤k≤p, if the end points of e is not at VsglobalIn, then delete e；
S6. sampling target point set quantity is adjusted: when | Vsglobal| < n, then increase the sampling impact point of sampler equably
Quantity, according to a kind of optimal way, updates n=n+ (n-|Vsglobal|)/p；If | Vsglobal| > n, then reduce equably and take out
Sample impact point quantity, according to a kind of optimal way, update n=n-(| Vsglobal|-n)/p；
If not arriving the collection sampling subgraph time point set, then, update sliding window according to the n after updating big
Little, updating principle is that n increases just appropriateness increase sliding window, and n reduces just appropriateness and reduces sliding window, returns step S1；
Otherwise perform step S8；
According to a kind of optimal way, the size updating sliding window be (2n-| Vsglobal|)/2。
S8. sampling subgraph is synthesized according to the sampling results of each sampler: sampling subgraph is represented by: Gsperiod=
(Vsperiod,Esperiod), wherein, Vsperiod=Vs1∪Vs2∪…∪VspFor the union of all sampler Point Sets, Esperiod=
Es1∪Es2∪…∪Esp∪Ed1∪Ed2∪…∪EdpFor limit collection in all samplers and the union of overall situation derivation limit collection；
S9. terminate.
In described step S3, to carry out the step of parallel processing as follows for each sampler:
A) (u, v) arrives certain sampler to streaming limit e=, and this sampler determines whether to produce a replacement, if can occur
Point is replaced, then perform b), otherwise perform f)；
Decision principle is:
If the some u ∈ V i. in streaming limitsk∪Vsglobal,v∈Vsk∪Vsglobal, V will not be causedskAdd new point, do not occur
Replace；
If the some u ∈ V ii. in streaming limitsk∪Vsglobal,Or v
∈Vsk∪Vsglobal, and the number of existing point | Vsk| < n ' does not replaces；Otherwise, point v or u needs to add V toskIn
And replace an existing point；
If the iii. point in streaming limitThe number of existing point | Vsk
| < n '-1 is not replaced；Otherwise, u and v is required for adding V toskIn and replace two existing points；
B) each sampler is each independent according to the subgraph G that samples in this samplerskThe degree of the point in (1≤j≤p) is special
Property, determine a replacement probability function fk(di),di∈Dj；Point v is calculated according to this probability functioniThe probability being replaced
Obtain replacing Making by Probability SetsWhereinIt is a viDegree, and haveWherein Dk
={ dk1,dk2,…,dkn′Be distributed for a degree at set midpoint；F to be found a functionk(di) in action scope [1, dmaxDullness in]
Successively decrease, wherein dmaxFor the number of degrees the highest in degree distributed collection；fk(di) linear function, inverse proportion function, logarithmic function can be used
Etc., the present embodiment preferably employs fk(di) it is inverse proportion function；
C) each sampler all uses the selection algorithm select (P) in genetic algorithm, and wherein P is calculated in being b)
Point replaces Making by Probability Sets, chooses some r to be replaced；Selection algorithm select (P) can adoption rate selection algorithm, determine that formula is sampled
Select, round gambles selection etc., and in the present embodiment, optimization algorithm select (P) is the ratio selection algorithm in genetic algorithm；
D) each sampler is according to replacing principle, it is judged that whether some r c) selected meets the requirements, if meeting, goes to e),
Otherwise go to c)；
Replacement principle is:
The most above-mentioned a)) in the case of, when selecting to be replaced, it is impossible to select the point in newly-increased limit, and follow-up
Isolated point delete time can not delete the point in newly-increased limit；
The most above-mentioned a)) in the case of, it is assumed that first add u, then add v；When first adding u, VskIn the most associated with it
Point, so the replacement occurred not does not limits；When adding v again, due to VskMiddle there is coupled some u, so replacing out
Point can not be u；In this case, first newly-increased point does not limit when replacing, and the second newly-increased point can not when replacing
Replace first the most newly-increased point, and first the most newly-increased point can not be deleted when follow-up isolated point is deleted.
E) each sampler is from VskMiddle deletion r, and from EskAnd EdkThe limit that middle deletion is associated with r；Again from VskMiddle deletion is only
Vertical point；Isolated point erasure request can not delete first point in newly-increased point；
F) each sampler adds subgraph G newly-increased point and limitskIn；
Wherein increase point with the principle on limit to subgraph is: if u, v are all at point set VskIn, then by e=, (u v) adds limit
Collection EskIn；If in u, v one at point set VskOne at global point set VsglobalIn, then e is added derivation limit collection Edk
In；If u, v are all at global point set VsglobalAnd not at point set VskIn, do not increase and a little the most do not increase limit in subgraph.
Below as a example by Fig. 2, introduce single edge handling process.Single edge handling process is with i-th (1≤i≤p) individual sampler
As a example by, it is assumed that currently processed limit is ecurrent(u, v), then, the process of the process of sampler i has a following step:
Step one: the target if current sample point number is not up to sampled, i.e. | Vsi|<ni, then directly choose limit
ecurrent(u, v), by the two of this limit end points u, v is added separately to the point set V that samplessiIn, and by ecurrentAdd local sampling
Limit collection EsiIn；
Step 2: if current sample is counted out reaches target of sampling, then will be to ecurrentCarry out trade-off decision.
First to ecurrentCarry out based on local point set VsiLimit derive, if u, v are at VsiIn, then directly choose ecurrent, and will
It joins set EsiIn.Otherwise it is necessary to according to the overall situation point set VsglobalAgain to ecurrentOnce derive；
Step 3: to ecurrentCarry out deriving, if u, v are at global point set V based on overall situation point setsglobalIn, then
Deriving successfully in limit based on global point, directly chooses ecurrent, and add it to another limit collection, i.e. overall situation derivation limit collection Edi
In.Otherwise it is necessary to ecurrentCarry out the selection of certain probability；
Step 4: after probability selection, if ecurrentThe most selected, then directly to abandon ecurrent.If ecurrentSelected
In, then by ecurrentJoin set EsiIn, two end points sampling add point set VsiIn, now produce to I haven't seen you for ages and once keep in
The replacement of point.On temporary some replacement policy, use the distribution replacement policy reciprocal of justice herein.In this strategy, point set
VsiIn the probability that is replaced of point be directly proportional to the degree inverse of this point, so, the point that degree of maintaining is high to a certain extent is not
Can be frequently replaced, be prevented again the over-agglomerate that the point that the deletion end is minimum terrifically causes.Select first with this strategy and want
The point being replaced, from VsiThis point of middle deletion, from EsiAll limits that middle deletion is relevant to this summit.
As it will be easily appreciated by one skilled in the art that and the foregoing is only presently preferred embodiments of the present invention, not in order to
Limit the present invention, all any amendment, equivalent and improvement etc. made within the spirit and principles in the present invention, all should comprise
Within protection scope of the present invention.
Claims (6)
1. the dynamic streaming figure parallel samples method synchronized based on dot information, it is characterised in that comprise the following steps:
S1. (u v) arrives sliding window, it is judged that sliding window is the fullest, if discontented, perform S1, otherwise hold streaming limit e=
Row S2；
S2. the limit in sliding window is sequentially randomly assigned to p sampler, until being assigned；
S3.p sampler, to allocated limit parallel processing, obtains Gsk=(Vsk,Esk,Edk)；The sampling that kth sampler is taken out
Subgraph is Gsk=(Vsk,Esk,Edk), 1≤k≤p, wherein, Vsk={ vsk1,vsk2,…,vskn′It is the some set of subgraph, vski, 1
≤ i≤n ' is the point in sampling subgraph, and the size of sampling point set | Vsk|=n '=n/p, wherein n is sampling impact point quantity；
Esk={ esk1,esk2,…,eskmBe subgraph limit set, e thereinski, 1≤i≤m is the limit in sampling subgraph；Edk=
{edk1,edk2,…,edktBe subgraph derivation limit set, wherein edki, 1≤i≤t is the foundation overall situation point set in sampling subgraph
VsglobalThe derivation limit obtained；
S4. global point synchronizing information: be successively read the point set V of each samplersk, remove the point repeated, refresh global point set
Vsglobal；
S5. the limit collection that global point is derived is updated: each sampler utilizes global point information V after updatingsglobal, for1≤k≤p, if the end points of e is not at VsglobalIn, then delete e；
S6. sampling target point set quantity is adjusted: when | Vsglobal| < n, then increase the sampling impact point quantity of sampler equably
n；If | Vsglobal| > n, then reduce sampling impact point quantity n equably；
If not arriving the collection sampling subgraph time point set, increasing according to sampling impact point quantity n and then increasing sliding window
Mouthful, impact point quantity n reduces the principle renewal sliding window size then reducing sliding window, returns step S1；Otherwise perform step
Rapid S8；
S8. sampling subgraph is synthesized according to the sampling results of each sampler: sampling sub-chart is shown as: Gsperiod=(Vsperiod,
Esperiod), wherein, Vsperiod=Vs1∪Vs2∪…∪VspFor the union of all sampler Point Sets, Esperiod=Es1∪Es2
∪…∪Esp∪Ed1∪Ed2∪…∪EdpFor limit collection in all samplers and the union of overall situation derivation limit collection；
S9. terminate.
The dynamic streaming figure parallel samples method synchronized based on dot information the most according to claim 1, it is characterised in that institute
Stating each sampler in step S3, to carry out the step of parallel processing as follows:
A) streaming limit e=(u, v) arrives certain sampler, and this sampler determines whether to produce to be replaced, if it occur that point is replaced,
Then perform b), otherwise perform f)；
Decision principle is:
If the some u ∈ V i. in streaming limitsk∪Vsglobal,v∈Vsk∪Vsglobal, V will not be causedskAdd new point, do not replace
Change；
If the some u ∈ V ii. in streaming limitsk∪Vsglobal,Orv∈Vsk
∪Vsglobal, and the number of existing point | Vsk| < n ' does not replaces；Otherwise, point v or u needs to add V toskIn and replace
Change an existing point；
If the iii. point in streaming limitThe number of existing point | Vsk|<
N '-1, does not replaces；Otherwise, u and v is required for adding V toskIn and replace two existing points；
B) each sampler is each independent according to the subgraph G that samples in this samplersk, 1≤j≤p, in the degree characteristic of point, really
Fixed point replaces probability function fk(di),di∈Dj, wherein DjIt it is the set of the degree of all nodes in jth sampler；Replace according to this
Change probability function and calculate some viThe probability being replacedObtain replacing Making by Probability Sets
WhereinIt is a viDegree, and haveWherein Dk={ dk1,dk2,…,dkn′Be distributed for a degree at set midpoint；
Wherein require to replace probability function fk(di) in action scope [1, dmax] interior monotone decreasing, wherein dmaxFor the highest in degree distributed collection
The number of degrees；
C) each sampler all uses the selection algorithm select (P) in genetic algorithm, wherein P in being b) calculated point replace
Change Making by Probability Sets, choose the some r being replaced；
D) each sampler is according to replacing principle, it is judged that whether some r c) selected meets is replaced requirement, if meeting, goes to
e)；Otherwise go to c)；
Replacement principle is as follows:
The most above-mentioned a)) in the case of, when selecting to be replaced, it is impossible to select the point in newly-increased limit, and follow-up orphan
The point in newly-increased limit can not be deleted during vertical point deletion；
The most above-mentioned a)) in the case of, it is assumed that first add u, then add v；When first adding u, VskIn the most associated there
Point, replaces and does not limit；When adding v again, due to VskMiddle there is coupled some u, the point being replaced can not be for u；
E) each sampler is from VskMiddle deletion r, and from EskAnd EakThe limit that middle deletion is associated with r；Again from VskMiddle deletion is isolated
Point；Isolated point erasure request can not delete first point in newly-increased point；
F) each sampler adds subgraph G newly-increased point and limitskIn；Wherein increase point with the principle on limit to subgraph is: if u, v
All at point set VskIn, then by e=, (u v) adds limit collection EskIn；If in u, v one at point set Vsk, another is in global point
Set VsglobalIn, then e is added derivation limit collection EdkIn；If u, v are all at global point set VsglobalAnd not at point set Vsk
In, do not increase and a little the most do not increase limit in subgraph.
The dynamic streaming figure parallel samples method synchronized based on dot information the most according to claim 2, it is characterised in that replace
Change probability function fk(di) it is inverse proportion function.
The dynamic streaming figure parallel samples method synchronized based on dot information the most according to claim 2, it is characterised in that choosing
Selecting algorithm select (P) is the ratio selection algorithm in genetic algorithm.
The dynamic streaming figure parallel samples method synchronized based on dot information the most according to claim 2, it is characterised in that institute
The detailed process stating step S6 adjustment sampling target point set quantity is: when | Vsglobal| < n increases the sampling of sampler equably
Impact point quantity i.e. n=n+ (n-| Vsglobal|)/p；If | Vsglobal| > n, reduce sampling impact point quantity i.e. n=n-equably
(|Vsglobal|-n)/p。
The dynamic streaming figure parallel samples method synchronized based on dot information the most according to claim 5, it is characterised in that more
The size of new sliding window be (2n-| Vsglobal|)/2。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610406969.8A CN106100921B (en) | 2016-06-08 | 2016-06-08 | Dynamic flow chart parallel sampling method based on point information synchronization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610406969.8A CN106100921B (en) | 2016-06-08 | 2016-06-08 | Dynamic flow chart parallel sampling method based on point information synchronization |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106100921A true CN106100921A (en) | 2016-11-09 |
| CN106100921B CN106100921B (en) | 2020-06-02 |
Family
ID=57228841
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610406969.8A Active CN106100921B (en) | 2016-06-08 | 2016-06-08 | Dynamic flow chart parallel sampling method based on point information synchronization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106100921B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109753797A (en) * | 2018-12-10 | 2019-05-14 | 中国科学院计算技术研究所 | For the intensive subgraph detection method and system of streaming figure |
| CN110032605A (en) * | 2019-03-26 | 2019-07-19 | 华中科技大学 | In relational network between entity connection relationship feature acquisition methods and system |
| CN110245135A (en) * | 2019-05-05 | 2019-09-17 | 华中科技大学 | A kind of extensive streaming diagram data update method based on NUMA architecture |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105005586A (en) * | 2015-06-24 | 2015-10-28 | 华中科技大学 | Degree feature replacement policy based stream type graph sampling method |
-
2016
- 2016-06-08 CN CN201610406969.8A patent/CN106100921B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105005586A (en) * | 2015-06-24 | 2015-10-28 | 华中科技大学 | Degree feature replacement policy based stream type graph sampling method |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109753797A (en) * | 2018-12-10 | 2019-05-14 | 中国科学院计算技术研究所 | For the intensive subgraph detection method and system of streaming figure |
| CN109753797B (en) * | 2018-12-10 | 2020-11-03 | 中国科学院计算技术研究所 | Dense subgraph detection method and system for stream graph |
| CN110032605A (en) * | 2019-03-26 | 2019-07-19 | 华中科技大学 | In relational network between entity connection relationship feature acquisition methods and system |
| CN110032605B (en) * | 2019-03-26 | 2021-04-06 | 华中科技大学 | Method and system for acquiring connection relation characteristics among users in social network |
| CN110245135A (en) * | 2019-05-05 | 2019-09-17 | 华中科技大学 | A kind of extensive streaming diagram data update method based on NUMA architecture |
| CN110245135B (en) * | 2019-05-05 | 2021-05-18 | 华中科技大学 | Large-scale streaming graph data updating method based on NUMA (non uniform memory access) architecture |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106100921B (en) | 2020-06-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Federated evaluation of on-device personalization | |
| CN107453919B (en) | Complex network node importance evaluation method and system | |
| US9189506B2 (en) | Database index management | |
| CN106100921A (en) | The dynamic streaming figure parallel samples method synchronized based on dot information | |
| US20150234876A1 (en) | Prospective search of objects using k-d forest | |
| US20200105133A1 (en) | Road segment speed prediction method, apparatus, server, medium, and program product | |
| WO2013061680A1 (en) | Information processing device, data access method, and program | |
| JP5427640B2 (en) | Decision tree generation apparatus, decision tree generation method, and program | |
| Lin et al. | Heavy-traffic analysis of mean response time under shortest remaining processing time | |
| WO2021027153A1 (en) | Method and apparatus for constructing traffic flow data analysis model | |
| WO2016206557A1 (en) | Risk identification method and apparatus | |
| US9097433B2 (en) | Apparatus and a method for controlling facility devices, and a non-transitory computer readable medium thereof | |
| CN107133268B (en) | Collaborative filtering method for Web service recommendation | |
| CN105005586A (en) | Degree feature replacement policy based stream type graph sampling method | |
| Zhang et al. | Optimizing video caching at the edge: A hybrid multi-point process approach | |
| CN110162716B (en) | Influence community searching method and system based on community retrieval | |
| CN106533979B (en) | A kind of network update method and device | |
| JP5609584B2 (en) | Shaping apparatus and method | |
| CN107291860B (en) | Seed user determination method | |
| CN105468728A (en) | Cross-section data acquisition method and system | |
| CN112861016A (en) | Data high-concurrency processing method of Feed stream system of healthy social platform | |
| CN111770152A (en) | Edge data management method, medium, edge server and system | |
| EP3207457B1 (en) | Hierarchical caching for online media | |
| CN115270955A (en) | Online data packet classification method based on rule prefix space | |
| CN114882713A (en) | Multi-scene-based signal control method, system, device and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |