CN105550704B - Distributed high dimensional data classification method based on mixing common factor analyzer - Google Patents
Distributed high dimensional data classification method based on mixing common factor analyzer Download PDFInfo
- Publication number
- CN105550704B CN105550704B CN201510916426.6A CN201510916426A CN105550704B CN 105550704 B CN105550704 B CN 105550704B CN 201510916426 A CN201510916426 A CN 201510916426A CN 105550704 B CN105550704 B CN 105550704B
- Authority
- CN
- China
- Prior art keywords
- data
- node
- high dimensional
- class
- training
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000012549 training Methods 0.000 claims abstract description 24
- 238000004364 calculation method Methods 0.000 claims abstract description 8
- 238000009792 diffusion process Methods 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000012804 iterative process Methods 0.000 claims description 3
- 239000013598 vector Substances 0.000 claims description 3
- CVXBEEMKQHEXEN-UHFFFAOYSA-N carbaryl Chemical compound C1=CC=C2C(OC(=O)NC)=CC=CC2=C1 CVXBEEMKQHEXEN-UHFFFAOYSA-N 0.000 claims description 2
- 239000004744 fabric Substances 0.000 claims 1
- 230000001149 cognitive effect Effects 0.000 abstract 1
- 230000003993 interaction Effects 0.000 abstract 1
- 150000004667 medium chain fatty acids Chemical class 0.000 description 9
- 238000012545 processing Methods 0.000 description 7
- 230000005540 biological transmission Effects 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 5
- 108090000695 Cytokines Proteins 0.000 description 4
- 102000004127 Cytokines Human genes 0.000 description 4
- 238000004891 communication Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 4
- 241001269238 Data Species 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000000556 factor analysis Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 208000019901 Anxiety disease Diseases 0.000 description 1
- 230000036506 anxiety Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001627 detrimental effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000005236 sound signal Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Complex Calculations (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses the distributed high dimensional data classification method based on mixing common factor analyzer, this method realizes the classification of high dimensional data by the cooperation between nodes different in network in a distributed way.The distribution for modeling its high dimensional data at each node with mixing common factor analyzer is entirely divided into two parts of training and identification.In the training process, then calculate three groups of intermediate variables are carried out broadcast diffusion by the initialization and local calculation for carrying out model parameter first, when node receives the intermediate variable that the broadcast of its neighbor node comes, it calculates joint statistic and completes the estimation of parameter, the continuous iteration of the process is until convergence.In cognitive phase, data to be sorted input any node, the log-likelihood of its model corresponding to the every a kind of data trained are calculated, using the corresponding classification of max log likelihood value as recognition result.The distributed dimensionality reduction of high dimensional data may be implemented using this method, each node in network obtains higher and consistent classification performance, the privacy that interaction intermediate variable can be effectively protected data is only transmitted between this exterior node.
Description
Technical field
The present invention relates to a kind of distributed high dimensional data classification methods based on mixing common factor analyzer, belong to data
The technical field of processing and application.
Background technique
With the continuous development of acquisition and storage technology, the dimension and quantity of data constantly increase, and higher-dimension big data is continuous
It emerges in large numbers.For example, facial image, video and webpage text commonplace in large-scale image retrieval and file retrieval based on content
This, biological tissue is carried out in the high dimensional feature vectors that unavoidably occur in voice and Audio Signal Processing, bioinformatics
Gene data etc. in clustering.It will be apparent that dimension is higher, data volume is bigger, can portray more fully hereinafter described
Object and preferably resolution object.It is born however, excessively high dimension, excessive data volume bring high processing and transmission
Load, especially in sensor network, the storage of individual node, processing and transmission communication capacity are all extremely limited, therefore, logarithm
According to analysis and the design of processing method propose new higher requirement and challenge.Specifically, on the one hand, for high-dimensional
For data or feature, traditional model and and its algorithm for estimating be easy to appear " dimension disaster " problem so that relevant issues are difficult
To understand and indicate, visualization is unlikely realized.Therefore, how to realize and high dimensional data accurately and efficiently analyzed and is handled,
Have become an extremely challenging basic research problems；On the other hand, when data volume is very big, single sensor section
Point is often unable to complete the analysis and processing task of data, and big data can be divided into different parts at this time, be respectively stored in
It is common to complete specified task by reasonably communicating and cooperating on multiple sensor nodes.How for big data design association
Strategy is dealt with, is also a problem to be solved.
Classification, which refers to the process of, is divided into multiple classes for data by certain method, in machine learning field, to data
Classification is the process of a supervised learning.In existing document and patent, the method for having had already appeared a large amount of classification, still
Under data volume is very big or the limited situation of individual node processing capacity, need to distribute data across on multiple nodes, at this time
How distributed treatment is completed, it is very crucial.Therefore, this patent is proposed method designs precisely in order to solve the problems, such as this
It is a kind of based on mixing common factor analyzer distributed high dimensional data classification method (1) hybrid cytokine analysis model can be effective
Processing high dimensional data；(2) it by cooperation mode between design node, only transmits intermediate result and is obtained with satisfied cluster knot
Fruit had not only reduced the expense of communication, but also protect the privacy information in data compared with transmitting initial data mode, it is ensured that
Data safety in network.
Summary of the invention
Present invention aims at the defects for solving the above-mentioned prior art, propose a kind of based on mixing common factor analysis
The distributed high dimensional data classification method of device, this method comprises the following steps:
Step 1: the acquisition of data；
A network is formed equipped with M node, each collected data of node come from V class, data dimension p.Its
In, in the collected all data of node m, the data set from v-th of class is
WhereinIt indicates at node m, the nth data from v-th of class,For the training data number of digital v-th of class；This
Outside, the neighbor node set expression of node m is Rm；
Step 2, training: public for the data mixing in all nodes from v-th of class
Factor minute parser (MCFA) describes its distribution, and completes the training of model using distributed way, estimates parameterIn the same way, it estimates corresponding to every a kind of data
The parameter set Θ of MCFA(v)(v=1 ..., V), training process are completed；
Step 3, identify: when any node in network collects new data x' for identification, calculate x' about
Θ(v)Log-likelihood log p (the x'| Θ of (v=1 ..., V)(v)) (v=1 ..., V):
Using the corresponding serial number of max log likelihood value as the recognition result v' of x':
The process of training of data described in step 2 of the present invention for v-th of class includes the following:
It in order to indicate succinct, and will not influence understanding and implementation, omitWith
In subscript " (v) ").
Step 2-1, initialization: the initial parameter value in setting MCFAWherein, each node
(the w at place1,...,wg,...,wG)=(1/G ..., 1/G ..., 1/G)；Each of L and E matrix element all from standard just
State, which is distributed in N (0,1), to be generated；{ξ1,...,ξg,...,ξGIn each element generate from the standardized normal distribution N (0,1)；
Ω1=...=Ωg=...=ΩG=Iq, wherein IqFor the unit matrix of (q × q).
Step 2-2, broadcast data number: each node l (l=1,2 ..., M) is by its collected data amount check NlExtensively
It broadcasts and gives its neighbor node.After the data amount check that all neighbor nodes broadcast that some node m receives it comes, which is calculated
Weight coefficient clm:
In addition, iteration count iter=1, starts iterative process；
Step 2-3, local calculation: at each node l in sensor network, according to the data X at current nodel
With the parameter value Θ estimated after last iterationold, i.e., (as iter=1, ΘoldFor the parameter value after initialization),
Calculate al,n,g,bg,hgWithIts formula are as follows:
Wherein,
Step 2-4, broadcast diffusion: each node l in sensor network calculate three groups of intermediate variables, it may be assumed that
It is placed in a data packet, so
Other backward node broadcasts spread the data packet.
Step 2-5, combined calculation: when node m (m=1 ..., M) is received from its all neighbor node l (l ∈ Rm) hair
After the data packet containing intermediate variable come, joint statistic is calculatedThat is:
Step 2-6, parameter Estimation: node m (m=1 ..., M) is according to the calculated joint statistic of step 2-5 and step
2-3 is calculated to be estimatedThat is:
Step 2-7, differentiate convergence: node m (m=1 ..., M) calculates the log-likelihood under current iteration, it may be assumed that
Wherein ΘnewIndicate the parameter value that current iteration estimates, ΘoldIndicate the estimation parameter value in last iteration.
If logp (Xm|Θnew)-logp(Xm|Θold) < ε, wherein ε=10-5, node m enters final state.Otherwise, step is turned to
2-3 starts next iteration.
After above-mentioned steps 2-1~step 2-7, Θ is estimated(v)。
The method of the present invention is applied to the parallel and distributed process of data.
The utility model has the advantages that
1. the mixing common factor analyzer that the present invention uses can carry out dimensionality reduction to high dimensional data, thus in the same of dimensionality reduction
When smoothly complete the modelings of data, obtain better classification performance, and reduce computational complexity.In addition, the present invention only passes
Defeated results of intermediate calculations rather than initial data greatly protect the privacy of transmission data.
2. the training and identification process based on mixing common factor analyzer that the present invention uses, so that each in network
Node can make full use of information included in the data of other nodes, so that classification performance is substantially superior to centralization side
Method.
Detailed description of the invention
Fig. 1 is the process of the distributed high dimensional data classification method of the present invention based on mixing common factor analyzer
Figure.
Fig. 2 is the qualitative comparative result schematic diagram of the classification performance of method according to the present invention and other methods.
Fig. 3 is the quantitative comparison result schematic diagram of the classification performance of method according to the present invention and other methods.
Specific embodiment
The invention is described in further detail with reference to the accompanying drawings of the specification.
As shown in Figure 1-3, the present invention provides a kind of distributed high dimensional datas point based on mixing common factor analyzer
Class method, this method comprises the following steps:
Step 1: the acquisition of data；
Equipped with M platform computer/calculate node (that is: node), a network is formed, each collected data of node come from
V class, data dimension p.Wherein, in the collected all data of node m, the data set from v-th of class isWhereinIt indicates at node m, the nth data from v-th of class,For
The training data number of digital v-th of class.
In addition, the data transmission range of each node is set as Dis, it is all to be less than Dis with its distance for present node m
Node be its neighbor node, the neighbor node set expression of node m is Rm.In the present invention, the connection relationship between node
(network topology) is determined in advance, it is only necessary to guarantee at least to have one directly between any two node or can arrive through multi-hop
The path reached.
Step 2: training；
For the data set in all nodes from v-th of classWith mixing common factor analysis
Device (mixture of common factor analyzers, abbreviation MCFA) describes its distribution.It is relevant to v-th of class
Its parameter set of MCFA model isWhereinFor mixed weight-value, meet (q n dimensional vector n) and((q × q) matrix) is respectively that q denapon corresponding with p dimension data is obeyed
The mean value and covariance matrix of Gaussian Profile, q take the arbitrary integer between p/2~p/8.It is completed using following distributed way
Training, specific training process are following (here with v class dataTraining process for, in order to indicate succinct, and will not
It influences to understand and implement, omitted in step belowWithIn subscript
" (v) "):
Step 2-1, initialization: the initial parameter value in setting MCFAWherein, each node
(the w at place1,...,wg,...,wG)=(1/G ..., 1/G ..., 1/G)；Each of L and E matrix element all from standard just
State, which is distributed in N (0,1), to be generated；{ξ1,...,ξg,...,ξGIn each element generate from the standardized normal distribution N (0,1)；
Ω1=...=Ωg=...=ΩG=Iq, wherein IqFor the unit matrix of (q × q).
Step 2-2, broadcast data number: each node l (l=1,2 ..., M) is by its collected data amount check NlExtensively
It broadcasts and gives its neighbor node.After the data amount check that all neighbor nodes broadcast that some node m receives it comes, which is calculated
Weight coefficient clm:
The meaning of the weight is each neighbor node l (l ∈ R for measuring node mm) information transmitted every time is in node m
The importance at place.In addition, iteration count iter=1, starts iterative process.
Step 2-3, local calculation: at each node l in sensor network, according to the data X at current nodel
With the parameter value Θ estimated after last iterationold(as iter=1, ΘoldFor the parameter value after initialization), it calculates
A outl,n,g,bg,hgWith
Wherein,That is:
Step 2-4, broadcast diffusion: each node l in sensor network calculate three intermediate variables, it may be assumed that
It is placed in a data packet, then
The data packet is spread to other node broadcasts.
Step 2-5, combined calculation: when node m (m=1 ..., M) is received from its all neighbor node l (l ∈ Rm) hair
After the data packet containing intermediate variable come, joint statistic is calculatedThat is:
Step 2-6, parameter Estimation: node m (m=1 ..., M) is according to the calculated joint statistic of step (5) and step
(3) calculated to estimate
Step 2-7, differentiate convergence: node m (m=1 ..., M) calculates the log-likelihood under current iteration, it may be assumed that
Wherein ΘnewIndicate the parameter value that current iteration estimates, ΘoldIndicate the estimation parameter value in last iteration.
If logp (Xm|Θnew)-logp(Xm|Θold) < ε, wherein ε=10-5, node m enters final state.Otherwise, step is turned to
2-3 starts next iteration.
The present invention passes through after step 2-1~step 2-7, estimates Θ(v).In the same way, every a kind of number is estimated
According to the parameter set Θ of corresponding MCFA(v)(v=1 ..., V), training process are completed.
Step 3: identification；When any node in network (is assumed to be node m) and collects new data x' for identification
When, x' is calculated about Θ(v)Log-likelihood logp (the x'| Θ of (v=1 ..., V)(v)) (v=1 ..., V), it may be assumed that
Using the corresponding serial number of max log likelihood value as the recognition result v' of x':
Performance evaluation of the invention includes the following:
Equipped with 15 nodes, form a network, guarantee in network between any two node can directly or indirectly into
Row communication.3844 0~9 handwritten numeral training datas collected in UCI data set are assigned randomly on 20 nodes, are instructed
The feature for practicing data is equally spaced to take the two-dimensional coordinate of 8 points to form on the digital handwriting track, totally 16 dimension.So, M=
15, p=16, q take 6,Using training side of the present invention
Method obtains Θ(0),Θ(1),...,Θ(9).After the completion of training, in order to test the overall discrimination of all nodes, by each test
Data x'(3500 in total) distinguish on all the nodes, it is identified using identification process of the present invention.Due to wait know
It is correct to obtain identification it is known that will be compared using recognition methods according to the present invention and its true value for other digital realistic value
Rate (that is: recognition correct rate=all nodes correctly identify quantity/(15*3500) of handwritten numeral), so as to evaluating and
Measure out the validity and accuracy of method according to the present invention.
For the Distributed Classification (abbreviation D-MCFA) more proposed by the present invention based on MCFA and other methods
Performance, the centralized Handwritten Digit Recognition method (abbreviation C-MCFA) here and based on MCFA, nothing between each node based on MCFA
The Handwritten Digit Recognition method (abbreviation N-MCFA) of cooperation, the Handwritten Digit Recognition method based on distributed gauss hybrid models
(abbreviation D-GMM) is compared.It should be noted that all nodes are needed original data transmissions to some in C-MCFA
Central node is completed training using traditional MCFA by central node and identified, result is then returned to each node again, this
Kind of mode in practice seldom, first is that transmission raw data communication expense is very big, once there is packet loss or data packet damage,
Very big on last recognition performance influence, two are detrimental to the secret protection in data, and network security causes anxiety.Recognition result difference
It is indicated with qualitative and quantitative two ways.In the qualitative representation of result, schemed using the hinton of confusion matrix, such as Fig. 2 institute
Show.In the figure, the recognition result of each list registration word 0~9 and each row indicate the true value of number 0~9.It is small on leading diagonal
Square indicates the case where correct identification, and the size of small cube shows that more greatly the number of the correct identification is more, and in other positions
There is small cube and shows the case where there are wrong identifications.It can be seen from this figure that C-MCFA and D-MCFA of the invention are (here only
One of node is provided, other node results are identical) better performances, and D-GMM performance is poor, that worst is N-MCFA.
In the quantificational expression of result, using the mean value and variance two indices of discrimination, as shown in Figure 3.In the figure, the present invention is set
The recognition correct rate and variance and C-MCFA of the D-MCFA of meter is essentially identical, and the variance of each node recognition correct rate compared with
It is low, and N-MCFA and D-GMM recognition correct rate and variance performance are all poor.In addition, will be using distributed of the present invention
D-MCFA and the existing distributed Handwritten Digit Recognition method based on the analysis of t hybrid cytokine are compared, the results showed that this hair
The discrimination of bright method is suitable with the distributed Handwritten Digit Recognition method recognition performance analyzed based on t hybrid cytokine, but D-
MCFA completes to train the required time to be 180 seconds, and the distributed Handwritten Digit Recognition method based on the analysis of t hybrid cytokine is completed
It is 420 seconds the time required to training, therefore, D-MCFA improves arithmetic speed.
Claims (3)
1. the distributed high dimensional data classification method based on mixing common factor analyzer, which is characterized in that the described method includes:
Step 1, the acquisition of data；
A network is formed equipped with M node, each collected data of node come from V class, data dimension p；Wherein, it saves
In the collected all data of point m, the data set from v-th of class isWhereinIt indicates at node m, the nth data from v-th of class,For the training data number of digital v-th of class；In addition, section
The neighbor node set expression of point m is Rm；
Step 2, training: for the data in all nodes from v-th of classWith mixing common factor
Analyzer (MCFA) describes its distribution, and completes the training of model using distributed way, estimates parameterThe parameter set of MCFA corresponding to every a kind of data is estimated in the same way
Θ(v)(v=1 ..., V)；
WhereinFor mixed weight-value, meet (q n dimensional vector n) and((q × q) matrix) is respectively to tie up with p
The mean value and covariance matrix for the Gaussian Profile that corresponding q denapon of data is obeyed, q takes any whole between p/2~p/8
Number, each of L and E matrix element are all generated from standardized normal distribution N (0,1)；
Step 3, it identifies: when any node in network collects new data x' for identification, calculating x' about Θ(v)
Log-likelihood logp (the x'| Θ of (v=1 ..., V)(v)) (v=1 ..., V):
Using the corresponding serial number of max log likelihood value as the recognition result v' of x':
2. the distributed high dimensional data classification method according to claim 1 based on mixing common factor analyzer, special
Sign is that the training process of the data in the step 2 for v-th of class includes:
Step 2-1, initialization: the initial parameter value in setting MCFAWherein, at each node
(w1,...,wg,...,wG)=(1/G ..., 1/G ..., 1/G)；Each of L and E matrix element is all from standard normal point
It is generated in cloth N (0,1)；{ξ1,...,ξg,...,ξGIn each element generate from the standardized normal distribution N (0,1)；Ω1
=...=Ωg=...=ΩG=Iq, wherein IqFor the unit matrix of (q × q)；
Step 2-2, broadcast data number: each node l (l=1,2 ..., M) is by its collected data amount check NlIt is broadcast to it
Neighbor node；After the data amount check that all neighbor nodes broadcast that some node m receives it comes, which calculates weight system
Number clm:
In addition, iteration count iter=1, starts iterative process；
Step 2-3, local calculation: at each node l in sensor network, according to the data X at current nodelWith it is upper
The parameter value Θ estimated after an iterationold, as iter=1, ΘoldFor the parameter value after initialization, calculate
al,n,g,bg,hgWithIts formula are as follows:
Wherein,
Step 2-4, broadcast diffusion: each node l in sensor network calculate three groups of intermediate variables:It is placed in a data packet, is then saved to other
The data packet is spread in point broadcast；
Step 2-5, combined calculation: when node m (m=1 ..., M) is received from its all neighbor node l (l ∈ Rm) containing of sending
After having the data packet of intermediate variable, joint statistic is calculated
Step 2-6, parameter Estimation: node m (m=1 ..., M) is according to the calculated joint statistic of step 2-5 and step 2-3
Calculated bg,gg,It estimates
Step 2-7, differentiate convergence: node m (m=1 ..., M) calculates the log-likelihood under current iteration, it may be assumed that
Wherein ΘnewIndicate the parameter value that current iteration estimates, ΘoldIndicate the estimation parameter value in last iteration；If
logp(Xm|Θnew)-logp(Xm|Θold) < ε, wherein ε=10-5, node m enters final state；Otherwise, step 2-3 is turned to
Start next iteration；
After step 2-1~step 2-7, Θ is estimated(v), complete the training of the data of v-th of class.
3. the distributed high dimensional data classification method according to claim 1 based on mixing common factor analyzer, special
Sign is: the method is applied to the parallel and distributed process of data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510916426.6A CN105550704B (en) | 2015-12-10 | 2015-12-10 | Distributed high dimensional data classification method based on mixing common factor analyzer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510916426.6A CN105550704B (en) | 2015-12-10 | 2015-12-10 | Distributed high dimensional data classification method based on mixing common factor analyzer |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105550704A CN105550704A (en) | 2016-05-04 |
| CN105550704B true CN105550704B (en) | 2019-01-01 |
Family
ID=55829887
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510916426.6A Expired - Fee Related CN105550704B (en) | 2015-12-10 | 2015-12-10 | Distributed high dimensional data classification method based on mixing common factor analyzer |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105550704B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8873836B1 (en) * | 2012-06-29 | 2014-10-28 | Emc Corporation | Cluster-based classification of high-resolution data |
| CN104484409A (en) * | 2014-12-16 | 2015-04-01 | 芜湖乐锐思信息咨询有限公司 | Data mining method for big data processing |
| CN104994170A (en) * | 2015-07-15 | 2015-10-21 | 南京邮电大学 | Distributed clustering method based on mixed factor analysis model in sensor network |
-
2015
- 2015-12-10 CN CN201510916426.6A patent/CN105550704B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8873836B1 (en) * | 2012-06-29 | 2014-10-28 | Emc Corporation | Cluster-based classification of high-resolution data |
| CN104484409A (en) * | 2014-12-16 | 2015-04-01 | 芜湖乐锐思信息咨询有限公司 | Data mining method for big data processing |
| CN104994170A (en) * | 2015-07-15 | 2015-10-21 | 南京邮电大学 | Distributed clustering method based on mixed factor analysis model in sensor network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105550704A (en) | 2016-05-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112085952B (en) | Method and device for monitoring vehicle data, computer equipment and storage medium | |
| WO2021017303A1 (en) | Person re-identification method and apparatus, computer device and storage medium | |
| CN102810161B (en) | Method for detecting pedestrians in crowding scene | |
| CN102571486B (en) | Traffic identification method based on bag of word (BOW) model and statistic features | |
| US10037495B2 (en) | Clustering coefficient-based adaptive clustering method and system | |
| CN109815788A (en) | A kind of picture clustering method, device, storage medium and terminal device | |
| CN105814582B (en) | Method and system for recognizing human face | |
| WO2014106363A1 (en) | Mobile device positioning system and method | |
| CN110166344B (en) | Identity identification method, device and related equipment | |
| CN105005760A (en) | Pedestrian re-identification method based on finite mixture model | |
| CN112016605A (en) | Target detection method based on corner alignment and boundary matching of bounding box | |
| CN111915015B (en) | Abnormal value detection method and device, terminal equipment and storage medium | |
| CN113422695B (en) | Optimization method for improving robustness of topological structure of Internet of things | |
| CN105024993A (en) | Protocol comparison method based on vector operation | |
| CN103390151A (en) | Face detection method and device | |
| CN107392254A (en) | A kind of semantic segmentation method by combining the embedded structural map picture from pixel | |
| CN106503631A (en) | A kind of population analysis method and computer equipment | |
| CN103310235A (en) | Steganalysis method based on parameter identification and estimation | |
| CN114783021A (en) | Intelligent detection method, device, equipment and medium for wearing of mask | |
| CN108304852B (en) | Method and device for determining road section type, storage medium and electronic device | |
| CN102722732A (en) | Image set matching method based on data second order static modeling | |
| CN103942526A (en) | Linear feature extraction method for discrete data point set | |
| CN103002472B (en) | The method that event boundaries in a kind of heterogeneous body sensor network detects, device and intelligent communication equipment | |
| CN105550704B (en) | Distributed high dimensional data classification method based on mixing common factor analyzer | |
| CN104462826B (en) | The detection of multisensor evidences conflict and measure based on Singular Value Decomposition Using |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 210003 new model road 66, Gulou District, Nanjing, Jiangsu Applicant after: NANJING University OF POSTS AND TELECOMMUNICATIONS Address before: 210023 9 Wen Yuan Road, Qixia District, Nanjing, Jiangsu. Applicant before: NANJING University OF POSTS AND TELECOMMUNICATIONS |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20201207 Address after: Gulou District of Nanjing City, Jiangsu Province, Beijing Road No. 20 210024 Patentee after: STATE GRID JIANGSU ELECTRIC POWER Co.,Ltd. INFORMATION & TELECOMMUNICATION BRANCH Address before: Room 214, building D5, No. 9, Kechuang Avenue, Zhongshan Science and Technology Park, Jiangbei new district, Nanjing, Jiangsu Province Patentee before: Nanjing Tian Gu Information Technology Co.,Ltd. Effective date of registration: 20201207 Address after: Room 214, building D5, No. 9, Kechuang Avenue, Zhongshan Science and Technology Park, Jiangbei new district, Nanjing, Jiangsu Province Patentee after: Nanjing Tian Gu Information Technology Co.,Ltd. Address before: 210003 Gulou District, Jiangsu, Nanjing new model road, No. 66 Patentee before: NANJING University OF POSTS AND TELECOMMUNICATIONS |
|
| TR01 | Transfer of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190101 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |