CN107122411A - A kind of collaborative filtering recommending method based on discrete multi views Hash - Google Patents
A kind of collaborative filtering recommending method based on discrete multi views Hash Download PDFInfo
- Publication number
- CN107122411A CN107122411A CN201710199300.0A CN201710199300A CN107122411A CN 107122411 A CN107122411 A CN 107122411A CN 201710199300 A CN201710199300 A CN 201710199300A CN 107122411 A CN107122411 A CN 107122411A
- Authority
- CN
- China
- Prior art keywords
- matrix
- article
- data
- represent
- hash
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of collaborative filtering recommending method based on discrete multi views Hash, comprise the following steps:1) the multi views anchor point figure for building data according to the data under different views is represented;2) collaborative filtering and anchor point figure are combined, learning model is obtained;3) obtained learning model is solved, obtains user's binary system Hash coding corresponding with article;4) Nearest Neighbor Search is carried out using obtained Hash coding, calculates preference of the specific user to candidate item, returned to several maximum articles of preference and be used as recommendation results.The present invention is integrated the data under different views, and the discrete feature of coding is remained when solving, the quality of recommendation results is improved.The fast search of similar users is realized using Hash coding simultaneously, the efficiency of recommendation results calculating is improved.
Description
Technical field
The present invention relates to personalized recommendation technology, more particularly to a kind of collaborative filtering recommending based on discrete multi views Hash
Method.
Background technology
The high speed development of Internet industry, brings the explosive increase of content.In order to help effective acquisition of user to believe
Breath, personalized recommendation system is just playing more and more important effect.Collaborative filtering is wide concerned in commending system
A class technology.Carry out that recommendations is different based on the direct filter analysis of content from traditional, collaborative filtering utilizes substantial amounts of user letter
Breath, chooses the user similar to targeted customer, or the article similar to target item, comes consequently recommended current goal user's
Possible article interested.
But in the application environment of reality, we tend to get a large amount of other informations in addition to scoring,
Including the social networks between user and user, the class relations between article and article etc..Traditional collaborative filtering recommending skill
Art can only often utilize the user profile under single view, and need to calculate pair by the computing between high dimension vector
The prediction scoring of user preference, this has had a strong impact on calculating and storage efficiency, and its conventional solutions also result in substantial amounts of information and lose
Lose.
The content of the invention
The purpose of the present invention is that there is provided a kind of collaborative filtering based on discrete multi views Hash in view of the shortcomings of the prior art
Recommendation method.
The purpose of the present invention is achieved through the following technical solutions:
A kind of collaborative filtering recommending method based on discrete multi views Hash, comprises the following steps:
1) according to the data under different views, the multi views anchor point figure for building data is represented;
2) collaborative filtering and anchor point figure are combined, learning model is obtained;
3) obtained learning model is solved, obtains user's binary system Hash coding corresponding with article;
4) Nearest Neighbor Search is carried out using obtained Hash coding, calculates preference of the specific user to candidate item,
Return to several maximum articles of preference and be used as recommendation results.
Above steps can use following specific implementation:
Step 1) include following sub-step:
1.1) the behavioral data matrix for training data under i-th of viewWherein N represents to regard for i-th
The quantity of data point under figure, diThe dimension of data point under i-th of view is represented, T is generated using K-means clustering methodsiIt is individual poly-
Class center, is used as the anchor point of data under the view, TiValue it is related to number of data points;
1.2) training data under different views is subjected to level connection joint and obtains matrix
Wherein M represents the quantity of view, dtotalRepresent the dimension sum of data under all views;
1.3) for each training data, several anchor points of its arest neighbors under each view, composition set are searched for
Diagonal matrix is built using the arest neighbors collection of anchors under different viewsIts
InRepresent the set of j anchor point composition under i-th of view, diRepresent the dimension of data under i-th of view;
1.4) for each data point, Nesterov gradient methods and Projected method solving-optimizing problem are utilizedWhereinRepresent the data point to the transition probability of all arest neighbors anchor points, initial value
It isxiRepresenting matrix X the i-th row;
1.5) data point is set as 0 to the transition probability of non-arest neighbors anchor point, according to obtained each data point to recently
The transition probability of adjacent anchor point, obtains all data points to the transition probability matrix of all anchor pointsWherein TtotalTable
Show the summation of all view anchorage quantity, this transition probability matrix is exactly the multi views anchor point chart of constructed data
Show.
Step 2) include following sub-step:
2.1) row and vector for obtaining transition probability matrix P are calculated
2.2) diagonal matrix is constructed
2.3) similarity matrix S=P Λ are calculated-1PT;
2.4) calculating matrix S degree matrixMake L=C-S;
2.5) tr (ULU are obtainedT)s.t.U∈{0,1}K×N, here it is the learning function obtained by anchor point figure, wherein tr
() represents trace function;
2.6) note score data W={ wij, i=1,2 ..., n;J=1,2 ..., m, n be number of users, m is article number
Amount, wijRepresent scorings of the user i to article j;
2.7) combine and obtain final learning modelAbout
Beam condition is Bl=0, Dl=0, BBT=nI, DDT=mI, wherein biAnd djThe respectively vector of matrix B and D;The respectively encoder matrix of user and article, I is unit matrix, L1And L2Represent respectively by B and
D passes through step 2.1) to the result that 2.4) calculating is obtained, r is code length.
Step 3) include following sub-step:
3.1) note X ∈ [- 1,1]r×n,Y∈[-1,1]r×mThe respectively corresponding continuous matrix of user and article, majorized function
It is expressed asConstrain bar
Part is Xl=0, Yl=0, XXT=nI, YYT=mI, wherein α, β, γ, η are parameter, | | | |FRepresent Frobenius norms;
3.2) calculateRepeat to update user's homography B up to convergence by row step-by-step, whereinWherein bikRepresent vector biK-th of element,Represent to remove kth
Vector b after individual elementiThe vector of remaining all elements composition;djkWithSimilarly, djkRepresent vector djK-th of element,Table
Show vector d after k-th of element of removingjThe vector of remaining all elements composition;xikAnd sijRespectively matrix X and S element;
3.3) calculateMore new article homography D is repeated by row step-by-step until restraining, whereinWherein yjkDefinition and step 3.2) in similar, representing matrix Y vector
yjK-th of element;
3.4) calculating matrixCalculate
3.5) to matrixCarry out feature decompositionObtain PbAnd Σb;
3.6) calculating matrixTo [Qb1] Gram-Schmidt orthogonalizations are carried out to obtain
3.7) pressUpdate matrix X;
3.8) calculating matrixCalculate
3.9) to matrixCarry out feature decompositionObtain Sb;
3.10) calculating matrixTo [Sb1] Gram-Schmidt orthogonalizations are carried out to obtain
3.11) pressUpdate matrix Y;
3.12) repeat step 3.2), step 3.3), step 3.7) and step 3.11), until binary coded matrix B and D
Convergence.
Step 4) include following sub-step:
4.1) according to step 3) the binary system Hash encoder matrix of obtained user and article, calculate specific user and be encoded to
Hamming distance between all items coding, chooses several minimum articles of Hamming distance;
4.2) specific user is gathered into corresponding preference article set and carries out merger, ignore what targeted customer once selected
Article, obtains the preference article candidate collection I of targeted customer;
4.3) for each article in preference article candidate collection, targeted customer is calculated pre- to the predilection grade of article
Measured value simultaneously sorts, and regard the candidate item of K before ranking as consequently recommended result.
The beneficial effects of the invention are as follows:Multi views of the present invention in personalized recommendation, large-scale data scene, will be many
View Hash learning algorithm is combined with the recommended technology based on collaborative filtering, has been merged separate sources, different types of has been regarded more
Diagram data, and the discrete feature that is encoded in solution procedure is always maintained at, so as to improve the quality of recommendation results;In addition, passing through
User and article are expressed as corresponding binary system Hash coding, quick similarity is realized, greatly improves recommendation knot
The efficiency that fruit calculates.
Brief description of the drawings
Fig. 1 is the collaborative filtering recommending method flow chart of the invention based on discrete multi views Hash.
Fig. 2 is the comparative result figure of the invention with other Hash proposed algorithms when being directed to data set MovieLens-1M
(MAP);
Fig. 3 is the comparative result figure of the invention with other Hash proposed algorithms when being directed to data set MovieLens-1M
(MNDCG);
Fig. 4 is the comparative result figure of the invention with other Hash proposed algorithms when being directed to data set Flixster
(MAP);
Fig. 5 is the comparative result figure of the invention with other Hash proposed algorithms when being directed to data set Flixster
(MNDCG);
Embodiment
The present invention is described in further detail below in conjunction with the accompanying drawings.
As shown in figure 1, a kind of collaborative filtering recommending method based on discrete multi views Hash of the present invention, including following step
Suddenly:
1) according to the data under different views, the multi views anchor point figure for building data is represented;Specifically include following sub-step:
1.1) the behavioral data matrix for training data under i-th of viewWherein N represents data point
Quantity, diThe dimension of data point under i-th of view is represented, T is generated using K-means clustering methodsiIndividual cluster centre, as
The anchor point of data, T under the viewiValue it is related to number of data points;
1.2) training data under different views is subjected to level connection joint and obtains matrix
Wherein M represents the quantity of view, dtotalRepresent the dimension sum of data under all views;
1.3) for each training data, several anchor points of its arest neighbors under each view, composition set are searched for
Diagonal matrix is built using the arest neighbors collection of anchors under different viewsIts
InRepresent the set of j anchor point composition under i-th of view, diRepresent the dimension of data under i-th of view;
1.4) for each data point, Nesterov gradient methods and Projected method solving-optimizing problem are utilizedWhereinRepresent the data point to the transition probability of all arest neighbors anchor points, initial value
It isxiRepresenting matrix X the i-th row;
1.5) data point is set as 0 to the transition probability of non-arest neighbors anchor point, according to obtained each data point to recently
The transition probability of adjacent anchor point, obtains all data points to the transition probability matrix of all anchor pointsWherein TtotalTable
Show the summation of all view anchorage quantity, this transition probability matrix is exactly the multi views anchor point chart of constructed data
Show.
2) collaborative filtering and anchor point figure are combined, learning model is obtained;Specifically include following sub-step:
2.1) row and vector for obtaining transition probability matrix P are calculated
2.2) diagonal matrix is constructed
2.3) similarity matrix S=P Λ are calculated-1PT;
2.4) calculating matrix S degree matrixMake L=C-S;
2.5) tr (ULU are obtainedT)s.t.U∈{0,1}K×N, here it is the learning function obtained by anchor point figure, wherein tr
() represents trace function;
2.6) note score data W={ wij, i=1,2 ..., n;J=1,2 ..., m, n be number of users, m is article number
Amount, wijRepresent scorings of the user i to article j;
2.7) combine and obtain final learning modelAbout
Beam condition is Bl=0, Dl=0, BBT=nI, DDT=mI, wherein biAnd djThe respectively vector of matrix B and D;B∈{±1}r×n,
D∈{±1}r×mThe respectively encoder matrix of user and article, I is unit matrix, L1And L2Represent to pass through step by B and D respectively
2.1) to the result that 2.4) calculating is obtained, r is code length.
3) obtained learning model is solved, obtains user's binary system Hash coding corresponding with article;Specific bag
Include following sub-step:
3.1) note X ∈ [- 1,1]r×n,Y∈[-1,1]r×mThe respectively corresponding continuous matrix of user and article, majorized function
It is expressed asConstrain bar
Part is Xl=0, Yl=0, XXT=nI, YYT=mI, wherein α, β, γ, η are parameter, | | | |FRepresent Frobenius norms;
3.2) calculateRepeat to update user's homography B up to convergence by row step-by-step, whereinWherein bikRepresent vector biK-th of element,Represent to remove kth
Vector b after individual elementiThe vector of remaining all elements composition;djkWithSimilarly, djkRepresent vector djK-th of element,Table
Show vector d after k-th of element of removingjThe vector of remaining all elements composition;xikAnd sijRespectively matrix X and S element;
3.3) calculateMore new article homography D is repeated by row step-by-step until restraining, whereinWherein yjkDefinition and step 3.2) in similar, representing matrix Y vector
yjK-th of element;
3.4) calculating matrixCalculate
3.5) to matrixCarry out feature decompositionObtain PbAnd Σb;
3.6) calculating matrixTo [Qb1] Gram-Schmidt orthogonalizations are carried out to obtain
3.7) pressUpdate matrix X;
3.8) calculating matrixCalculate
3.9) to matrixCarry out feature decompositionObtain Sb;
3.10) calculating matrixTo [Sb1] Gram-Schmidt orthogonalizations are carried out to obtain
3.11) pressUpdate matrix Y;
3.12) repeat step 3.2), step 3.3), step 3.7) and step 3.11), until binary coded matrix B and D
Convergence.
4) Nearest Neighbor Search is carried out using obtained Hash coding, calculates preference of the specific user to candidate item,
Return to several maximum articles of preference and be used as recommendation results.Specifically include following sub-step:
4.1) according to step 3) the binary system Hash encoder matrix of obtained user and article, calculate specific user and be encoded to
Hamming distance between all items coding, chooses n minimum article of Hamming distance;
4.2) specific user is gathered into corresponding preference article set and carries out merger, ignore what targeted customer once selected
Article, obtains the preference article candidate collection I of targeted customer;
4.3) for each article in preference article candidate collection, targeted customer is calculated pre- to the predilection grade of article
Measured value simultaneously sorts, and using the candidate item of K before ranking as consequently recommended result, in actual applications, it is left that K typically can use 5 to 20
It is right.
Embodiment
The above method is applied to survey in real data collection MovieLens-1M and Flixster personalized recommendation system
Effect is tried, specific steps are repeated no more.Contrast index Average Accuracy average (Mean Average are set simultaneously
Precision, MAP) and normalization lose storage gain average (Mean Normalized Discounted Cumulative
Gain, MNDCG), collaboration Hash (Collaborative Hashing, CH), local sensitivity Hash is respectively adopted in control methods
(Locality Sensitive Hashing, LSH), multi views anchor point figure Hash (Multi-view Anchor Graph
Hashing, MVAGH) and discrete collaborative filtering (Discrete Collaborative Hashing, DCF).Its result such as Fig. 2
Shown in~5, show on the two indices of two datasets, this method all achieves more preferable effect.
Claims (5)
1. a kind of collaborative filtering recommending method based on discrete multi views Hash, it is characterised in that comprise the following steps:
1) according to the data under different views, the multi views anchor point figure for building data is represented;
2) collaborative filtering and anchor point figure are combined, learning model is obtained;
3) obtained learning model is solved, obtains user's binary system Hash coding corresponding with article;
4) Nearest Neighbor Search is carried out using obtained Hash coding, calculates preference of the specific user to candidate item, return
Several maximum articles of preference are used as recommendation results.
2. a kind of collaborative filtering recommending method based on discrete multi views Hash according to claim 1, it is characterised in that institute
The step 1 stated) include following sub-step:
1.1) the behavioral data matrix for training data under i-th of viewWherein N is represented under i-th of view
Data point quantity, diThe dimension of data point under i-th of view is represented, T is generated using K-means clustering methodsiIn individual cluster
The heart, is used as the anchor point of data under the view, TiValue it is related to number of data points;
1.2) training data under different views is subjected to level connection joint and obtains matrix
Wherein M represents the quantity of view, dtotalRepresent the dimension sum of data under all views;
1.3) for each training data, several anchor points of its arest neighbors under each view, composition set are searched for
Diagonal matrix is built using the arest neighbors collection of anchors under different viewsIts
InRepresent the set of j anchor point composition under i-th of view, diRepresent the dimension of data under i-th of view;
1.4) for each data point, Nesterov gradient methods and Projected method solving-optimizing problem are utilizedWhereinRepresent the data point to the transition probability of all arest neighbors anchor points, initial value
It isxiRepresenting matrix X the i-th row;
1.5) data point is set as 0 to the transition probability of non-arest neighbors anchor point, according to obtained each data point to arest neighbors anchor
The transition probability of point, obtains all data points to the transition probability matrix of all anchor pointsWherein TtotalRepresent institute
There is the summation of view anchorage quantity, this transition probability matrix is exactly that the multi views anchor point figure of constructed data is represented.
3. a kind of collaborative filtering recommending method based on discrete multi views Hash according to claim 1, it is characterised in that institute
The step 2 stated) include following sub-step:
2.1) row and vector for obtaining transition probability matrix P are calculated
2.2) diagonal matrix is constructed
2.3) similarity matrix S=P Λ are calculated-1PT;
2.4) calculating matrix S degree matrixMake L=C-S;
2.5) tr (ULU are obtainedT)s.t. U∈{0,1}K×N, here it is the learning function obtained by anchor point figure, wherein tr ()
Represent trace function;
2.6) note score data W={ wij, i=1,2 ..., n;J=1,2 ..., m, n be number of users, m is number of articles,
wijRepresent scorings of the user i to article j;
2.7) combine and obtain final learning modelConstrain bar
Part is Bl=0, Dl=0, BBT=nI, DDT=mI, wherein biAnd djThe respectively vector of matrix B and D;
B∈{±1}r×n,D∈{±1}r×mThe respectively encoder matrix of user and article, I is unit matrix, L1And L2Difference table
Show and step 2.1 passed through by B and D) to the result that 2.4) calculating is obtained, r is code length.
4. a kind of collaborative filtering recommending method based on discrete multi views Hash according to claim 1, it is characterised in that institute
The step 3 stated) include following sub-step:
3.1) note X ∈ [- 1,1]r×n,Y∈[-1,1]r×mThe respectively corresponding continuous matrix of user and article, majorized function is represented
ForConstraints is
Xl=0, Yl=0, XXT=nI, YYT=mI, wherein α, β, γ, η are parameter, | | | |FRepresent Frobenius norms;
3.2) calculateRepeat to update user's homography B up to convergence by row step-by-step, whereinWherein bikRepresent vector biK-th of element,Represent to remove kth
Vector b after individual elementiThe vector of remaining all elements composition;djkRepresent vector djK-th of element,Represent to remove k-th
Vector d after elementjThe vector of remaining all elements composition;xikAnd sijRespectively matrix X and S element;
3.3) calculateMore new article homography D is repeated by row step-by-step until restraining, whereinWherein yjkRepresenting matrix Y vectorial yjK-th of element;
3.4) calculating matrixCalculate
3.5) to matrixCarry out feature decompositionObtain PbAnd Σb;
3.6) calculating matrixTo [Qb1] Gram-Schmidt orthogonalizations are carried out to obtain
3.7) pressUpdate matrix X;
3.8) calculating matrixCalculate
3.9) to matrixCarry out feature decompositionObtain Sb;
3.10) calculating matrixTo [Sb1] Gram-Schmidt orthogonalizations are carried out to obtain
3.11) pressUpdate matrix Y;
3.12) repeat step 3.2), step 3.3), step 3.7) and step 3.11), until binary coded matrix B and D receive
Hold back.
5. a kind of collaborative filtering recommending method based on discrete multi views Hash according to claim 1, it is characterised in that institute
The step 4 stated) include following sub-step:
4.1) according to step 3) the binary system Hash encoder matrix of obtained user and article, calculate specific user be encoded to it is all
Hamming distance between article code, chooses several minimum articles of Hamming distance;
4.2) specific user is gathered into corresponding preference article set and carries out merger, ignore the thing that targeted customer once selected
Product, obtain the preference article candidate collection I of targeted customer;
4.3) for each article in preference article candidate collection, predilection grade predicted value of the targeted customer to article is calculated
And sort, it regard the candidate item of K before ranking as consequently recommended result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710199300.0A CN107122411B (en) | 2017-03-29 | 2017-03-29 | Collaborative filtering recommendation method based on discrete multi-view Hash |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710199300.0A CN107122411B (en) | 2017-03-29 | 2017-03-29 | Collaborative filtering recommendation method based on discrete multi-view Hash |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107122411A true CN107122411A (en) | 2017-09-01 |
CN107122411B CN107122411B (en) | 2020-08-14 |
Family
ID=59718259
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710199300.0A Active CN107122411B (en) | 2017-03-29 | 2017-03-29 | Collaborative filtering recommendation method based on discrete multi-view Hash |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107122411B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109634952A (en) * | 2018-11-02 | 2019-04-16 | 宁波大学 | A kind of adaptive nearest neighbor querying method towards large-scale data |
CN109886787A (en) * | 2019-02-22 | 2019-06-14 | 清华大学 | Discrete social recommendation method and system |
CN110188825A (en) * | 2019-05-31 | 2019-08-30 | 山东师范大学 | Image clustering method, system, equipment and medium based on discrete multiple view cluster |
CN111104604A (en) * | 2019-11-25 | 2020-05-05 | 北京交通大学 | Lightweight social recommendation method based on Hash learning |
CN111405070A (en) * | 2020-05-29 | 2020-07-10 | 成都信息工程大学 | Multi-core collaborative cross-chain method based on cloud manufacturing platform |
CN113010800A (en) * | 2021-03-04 | 2021-06-22 | 浙江大学 | Combined coding-based collaborative ranking recommendation method |
CN113158032A (en) * | 2021-03-18 | 2021-07-23 | 北京京东乾石科技有限公司 | Information pushing method and device |
CN115309997A (en) * | 2022-10-10 | 2022-11-08 | 浙商银行股份有限公司 | Commodity recommendation method and device based on multi-view self-coding features |
CN116484970A (en) * | 2023-04-13 | 2023-07-25 | 南京大学 | Learning piece searching method based on anchor learning piece |
CN117522532A (en) * | 2024-01-08 | 2024-02-06 | 浙江大学 | Popularity deviation correction recommendation method and device, electronic equipment and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1547351A (en) * | 2003-12-04 | 2004-11-17 | 上海交通大学 | Collaborative filtering recommendation approach for dealing with ultra-mass users |
US20110231405A1 (en) * | 2010-03-17 | 2011-09-22 | Microsoft Corporation | Data Structures for Collaborative Filtering Systems |
CN104462560A (en) * | 2014-12-25 | 2015-03-25 | 广东电子工业研究院有限公司 | Personalized recommendation system and method |
CN104679835A (en) * | 2015-02-09 | 2015-06-03 | 浙江大学 | Book recommending method based on multi-view hash |
CN105608219A (en) * | 2016-01-07 | 2016-05-25 | 上海通创信息技术有限公司 | Stream-oriented recommended engine, recommendation system and recommendation method based on clustering |
CN105956093A (en) * | 2016-04-29 | 2016-09-21 | 浙江大学 | Individual recommending method based on multi-view anchor graph Hash technology |
-
2017
- 2017-03-29 CN CN201710199300.0A patent/CN107122411B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1547351A (en) * | 2003-12-04 | 2004-11-17 | 上海交通大学 | Collaborative filtering recommendation approach for dealing with ultra-mass users |
US20110231405A1 (en) * | 2010-03-17 | 2011-09-22 | Microsoft Corporation | Data Structures for Collaborative Filtering Systems |
CN104462560A (en) * | 2014-12-25 | 2015-03-25 | 广东电子工业研究院有限公司 | Personalized recommendation system and method |
CN104679835A (en) * | 2015-02-09 | 2015-06-03 | 浙江大学 | Book recommending method based on multi-view hash |
CN105608219A (en) * | 2016-01-07 | 2016-05-25 | 上海通创信息技术有限公司 | Stream-oriented recommended engine, recommendation system and recommendation method based on clustering |
CN105956093A (en) * | 2016-04-29 | 2016-09-21 | 浙江大学 | Individual recommending method based on multi-view anchor graph Hash technology |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109634952B (en) * | 2018-11-02 | 2021-08-17 | 宁波大学 | Self-adaptive nearest neighbor query method for large-scale data |
CN109634952A (en) * | 2018-11-02 | 2019-04-16 | 宁波大学 | A kind of adaptive nearest neighbor querying method towards large-scale data |
CN109886787A (en) * | 2019-02-22 | 2019-06-14 | 清华大学 | Discrete social recommendation method and system |
CN110188825A (en) * | 2019-05-31 | 2019-08-30 | 山东师范大学 | Image clustering method, system, equipment and medium based on discrete multiple view cluster |
CN110188825B (en) * | 2019-05-31 | 2020-01-31 | 山东师范大学 | Image clustering method, system, device and medium based on discrete multi-view clustering |
CN111104604A (en) * | 2019-11-25 | 2020-05-05 | 北京交通大学 | Lightweight social recommendation method based on Hash learning |
CN111104604B (en) * | 2019-11-25 | 2023-07-21 | 北京交通大学 | Lightweight socialization recommendation method based on Hash learning |
CN111405070A (en) * | 2020-05-29 | 2020-07-10 | 成都信息工程大学 | Multi-core collaborative cross-chain method based on cloud manufacturing platform |
CN113010800A (en) * | 2021-03-04 | 2021-06-22 | 浙江大学 | Combined coding-based collaborative ranking recommendation method |
CN113010800B (en) * | 2021-03-04 | 2022-03-18 | 浙江大学 | Combined coding-based collaborative ranking recommendation method |
CN113158032A (en) * | 2021-03-18 | 2021-07-23 | 北京京东乾石科技有限公司 | Information pushing method and device |
CN113158032B (en) * | 2021-03-18 | 2024-03-01 | 北京京东乾石科技有限公司 | Information pushing method and device |
CN115309997A (en) * | 2022-10-10 | 2022-11-08 | 浙商银行股份有限公司 | Commodity recommendation method and device based on multi-view self-coding features |
CN116484970A (en) * | 2023-04-13 | 2023-07-25 | 南京大学 | Learning piece searching method based on anchor learning piece |
CN116484970B (en) * | 2023-04-13 | 2024-04-02 | 南京大学 | Learning piece searching method based on anchor learning piece |
CN117522532A (en) * | 2024-01-08 | 2024-02-06 | 浙江大学 | Popularity deviation correction recommendation method and device, electronic equipment and storage medium |
CN117522532B (en) * | 2024-01-08 | 2024-04-16 | 浙江大学 | Popularity deviation correction recommendation method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN107122411B (en) | 2020-08-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107122411A (en) | A kind of collaborative filtering recommending method based on discrete multi views Hash | |
CN105469096B (en) | A kind of characteristic bag image search method based on Hash binary-coding | |
CN106021364B (en) | Foundation, image searching method and the device of picture searching dependency prediction model | |
CN106682233A (en) | Method for Hash image retrieval based on deep learning and local feature fusion | |
CN110674407B (en) | Hybrid recommendation method based on graph convolution neural network | |
CN108304911A (en) | Knowledge Extraction Method and system based on Memory Neural Networks and equipment | |
CN106407352A (en) | Traffic image retrieval method based on depth learning | |
CN106802956A (en) | A kind of film based on weighting Heterogeneous Information network recommends method | |
CN105956093B (en) | A kind of personalized recommendation method based on multiple view anchor point figure Hash technology | |
CN104834748A (en) | Image retrieval method utilizing deep semantic to rank hash codes | |
CN104462196B (en) | Multiple features combining Hash information search method | |
CN106354735A (en) | Image target searching method and device | |
CN103810299A (en) | Image retrieval method on basis of multi-feature fusion | |
CN104298787A (en) | Individual recommendation method and device based on fusion strategy | |
CN105718960A (en) | Image ordering model based on convolutional neural network and spatial pyramid matching | |
CN103473327A (en) | Image retrieval method and image retrieval system | |
CN103778227A (en) | Method for screening useful images from retrieved images | |
CN104317838B (en) | Cross-media Hash index method based on coupling differential dictionary | |
CN102129470A (en) | Tag clustering method and system | |
CN104268629B (en) | Complex network community detecting method based on prior information and network inherent information | |
CN103186538A (en) | Image classification method, image classification device, image retrieval method and image retrieval device | |
CN111931505A (en) | Cross-language entity alignment method based on subgraph embedding | |
CN107291895B (en) | Quick hierarchical document query method | |
CN107833142A (en) | Academic social networks scientific research cooperative person recommends method | |
CN114358188A (en) | Feature extraction model processing method, feature extraction model processing device, sample retrieval method, sample retrieval device and computer equipment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |