CN107229702A - Micro- video popularity Forecasting Methodology with various visual angles Fusion Features is constrained based on low-rank - Google Patents
Micro- video popularity Forecasting Methodology with various visual angles Fusion Features is constrained based on low-rank Download PDFInfo
- Publication number
- CN107229702A CN107229702A CN201710378158.6A CN201710378158A CN107229702A CN 107229702 A CN107229702 A CN 107229702A CN 201710378158 A CN201710378158 A CN 201710378158A CN 107229702 A CN107229702 A CN 107229702A
- Authority
- CN
- China
- Prior art keywords
- micro
- rank
- low
- video
- feature
- 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
- 230000000007 visual effect Effects 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims abstract description 48
- 230000004927 fusion Effects 0.000 title claims abstract description 33
- 239000011159 matrix material Substances 0.000 claims abstract description 40
- 238000010219 correlation analysis Methods 0.000 claims abstract description 12
- 238000012545 processing Methods 0.000 claims abstract description 8
- 238000000605 extraction Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 101100025318 Danio rerio mvd gene Proteins 0.000 description 1
- 235000010627 Phaseolus vulgaris Nutrition 0.000 description 1
- 244000046052 Phaseolus vulgaris Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
- 238000012795 verification 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
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/735—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/783—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/7834—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using audio features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/783—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/7844—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using original textual content or text extracted from visual content or transcript of audio data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/7867—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, title and artist information, manually generated time, location and usage information, user ratings
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Library & Information Science (AREA)
- Databases & Information Systems (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Primary Health Care (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Computational Linguistics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of micro- video popularity Forecasting Methodology constrained based on low-rank with various visual angles Fusion Features, methods described includes：Low-rank approximate processing is carried out respectively to 4 kinds of visual angle modal characteristics, obtains removing 4 kinds of low-rank characteristic informations of noise；Fusion Features are carried out to 4 kinds of low-rank characteristic informations by various visual angles information canonical correlation analysis；Using the characteristic information after fusion, the Laplacian Matrix for representing the figure relation between each micro- video is set up；Based on Laplacian Matrix, the popularity of micro- video is predicted using semi-supervised method.Present invention, avoiding limitation of the feature of single visual angle for Popularity prediction, the feature for handling each visual angle is constrained using low-rank so that the Laplacian Matrix between the feature of foundation has higher stability.
Description
Technical field
Melt the present invention relates to micro- video Popularity prediction field, more particularly to a kind of constrained based on low-rank with various visual angles feature
The micro- video popularity Forecasting Methodology closed.
Background technology
With the popularization of network technology and social platform, micro- video as a kind of new user content, is received more
Carry out more concerns.Micro- video refers to short then 30 seconds, video clip that is long then being no more than 20 minutes.The appearance of micro- video, not only
Meet network viewing custom and the mobile terminal characteristic under modern society's fast pace life mode, can also meet amusement blast, note
The autonomous sense of participation of the meaning rare epoch consumer of power and the demand of notice return rate, it is contemplated that " micro- video " brings masses
Will be that random whenever and wherever possible video is enjoyed.And the prediction of micro- video popularity is in advertisement pushing, video recommendations and reserved
Directive function is respectively provided with terms of bandwidth, therefore, the prediction for micro- video popularity has great importance.
In actual life, each object can be indicated with a variety of different visual angle characteristics, for example：Micro- video
Expression there may be acoustic feature, visual signature, the diversified forms such as social property feature and text feature, different visual angles
Feature can play different effects for the prediction of micro- video popularity, therefore Fusion Features and feature selecting are also current ratio
The method of more popular processing various visual angles feature[1]。
Inventor is during the present invention is realized, discovery at least has the following disadvantages and not enough in the prior art：
In actual application, because the change of external environment condition and the shake of camera cause micro- video to be contaminated, depending on
There is the feature that frequency is extracted noise can not be completely dependent on, and current method does not have solves noise for feature well
The problem of influence, it is impossible to meet a variety of needs in practical application.
The content of the invention
The invention provides a kind of micro- video popularity Forecasting Methodology constrained based on low-rank with various visual angles Fusion Features, sheet
Invention avoids limitation of the feature for Popularity prediction of single visual angle, and the feature for handling each visual angle is constrained using low-rank,
So that the Laplacian Matrix between the feature set up has higher stability, it is described below：
A kind of micro- video popularity Forecasting Methodology constrained based on low-rank with various visual angles Fusion Features, methods described is included：
Low-rank approximate processing is carried out respectively to 4 kinds of visual angle modal characteristics, obtains removing 4 kinds of low-rank characteristic informations of noise；
Fusion Features are carried out to 4 kinds of low-rank characteristic informations by various visual angles information canonical correlation analysis；
Using the characteristic information after fusion, the Laplacian Matrix for representing the figure relation between each micro- video is set up；It is based on
Laplacian Matrix, is predicted using semi-supervised method to the popularity of micro- video.
Methods described also includes：To given 4 kinds of visual angle modal characteristics of micro- video extraction.
4 kinds of visual angle modal characteristics are specially：Visual signature, acoustic feature, text feature and social property feature.
It is described to be specially to 4 kinds of low-rank characteristic informations progress Fusion Features by various visual angles information canonical correlation analysis：
Make the COS distance on the low-rank Projection Character to public subspace at each visual angle maximum, with the public son of low-rank feature
Feature space after space representation fusion, carries out the prediction of micro- video popularity on this basis.
The characteristic information using after fusion, sets up the Laplacian Matrix tool for representing the figure relation between each micro- video
Body is：
Wherein, L is low-rank proper subspaceNormalization Laplacian Matrix, D is diagonal matrix, and its value isEach row
Numerical value sum.
The beneficial effect for the technical scheme that the present invention is provided is：
1st, low-rank approximate processing is carried out to the characteristic information at each visual angle, obtains even closer architectural feature, and adopt
With noise processed is removed, make the Laplacian Matrix got that there is higher stability；
2nd, 4 kinds of Viewing-angle informations are learnt using the method for various visual angles canonical correlation analysis, reaches the mesh of Fusion Features
, study obtains public subspace, and single features space is eliminated with this for the limitation that predicts the outcome；
3rd, the figure relation between each micro- video features is represented using the Laplacian Matrix got, prevalence is improved with this
The precision of prediction is spent, a variety of needs in practical application are met.
Brief description of the drawings
Fig. 1 is a kind of flow chart constrained based on low-rank with micro- video popularity Forecasting Methodology of various visual angles Fusion Features；
Fig. 2 is method proposed by the present invention and the schematic diagram of the comparing result of other Popularity prediction algorithms.
Embodiment
To make the object, technical solutions and advantages of the present invention clearer, further is made to embodiment of the present invention below
It is described in detail on ground.
Embodiment 1
In order to reach preferable prediction effect, it is desirable to be able to comprehensively, automatically, accurately carry out the side of micro- video Popularity prediction
Method.Research shows：Micro- video between close feature has similar popularity.The embodiment of the present invention proposes a kind of based on low
Micro- video popularity Forecasting Methodology of order constraint and various visual angles Fusion Features, it is described below referring to Fig. 1：
101：Low-rank approximate processing is carried out respectively to 4 kinds of visual angle modal characteristics, obtains removing 4 kinds of low-rank features letter of noise
Breath；
102：Fusion Features are carried out to 4 kinds of low-rank characteristic informations by various visual angles information canonical correlation analysis；
103：Using the characteristic information after fusion, the Laplacian Matrix for representing the figure relation between each micro- video is set up；
Based on Laplacian Matrix, the popularity of micro- video is predicted using semi-supervised method.
Wherein, before step 101, this method also includes：To given 4 kinds of visual angle modal characteristics of micro- video extraction.
Further, above-mentioned 4 kinds of visual angle modal characteristics are specially：Visual signature, acoustic feature, text feature and society
Attributive character.
Wherein, 4 kinds of low-rank characteristic informations progress features are melted by various visual angles information canonical correlation analysis in step 102
Conjunction is specially：
Make the COS distance on the low-rank Projection Character to public subspace at each visual angle maximum, with the public son of low-rank feature
Feature space after space representation fusion, carries out the prediction of micro- video popularity on this basis.
Wherein, the characteristic information utilized after fusion in step 103, sets up the drawing for representing the figure relation between each micro- video
This matrix of pula is specially：
Wherein, L is low-rank proper subspaceNormalization Laplacian Matrix, D is diagonal matrix, and its value isEach row
Numerical value sum.
In summary, the embodiment of the present invention by above-mentioned steps 101- steps 103 avoid the feature of single visual angle for
The limitation of Popularity prediction, the feature for handling each visual angle is constrained using low-rank so that the Laplce between the feature of foundation
Matrix has higher stability.
Embodiment 2
The scheme in embodiment 1 is further introduced with reference to specific calculation formula, example, it is as detailed below
Description：
201：To given 4 kinds of visual angle modal characteristics of micro- video extraction, i.e.,：Visual signature, acoustic feature, text feature with
And social property feature；
Common feature of the embodiment of the present invention first to the given micro- video research of 4 kinds of micro- video extraction, including：Vision is special
Levy, acoustic feature, text feature and social property feature.
1st, visual signature includes：Object information in color histogram information, micro- video (can use convolutional neural networks
Method is obtained, it would however also be possible to employ other method is obtained, and the embodiment of the present invention is without limitation) and aesthetic features.
2nd, acoustic feature includes：The feature of music and other main background sounds in micro- video.
3rd, text feature includes：Text marking in micro- video etc., can use word2vec[2]Method is directly obtained.
4th, social property feature refers to the information of user account, including：Whether account is by verifying, the information such as bean vermicelli number.
Influence is played in the prediction of the feature of this 4 kinds of mode energy pop degree, and is complemented one another.
Wherein, above-mentioned visual signature, acoustic feature, text feature and social property are characterized as known to micro- video field
Technical term, the embodiment of the present invention only does simple introduction to this, will not be described here.
202：4 kinds of visual angle modal characteristics are respectively processed using the approximate method of low-rank, obtained after low-rank processing
Remove 4 kinds of low-rank characteristic informations of noise；
In 4 kinds of visual angle modal characteristics of actual extracting, due to the noise of micro- video in itself, the influence such as visual angle can cause
The video figure relation applicability of foundation is not high, therefore micro- video modality feature of extraction is handled using low-rank, removes
The polluted informations such as noise so that the Laplacian Matrix of foundation is more stablized.The formula of implicit low-rankization processing is expressed as follows：
s.t.Xk=AkZk+Ek (1)
Wherein, λ is the equilibrium constant, | | | |1The l-1 norms of representing matrix, | | | |*The trace norm of representing matrix, XkFor
The initial characteristic data at k visual angle, ZkRepresent low-rank transition matrix, EkRepresent noise information, AkRepresent dictionary square set in advance
Battle array, in general, from practical application angle, generally selectes Ak=Xk, then the low-rank table of original feature space is obtained
Show resultThe minimum of above-mentioned object function trace norm can shrink (SVT) Algorithm for Solving, tool using singular value
Body solution procedure is known to those skilled in the art, and the embodiment of the present invention is not repeated this.
203：4 kinds of low-rank characteristic informations are handled using various visual angles information canonical correlation analysis, Fusion Features are carried out；
The low-rank of the characteristic at 4 visual angles has been obtained in step 202. as a result, it is desirable to utilize various visual angles canonical correlation
The method of analysis carries out Fusion Features, to obtain public subspace to consider the information at each visual angle.Various visual angles typical case
The formula of correlation analysis is as follows：
Wherein, W1,...,WKFor the Feature Conversion matrix in various visual angles canonical correlation analysis, SijMicro- for different visual angles regards
Covariance matrix between frequency, SiiFor auto-variance matrix, K is the quantity of visual angle characteristic, DiFor the intrinsic dimensionality of i-th of mode, T
Transposition is represented, i, j is the label at the feature visual angle of micro- video, and I is unit matrix, and K value is the positive integer more than 1.
W1,...,WKSolve can use standard TraceRatio[3]The step of method, specific solution is those skilled in the art institute
Known, the embodiment of the present invention is not repeated this.
The purpose of various visual angles canonical correlation analysis is to calculate a public subspace so that the Projection Character at each visual angle is arrived
COS distance on the public subspace is maximum, i.e., closer to can then represent the feature after merging with the public subspace
Space, carries out the prediction of micro- video popularity on this basis.
204：Using the characteristic information after fusion, the Laplacian Matrix for representing the figure relation between each micro- video is set up；
In the presence of a priori, i.e., similar micro- video should have similar popularity fraction, in this priori
On the basis of, it is desirable to the figure relation set up between each micro- video.The method of figure relation between commonplace each video of sign
It is to set up Laplacian Matrix, the specific method that it calculates Laplacian Matrix is as follows：
Wherein,It is the radial direction cardinal distance between micro- video of low-rank proper subspace from representing the phase between each micro- video
Like degree, σ0Represent low-rank proper subspaceEuclidean distance intermediate value.It is expressed as i-th and j-th
Micro- video is in low-rank proper subspaceIn feature.
On this basis, normalization Laplacian Matrix can be calculated as follows：
Wherein, L is low-rank proper subspaceNormalization Laplacian Matrix, D is diagonal matrix, and its value isEach row
Numerical value sum.Then the similar micro- video of feature may there can be this priori of similar popularity with mathematical expression
The form of formula is write out as follows：
Wherein, f represents micro- video popularity fraction of prediction.
205：Based on Laplacian Matrix, the popularity of micro- video is predicted using semi-supervised method.
On the basis of Laplacian Matrix, the prediction of popularity fraction is carried out by the way of Semi-Supervised Regression, specifically
Operation is as follows：
Wherein, α is coefficient of balance, and the popularity that f obtains for prediction, y is real popularity fraction, and M is diagonal matrix,
Wherein markd micro- pixel value is 1, and unlabelled micro- pixel value is 0, in the Popularity prediction, it is only necessary to ensure training set
Popularity it is close with actual value, the popularity fraction of test set is then defined with figure relation.The solution of object function can be with
Solved using the derivation mode of standard, specific solution procedure is known to those skilled in the art, the embodiment of the present invention pair
This is not repeated.
In summary, the embodiment of the present invention by above-mentioned steps 201- steps 205 avoid the feature of single visual angle for
The limitation of Popularity prediction, the feature for handling each visual angle is constrained using low-rank so that the Laplce between the feature of foundation
Matrix has higher stability.
Embodiment 3
Feasibility checking is carried out to the scheme in Examples 1 and 2 with reference to specific example, it is described below：
First, test data set
The test data set that this experiment is used is the micro- video set downloaded from Vine social network sites, and its micro- video length is equal
For 6S.
2nd, evaluation criteria
Micro- video popularity estimated performance of this method is weighed using mean square error and p value, mean square error (nMSE) is characterized
The accuracy of prediction, p value (P-value) characterizes forecasting reliability.
3rd, algorithm is contrasted
This method is contrasted with a variety of methods in experiment, including TMALL[1],MLR[2],Lasso[3],SVR[4],
RegMVMT[5],MLHR[6],MSNL[7],MvDA[8]Deng 8 kinds of micro- video popularity Forecasting Methodologies commonplace in the recent period.
4th, experimental result
Fig. 2 is this method and the nMSE of other 8 kinds micro- video popularity prediction algorithms and the comparing result of p value index.By
Contrast understands that TLRMVR (method proposed by the present invention) its degree of accuracy (nMSE values are minimum) predicted on available data collection is higher than
Other control methods, stability (nMSE mean square error is smaller) is preferably.P is calculated for control with this experiment by other method
Value, p value is smaller value, it was demonstrated that the reliability of this method.The experimental verification feasibility and superiority of this method.
Bibliography：
[1]Chen J,Song X,Nie L,et al.Micro tells macro:predicting the
popularity of micro-videos via a transductive model[C]//Proceedings of the
2016ACM on Multimedia Conference.ACM,2016:898-907.
[2]T.Mikolov,I.Sutskever,K.Chen,G.S.Corrado,and J.Dean.Distributed
representations of words and phrases and their compositionality.In
Proceedings of the Annual Conference on Neural Information Processing
Systems,pages 3111–3119.NIPS Foundation,2013.
[3]Yangqing Jia,Feiping Nie,Changshui Zhang.Trace Ratio Problem
Revisited.IEEE Transactions on Neural Networks(TNN),Volume 20,Issue 4,Pages
729-735,2009.
[4]A.J.Smola and B.Scholkopf,“A tutorial on support vector
regression,”Statistics and computing,vol.14,no.3,pp.199–222,2004.
[5]J.Zhang and J.Huan,“Inductive multi-task learning with multiple
view data,”in Proceedings of ACM International Conference on Knowledge
Discovery and Data Mining.ACM,2012,pp.543–551.
[6]Y.Yang,J.Song,Z.Huang,and Z.Ma,“Multi-feature fusion via
hierarchical regression for multimedia analysis,”IEEE Transactions on
Multimedia,vol.15,no.3,pp.572–581,2013.
[7]X.Song,L.Nie,L.Zhang,M.Akbari,and T.-S.Chua,“Multiple social
network learning and its application in volunteerism tendency prediction,”in
Proceedings of ACM SIGIR Conference on Research and Development in
Information Retrieval.ACM,2015,pp.213–222.
[8]M.Kan,S.Shan,H.Zhang,S.Lao,and X.Chen,“Multi-view discriminant
analysis,”IEEE Transactions on Pattern Analysis and Machine Intelligence,
vol.38,no.1,pp.188–194,2016.
The embodiment of the present invention is to the model of each device in addition to specified otherwise is done, and the model of other devices is not limited,
As long as the device of above-mentioned functions can be completed.
It will be appreciated by those skilled in the art that accompanying drawing is the schematic diagram of a preferred embodiment, the embodiments of the present invention
Sequence number is for illustration only, and the quality of embodiment is not represented.
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc. should be included in the scope of the protection.
Claims (5)
1. a kind of micro- video popularity Forecasting Methodology constrained based on low-rank with various visual angles Fusion Features, it is characterised in that described
Method includes：
Low-rank approximate processing is carried out respectively to 4 kinds of visual angle modal characteristics, obtains removing 4 kinds of low-rank characteristic informations of noise；
Fusion Features are carried out to 4 kinds of low-rank characteristic informations by various visual angles information canonical correlation analysis；
Using the characteristic information after fusion, the Laplacian Matrix for representing the figure relation between each micro- video is set up；It is general based on drawing
Lars matrix, is predicted using semi-supervised method to the popularity of micro- video.
2. a kind of micro- video Popularity prediction side constrained based on low-rank with various visual angles Fusion Features according to claim 1
Method, it is characterised in that methods described also includes：To given 4 kinds of visual angle modal characteristics of micro- video extraction.
3. a kind of micro- video popularity based on low-rank constraint and various visual angles Fusion Features according to claim 1 or 2 is pre-
Survey method, it is characterised in that 4 kinds of visual angle modal characteristics are specially：Visual signature, acoustic feature, text feature and society
Can attributive character.
4. a kind of micro- video Popularity prediction side constrained based on low-rank with various visual angles Fusion Features according to claim 1
Method, it is characterised in that described specific to 4 kinds of low-rank characteristic informations progress Fusion Features by various visual angles information canonical correlation analysis
For：
Make the COS distance on the low-rank Projection Character to public subspace at each visual angle maximum, with the public subspace of low-rank feature
The feature space after fusion is represented, the prediction of micro- video popularity is carried out on this basis.
5. a kind of micro- video Popularity prediction side constrained based on low-rank with various visual angles Fusion Features according to claim 1
Method, it is characterised in that the characteristic information using after fusion, sets up the Laplce for representing the figure relation between each micro- video
Matrix is specially：
<mrow>
<mi>L</mi>
<mo>=</mo>
<mi>I</mi>
<mo>-</mo>
<msup>
<mi>D</mi>
<mrow>
<mo>-</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
</mrow>
</msup>
<msubsup>
<mi>S</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mn>0</mn>
</msubsup>
<msup>
<mi>D</mi>
<mrow>
<mo>-</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
</mrow>
</msup>
</mrow>
Wherein, L is low-rank proper subspaceNormalization Laplacian Matrix, D is diagonal matrix, and its value isThe number respectively arranged
It is worth sum.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710378158.6A CN107229702B (en) | 2017-05-24 | 2017-05-24 | Micro video popularity prediction method based on low-rank constraint and multi-view feature fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710378158.6A CN107229702B (en) | 2017-05-24 | 2017-05-24 | Micro video popularity prediction method based on low-rank constraint and multi-view feature fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107229702A true CN107229702A (en) | 2017-10-03 |
| CN107229702B CN107229702B (en) | 2020-09-11 |
Family
ID=59934315
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710378158.6A Expired - Fee Related CN107229702B (en) | 2017-05-24 | 2017-05-24 | Micro video popularity prediction method based on low-rank constraint and multi-view feature fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107229702B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109189930A (en) * | 2018-09-01 | 2019-01-11 | 网易（杭州）网络有限公司 | Text character extraction and extraction model optimization method and medium, device and equipment |
| CN110110610A (en) * | 2019-04-16 | 2019-08-09 | 天津大学 | A kind of event detecting method for short-sighted frequency |
| CN110505520A (en) * | 2018-03-12 | 2019-11-26 | 腾讯科技（深圳）有限公司 | Information recommendation method and system, medium and electronic equipment |
| CN111723241A (en) * | 2020-05-08 | 2020-09-29 | 天津大学 | Short video automatic labeling method based on feature and multi-label enhanced representation |
| CN111984872A (en) * | 2020-09-09 | 2020-11-24 | 北京中科研究院 | Multi-modal information social media popularity prediction method based on iterative optimization strategy |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090034808A1 (en) * | 2006-07-10 | 2009-02-05 | Siemens Corporate Research, Inc. | Automatic Cardiac View Classification of Echocardiography |
| CN105590115A (en) * | 2015-12-18 | 2016-05-18 | 北京畅景立达软件技术有限公司 | Visual-angle-invariant human feature expression method |
| US20160283858A1 (en) * | 2015-03-24 | 2016-09-29 | International Business Machines Corporation | Multimodal Data Fusion by Hierarchical Multi-View Dictionary Learning |
-
2017
- 2017-05-24 CN CN201710378158.6A patent/CN107229702B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090034808A1 (en) * | 2006-07-10 | 2009-02-05 | Siemens Corporate Research, Inc. | Automatic Cardiac View Classification of Echocardiography |
| US20160283858A1 (en) * | 2015-03-24 | 2016-09-29 | International Business Machines Corporation | Multimodal Data Fusion by Hierarchical Multi-View Dictionary Learning |
| CN105590115A (en) * | 2015-12-18 | 2016-05-18 | 北京畅景立达软件技术有限公司 | Visual-angle-invariant human feature expression method |
Non-Patent Citations (2)
| Title |
|---|
| JINGYUAN CHEN等: "《Micro tells macro:predicting the popularity of micro-videos via a transductive model》", 《PROCEEDINGS OF THE 2016ACM ON MULTIMEDIA CONFERENCE.ACM》 * |
| SHUAI ZHENG等: "《A Closed Form Solution to Multi-View Low-Rank Regression》", 《PROCEEDINGS OF THE TWENTY-NINTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110505520A (en) * | 2018-03-12 | 2019-11-26 | 腾讯科技（深圳）有限公司 | Information recommendation method and system, medium and electronic equipment |
| CN110505520B (en) * | 2018-03-12 | 2021-07-13 | 腾讯科技（深圳）有限公司 | Information recommendation method and system, medium and electronic device |
| CN109189930A (en) * | 2018-09-01 | 2019-01-11 | 网易（杭州）网络有限公司 | Text character extraction and extraction model optimization method and medium, device and equipment |
| CN110110610A (en) * | 2019-04-16 | 2019-08-09 | 天津大学 | A kind of event detecting method for short-sighted frequency |
| CN110110610B (en) * | 2019-04-16 | 2023-06-30 | 天津大学 | Event detection method for short video |
| CN111723241A (en) * | 2020-05-08 | 2020-09-29 | 天津大学 | Short video automatic labeling method based on feature and multi-label enhanced representation |
| CN111723241B (en) * | 2020-05-08 | 2023-11-03 | 天津大学 | Short video automatic labeling method based on feature and multi-label enhancement representation |
| CN111984872A (en) * | 2020-09-09 | 2020-11-24 | 北京中科研究院 | Multi-modal information social media popularity prediction method based on iterative optimization strategy |
| CN111984872B (en) * | 2020-09-09 | 2021-03-16 | 北京中科研究院 | Multi-modal information social media popularity prediction method based on iterative optimization strategy |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107229702B (en) | 2020-09-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107229702A (en) | Micro- video popularity Forecasting Methodology with various visual angles Fusion Features is constrained based on low-rank | |
| CN109213868A (en) | Entity level sensibility classification method based on convolution attention mechanism network | |
| CN110990631A (en) | Video screening method and device, electronic equipment and storage medium | |
| CN109635273A (en) | Text key word extracting method, device, equipment and storage medium | |
| CN110633745A (en) | Image classification training method and device based on artificial intelligence and storage medium | |
| CN109961041A (en) | A kind of video frequency identifying method, device and storage medium | |
| CN109657229A (en) | A kind of intention assessment model generating method, intension recognizing method and device | |
| CN112700252B (en) | Information security detection method and device, electronic equipment and storage medium | |
| CN109766557A (en) | A kind of sentiment analysis method, apparatus, storage medium and terminal device | |
| CN106339368A (en) | Text emotional tendency acquiring method and device | |
| CN106339507A (en) | Method and device for pushing streaming media message | |
| CN108509407A (en) | Text semantic similarity calculating method, device and user terminal | |
| CN108491389A (en) | Click bait title language material identification model training method and device | |
| CN110347790A (en) | Text duplicate checking method, apparatus, equipment and storage medium based on attention mechanism | |
| WO2017136674A1 (en) | Generating feature embeddings from a co-occurrence matrix | |
| CN110889759A (en) | Credit data determination method, device and storage medium | |
| CN111400615A (en) | Resource recommendation method, device, equipment and storage medium | |
| CN108305057A (en) | Dispensing apparatus, method and the computer readable storage medium of electronics red packet | |
| CN114386410A (en) | Training method and text processing method of pre-training model | |
| CN107609570B (en) | Micro video popularity prediction method based on attribute classification and multi-view feature fusion | |
| CN111241258A (en) | Data cleaning method and device, computer equipment and readable storage medium | |
| CN110046344A (en) | Add the method and terminal device of separator | |
| CN113362852A (en) | User attribute identification method and device | |
| CN109902273A (en) | The modeling method and device of keyword generation model | |
| CN109933741A (en) | User network behaviors feature extracting method, device and storage medium |
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 | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Su Yuting Inventor after: Li Yang Inventor after: Bai Xu Inventor after: Jing Peiguang Inventor after: Zhang Jing Inventor before: Su Yuting Inventor before: Bai Xu Inventor before: Jing Peiguang Inventor before: Zhang Jing |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200911 Termination date: 20210524 |