CN113312542B - Processing method and device of object recommendation model and electronic equipment - Google Patents
Processing method and device of object recommendation model and electronic equipment Download PDFInfo
- Publication number
- CN113312542B CN113312542B CN202010120167.7A CN202010120167A CN113312542B CN 113312542 B CN113312542 B CN 113312542B CN 202010120167 A CN202010120167 A CN 202010120167A CN 113312542 B CN113312542 B CN 113312542B
- Authority
- CN
- China
- Prior art keywords
- target
- model
- object recommendation
- index set
- sub
- 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.)
- Active
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 55
- 238000012549 training Methods 0.000 claims abstract description 123
- 238000000034 method Methods 0.000 claims description 60
- 238000012545 processing Methods 0.000 claims description 48
- 230000006399 behavior Effects 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000004931 aggregating effect Effects 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 11
- 230000008569 process Effects 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 235000012054 meals Nutrition 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
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
- 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
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- 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
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Abstract
The application provides a processing method of an object recommendation model, which comprises the following steps: obtaining a real interference index set corresponding to a target user terminal; obtaining a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal; according to the target interference index set, obtaining a target object recommendation sub-model to be trained from a server; model training is carried out on the target object recommendation sub-model to be trained, and a target object recommendation sub-model to be uploaded is obtained; and providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set. According to the processing method of the object recommendation model, the object recommendation sub-model to be trained is obtained through the object interference index set, and the object recommendation sub-model to be uploaded is provided for the server, so that the efficiency of recommendation model training can be improved, and the complexity of recommendation model training can be reduced on the basis of ensuring the safety of user terminal data.
Description
Technical Field
The present application relates to the field of computer technology. The application relates to a processing method of an object recommendation model, and simultaneously relates to a processing device of the recommendation model, electronic equipment and a storage medium.
Background
With the rapid development of computing technology and internet technology, network platforms have also begun to emerge, wherein many network platforms tend to provide users with accurate relevant recommendations for better service and attraction, such as: the electronic commerce platform often provides accurate commodity recommendation for users, and the short video platform often provides accurate short video recommendation for users. When different network platforms provide accurate relevant recommendation for users, the accurate relevant recommendation is provided for the users by obtaining user data, and when an e-commerce platform is provided for providing accurate commodity recommendation, the e-commerce platform generally collects and obtains identity characteristic data, historical behavior data and the like of the users through a server to obtain commodities meeting user preferences as commodities recommended to the users. Because of the user's identity data, user history behavior data, etc., there is a high probability that the user's personal privacy is involved, resulting in that users sensitive to security privacy may refuse to share user data, and that any institution or company cannot upload and store user data without obtaining explicit permission from the eu user, as specified by the GDPR (General Data ProtectionRegulation, general data protection regulations) that is currently formally in force. Today, the task of providing accurate recommendations to users is very challenging and urgent to solve in situations where user data is not available.
The conventional common recommendation method for providing accurate recommendation for users under the condition that user data cannot be obtained is as follows: based on the joint learning framework, training a recommendation model, and providing accurate recommendation for a user according to the trained recommendation model. Specifically, the plurality of user terminals respectively download the global/complete model from the server, perform local model training under the coordination of the server, and submit the update of the complete model to the server, so that the recommendation model can be trained without uploading the original user data to the server by the user terminals. However, each user terminal needs to download the global/complete model for local training and submit the update of the complete model to the server, so that the complexity is high and the training efficiency of the recommended model is poor.
Disclosure of Invention
The application provides a processing method, a processing device, electronic equipment and a storage medium of an object recommendation model, so that the efficiency of recommendation model training is improved and the complexity of recommendation model training is reduced on the basis of ensuring the safety of user terminal data.
The method comprises the steps of obtaining a real interference index set corresponding to a target user terminal, wherein the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of a target object recommendation sub-model corresponding to the target object identification set in an object recommendation model;
obtaining a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to the union of the target real index set and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained, which comprises the target object recommendation sub-model, in the object recommendation model;
according to the target interference index set, obtaining a target object recommendation sub-model to be trained from a server;
model training is carried out on the target object recommendation sub-model to be trained, and a target object recommendation sub-model to be uploaded is obtained;
and providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set.
Optionally, the obtaining the real interference index set corresponding to the target ue includes:
obtaining a target object identification set corresponding to the target user terminal;
and obtaining the target real index set according to the target object identification set.
Optionally, the obtaining the target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal includes:
obtaining real index sets corresponding to the plurality of user terminals;
obtaining a union set of the target real index set and the real index sets corresponding to the plurality of user terminals according to the target real index set and the real index sets corresponding to the plurality of user terminals;
and obtaining the target interference index set according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals.
Optionally, the obtaining the target interference index set according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals includes:
obtaining a union set of the target real index set and real index sets corresponding to the plurality of user terminals;
generating a specified problem for the plurality of user terminals according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals;
Obtaining random answers of the plurality of user terminals to the specified questions;
and obtaining the target interference index set according to random answers of the plurality of user terminals to the specified questions.
Optionally, the obtaining, according to the target real index set and the real index sets corresponding to the plurality of user terminals, a union set of the target real index set and the real index sets corresponding to the plurality of user terminals includes:
according to the target real index set and the real index sets corresponding to the plurality of user terminals, a bloom counter corresponding to the target real index set is obtained, and bloom counters corresponding to the plurality of user terminals are obtained;
and performing set union calculation on bloom counters corresponding to the target real index set and bloom counters corresponding to the plurality of user terminals to obtain a union of the target real index set and real index sets corresponding to the plurality of user terminals.
Optionally, the obtaining, from the server, the target object recommendation sub-model to be trained according to the target interference index set includes:
obtaining an object identification set corresponding to the target interference index set;
obtaining the position of the object identification set corresponding to the target interference index set in the object recommendation model according to the object identification set corresponding to the target interference index set;
And obtaining the target object recommendation sub-model to be trained from a server according to the position of the object identification set corresponding to the target interference index set in the object recommendation model.
Optionally, the performing model training on the target object recommendation sub-model to be trained to obtain a target object recommendation sub-model to be uploaded includes:
obtaining a target succinct index set corresponding to the target user terminal according to the target real index set, wherein the target succinct index set is an index set generated according to the union of the target real index set and the real index sets corresponding to the plurality of user terminals and the intersection of the target real index set, and is used for indicating the position of the target object recommendation sub-model in the target object recommendation sub-model to be trained;
according to the target succinct index set, obtaining a target object recommendation sub-model from the target object recommendation sub-model to be trained;
acquiring target training data for training the target object recommendation sub-model according to the target succinct index set;
according to the target training data, performing model training on the target object recommendation sub-model to obtain a target object recommendation sub-model to be updated;
And replacing the target object recommendation sub-model in the target object recommendation sub-model to be trained with the target object recommendation sub-model to be updated according to the target succinct index set, and obtaining the target object recommendation sub-model to be uploaded.
Optionally, the obtaining, according to the target real index set, a target succinct index set corresponding to the target user terminal includes:
obtaining a union set of the target real index set and real index sets corresponding to the plurality of user terminals and an intersection set of the union set and the target real index set;
and taking the union set of the target real index set and the real index sets corresponding to the plurality of user terminals and the intersection set of the target real index set as the target succinct index set.
Optionally, the obtaining, according to the target succinct index set, the target object recommendation sub-model from the target object recommendation sub-model to be trained includes: and obtaining the target object recommendation sub-model from the target object recommendation sub-model to be trained according to the position of the object identification set corresponding to the target succinct index set in the target object recommendation sub-model to be trained.
Optionally, the obtaining, according to the target succinct index set, target training data for training the target object recommendation sub-model includes:
Obtaining user behavior data related to the object identification set corresponding to the target succinct index set according to the object identification set corresponding to the target succinct index set;
and taking the user behavior data related to the object identification set corresponding to the target succinct index set as target training data.
Optionally, performing model training on the target object recommendation sub-model according to the target training data to obtain a target object recommendation sub-model to be updated, including: and carrying out model training by taking the target training data and the target object recommendation sub-model as input to obtain the target object recommendation sub-model to be updated.
Optionally, replacing the target object recommendation sub-model in the target object recommendation sub-model to be trained with the target object recommendation sub-model to be updated according to the target succinct index set, to obtain the target object recommendation sub-model to be uploaded includes: and replacing the target object recommendation sub-model in the target object recommendation sub-model to be updated with the target object recommendation sub-model to be updated according to the position of the object identification set corresponding to the target succinct index set in the target object recommendation sub-model to be trained, so as to obtain the target object recommendation sub-model to be uploaded.
Optionally, the method further comprises:
according to the target succinct index set, an object identification set corresponding to the target succinct index set is obtained;
and obtaining the position of the object identification set corresponding to the target succinct index set in the target object recommendation sub-model to be trained according to the object identification set corresponding to the target succinct index set.
Optionally, the providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set includes:
obtaining the position of the target object recommendation sub-model to be uploaded in the object recommendation model according to the target interference index set;
and providing the target object recommendation sub-model to be uploaded for the server side according to the position of the target object recommendation sub-model to be uploaded in the object recommendation model.
Optionally, the method further comprises: acquiring a request message which is sent by the server and is used for requesting a training object recommendation sub-model;
the providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set includes: and for the request message, providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set.
In another aspect of the present application, there is provided a processing apparatus for an object recommendation model, including:
the real interference index set obtaining unit is used for obtaining a real interference index set corresponding to a target user terminal, wherein the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of a target object recommendation sub-model corresponding to the target object identification set in the object recommendation model;
a real interference index set obtaining unit, configured to obtain a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal, where the target interference index set is an index set generated according to a union set of the target real index set and real index sets corresponding to a plurality of user terminals, and is configured to indicate a position of a target object recommendation sub-model to be trained including the target object recommendation sub-model in the object recommendation model;
the object recommendation sub-model obtaining unit is used for obtaining the object recommendation sub-model to be trained from the server according to the object interference index set;
the object recommendation sub-model to be uploaded is used for carrying out model training on the target object recommendation sub-model to be trained to obtain the target object recommendation sub-model to be uploaded;
And the object recommendation sub-model to be uploaded is provided for providing the object recommendation sub-model to be uploaded to the server according to the target interference index set.
In another aspect of the present application, there is provided an electronic device, including:
a processor; and
a memory for storing a program of a processing method for an object recommendation model, the apparatus being powered on and executing the program of the processing method for an object recommendation model by the processor, and executing the steps of:
obtaining a real interference index set corresponding to a target user terminal, wherein the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of a target object recommendation sub-model corresponding to the target object identification set in the object recommendation model;
obtaining a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to the union of the target real index set and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained, which comprises the target object recommendation sub-model, in the object recommendation model;
According to the target interference index set, obtaining a target object recommendation sub-model to be trained from a server;
model training is carried out on the target object recommendation sub-model to be trained, and a target object recommendation sub-model to be uploaded is obtained;
and providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set.
In another aspect of the present application, there is provided a storage medium storing a program of a processing method for an object recommendation model, the program being executed by a processor to perform the steps of:
obtaining a real interference index set corresponding to a target user terminal, wherein the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of a target object recommendation sub-model corresponding to the target object identification set in the object recommendation model;
obtaining a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to the union of the target real index set and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained, which comprises the target object recommendation sub-model, in the object recommendation model;
According to the target interference index set, obtaining a target object recommendation sub-model to be trained from a server;
model training is carried out on the target object recommendation sub-model to be trained, and a target object recommendation sub-model to be uploaded is obtained;
and providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set.
In another aspect of the present application, a method for processing an object recommendation model is provided, including:
obtaining a target object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained comprising the target object recommendation sub-model in the object recommendation model, the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of the target object recommendation sub-model corresponding to the target object identification set in the object recommendation model, and the target object recommendation sub-model to be uploaded is an object sub-model obtained after model training is carried out on the target object recommendation sub-model by the target user terminal;
And obtaining a target object recommendation model according to the target object recommendation sub-model to be uploaded, wherein the target object recommendation model is used for recommending the object for the user terminal according to the object identification set corresponding to the user terminal.
Optionally, the method further comprises: obtaining object recommendation sub-models to be uploaded corresponding to a plurality of user terminals, wherein the object recommendation sub-models are uploaded for the plurality of user terminals according to interference index sets corresponding to the plurality of user terminals;
the obtaining the target object recommendation model according to the target object recommendation sub-model to be uploaded comprises the following steps:
acquiring weight values of the target object recommendation sub-model to be uploaded and the object recommendation sub-models to be uploaded corresponding to the plurality of user terminals;
and aggregating the target object recommendation sub-model to be uploaded and the object recommendation sub-model to be uploaded corresponding to the plurality of user terminals according to the weight values of the target object recommendation sub-model to be uploaded and the object recommendation sub-models to be uploaded corresponding to the plurality of user terminals, so as to obtain the target object recommendation model.
Optionally, the method further comprises: transmitting a request message for requesting a training object recommendation sub-model to the target user terminal;
The obtaining the target object recommendation sub-model to be uploaded, which is uploaded by the target user terminal according to the target interference index set corresponding to the target user terminal, includes: and aiming at the request message, obtaining a target object recommendation sub-model to be uploaded, which is uploaded by the target user terminal according to the target interference index set corresponding to the target user terminal.
Optionally, the method further comprises:
obtaining an object identification set corresponding to a user terminal to be recommended;
and inputting the object identification set corresponding to the user terminal to be recommended into the target object recommendation model to obtain a recommended object recommended for the user terminal to be recommended.
In another aspect of the present application, there is provided a processing apparatus for an object recommendation model, including:
the object recommendation sub-model to be uploaded is used for obtaining an object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals, the index set is used for indicating the position of a target object recommendation sub-model to be trained containing the object recommendation sub-model in the object recommendation model, the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal, the index set is used for indicating the position of the target object recommendation sub-model corresponding to the target object identification set in the object recommendation model, and the object recommendation sub-model to be uploaded is an object sub-model obtained after model training is carried out on the target object recommendation sub-model by the target user terminal;
The target object recommendation model obtaining unit is used for obtaining a target object recommendation model according to the target object recommendation sub-model to be uploaded, and the target object recommendation model is used for carrying out object recommendation on the user terminal according to an object identification set corresponding to the user terminal.
In another aspect of the present application, there is provided an electronic device, including:
a processor; and
a memory for storing a program of a processing method for an object recommendation model, the apparatus being powered on and executing the program of the processing method for an object recommendation model by the processor, and executing the steps of:
obtaining a target object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained comprising the target object recommendation sub-model in the object recommendation model, the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of the target object recommendation sub-model corresponding to the target object identification set in the object recommendation model, and the target object recommendation sub-model to be uploaded is an object sub-model obtained after model training is carried out on the target object recommendation sub-model by the target user terminal;
And obtaining a target object recommendation model according to the target object recommendation sub-model to be uploaded, wherein the target object recommendation model is used for recommending the object for the user terminal according to the object identification set corresponding to the user terminal.
In another aspect of the present application, there is provided a storage medium storing a program of a processing method for an object recommendation model, the program being executed by a processor to perform the steps of:
obtaining a target object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained comprising the target object recommendation sub-model in the object recommendation model, the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of the target object recommendation sub-model corresponding to the target object identification set in the object recommendation model, and the target object recommendation sub-model to be uploaded is an object sub-model obtained after model training is carried out on the target object recommendation sub-model by the target user terminal;
And obtaining a target object recommendation model according to the target object recommendation sub-model to be uploaded, wherein the target object recommendation model is used for recommending the object for the user terminal according to the object identification set corresponding to the user terminal.
Compared with the prior art, the application has the following advantages:
according to the processing method of the object recommendation model, firstly, a real interference index set for indicating the position of a target object recommendation sub-model corresponding to a target object identification set in the object recommendation model is obtained, and further, according to the obtained real interference index set, a target interference index set for indicating the position of a target object recommendation sub-model to be trained containing the target object recommendation sub-model in the object recommendation model is obtained; secondly, according to the target interference index set, a target object recommendation sub-model to be trained is obtained from the server; thirdly, performing model training on the target object recommendation sub-model to be trained to obtain a target object recommendation sub-model to be uploaded; and finally, providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set. According to the object recommendation model processing method, the object recommendation sub-model to be uploaded is obtained from the server to be subjected to model training through the index set generated according to the union of the target real index set and the real index sets corresponding to the plurality of user terminals, and the object recommendation sub-model to be uploaded is provided for the server, so that the recommendation model training efficiency is improved and the recommendation model training complexity is reduced on the basis of ensuring the safety of user terminal data.
Drawings
Fig. 1 is a first schematic diagram of an application scenario of a method for processing an object recommendation model provided in the present application.
Fig. 1A is a second schematic diagram of an application scenario of a method for processing an object recommendation model provided in the present application.
Fig. 2 is a flowchart of a processing method of an object recommendation model according to a first embodiment of the present application.
Fig. 3 is a flowchart of a method for obtaining a target interference index set according to the first embodiment of the present application.
Fig. 4 is a flowchart of a method for obtaining a recommendation sub-model of a target object to be uploaded according to a first embodiment of the present application.
Fig. 5 is a schematic diagram of a processing device of a recommendation model according to a second embodiment of the present application.
Fig. 6 is a schematic diagram of an electronic device provided in an embodiment of the present application.
Fig. 7 is a flowchart of a processing method of an object recommendation model according to a fifth embodiment of the present application.
Fig. 8 is a schematic diagram of a processing device of a recommendation model according to a sixth embodiment of the present application.
Detailed Description
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. This application is, however, susceptible of embodiment in many other ways than those herein described and similar generalizations can be made by those skilled in the art without departing from the spirit of the application and the application is therefore not limited to the specific embodiments disclosed below.
In order to more clearly show the processing method of the object recommendation model provided by the application, an application scenario of the processing method of the object recommendation model provided by the application is introduced.
The method for processing the object recommendation model is generally applied to training a scene of the object recommendation model for a commodity recommendation model or training a scene of the object recommendation model for a short view screen, and is specifically described based on the object recommendation model being the commodity recommendation model. Fig. 1 is a first schematic diagram of an application scenario of a processing method of an object recommendation model provided in the present application.
The server 101 selects a plurality of user terminals meeting the preset rules to participate in the merchandise recommendation model training, and each selected user terminal is called a target user terminal 102. Specifically, in an application scenario of the method for processing an object recommendation model provided in the present application, a user terminal meeting a preset rule generally needs to meet the following states: the user terminal is in a standby state, the user terminal is in a wireless network connection state, the user terminal is in a charging state, and the like. The user terminal is typically a smart phone, a PC (Personal Computer ) terminal, a tablet PC, etc., and is guaranteed to be in a standby state to prevent normal use of the user from being affected, a wireless network connection state to prevent the user terminal from consuming network data for use, and a charging state to prevent battery life of the user terminal from being affected.
After selecting the target user terminal 102, the server 101 sends a request message to the target user terminal 102 for training the commodity recommendation sub-model. After obtaining the request message sent by the server 101 for requesting training of the commodity recommendation sub-model, the target user terminal 102 first downloads the target commodity recommendation sub-model 101-2 belonging to the commodity recommendation model 101-1 from the server 101 for training, and obtains the target commodity recommendation sub-model 101-3 to be uploaded. The target commodity recommendation sub-model 101-2 includes a sub-model of a target commodity recommendation sub-model corresponding to the target commodity identification set. Then, the target commodity recommendation sub-model 101-3 to be uploaded is provided to the server 101, and the server 101 performs security aggregation according to the target commodity recommendation sub-model 101-3 to be uploaded and a plurality of commodity recommendation sub-models to be uploaded provided by a plurality of other target user terminals, and a target commodity recommendation model 101-4 is obtained by aiming at the target commodity recommendation sub-model 101-3 to be uploaded and a plurality of commodity recommendation sub-models to be uploaded provided by a plurality of other target user terminals.
In order to ensure the security of the target user terminal data, in the application scenario of the processing method of the object recommendation model provided in the present application, the target user terminal needs to process according to the following steps when downloading and providing the related sub-model, and specifically refer to fig. 1A, which is a second schematic diagram of the application scenario of the processing method of the object recommendation model provided in the present application.
Step S101: and obtaining a target commodity identification set corresponding to the target user terminal.
After obtaining a request message sent by a server side and requesting training of a commodity recommendation sub-model, the target user terminal obtains a target commodity identification set corresponding to the target user terminal according to the request message. The target commodity identification set is used for indicating various expressions and indications of commodity information, and the commodity identification set is data extracted from local user behavior data in the target user terminal.
Step S102: and obtaining a target real index set according to the target commodity identification set.
The target real index set is an index set generated according to a target commodity identification set corresponding to the target user terminal and is used for indicating the position of a target commodity recommendation sub-model corresponding to the target commodity identification set in the commodity recommendation model.
Step S103: and obtaining a target interference index set according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals.
The target interference index set is an index set generated according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals and is used for indicating the position of the target commodity recommendation sub-model to be trained, which comprises the target commodity recommendation sub-model, in the commodity recommendation model. The plurality of other target user terminals are other target user terminals which are selected by the server and participate in the same commodity recommendation model training with the target user terminals.
In order to obtain the target interference index set according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals, the real index sets corresponding to the plurality of other target user terminals need to be obtained first, and then the union set of the target real index set and the real index sets corresponding to the plurality of user terminals is obtained according to the target real index set and the real index sets corresponding to the plurality of user terminals. Specifically, according to a target real index set and real index sets corresponding to a plurality of user terminals, a bloom counter corresponding to the target real index set is obtained, and bloom counters corresponding to the plurality of user terminals are obtained; and performing set union calculation on the bloom counter corresponding to the target real index set and bloom counters corresponding to the plurality of user terminals to obtain a union of the target real index set and the real index sets corresponding to the plurality of user terminals. The bloom filter is a bit vector, a plurality of hash functions are adopted to map elements of a target real index set corresponding to a sparse target interference index set, and the counting bloom filter adopts a plurality of hash functions to map elements of a union set of the target real index set corresponding to the sparse target interference index set and the target real index set corresponding to a plurality of other target user terminals.
Obtaining a target interference index set according to a union set of a target real index set and real index sets corresponding to a plurality of user terminals, wherein the method comprises the following steps: first, a union of a target real index set and real index sets corresponding to a plurality of user terminals is obtained. And secondly, generating specified problems aiming at a plurality of user terminals according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals. Again, a random answer to the specified question is obtained for the plurality of user terminals. And finally, obtaining a target interference index set according to random answers of the plurality of user terminals to the specified questions. Specifically, the specified problem is two times, and the two problems are: "do you own a certain index? The index range aimed at by the problem is the union of the real index sets corresponding to all target user terminals which participate in the same round of commodity recommendation model training when the server is selected. When some other target user terminal answers yes with probability 1 and answers no with probability 2 for the first random answer, the target user terminal needs to record an index memo, that is, needs to record an index set corresponding to all answers yes and an index set corresponding to answers no in the first random answer. When some other target user terminal carries out random answer for the second time aiming at the appointed question, if the index aimed at the appointed question is marked as 'yes' in the memo, putting the index into an interference index set with probability of 3; otherwise, the interference index set is put in the probability 4.
Step S104: and obtaining a target commodity recommendation sub-model to be trained from the server according to the target interference index set.
According to the target interference index set, the process of obtaining the target commodity recommendation sub-model to be trained from the server side comprises the following steps: first, a commodity identification set corresponding to a target interference index set is obtained. And then, according to the commodity identification set corresponding to the target interference index set, obtaining the position of the commodity identification set corresponding to the target interference index set in the commodity recommendation model. And finally, according to the position of the commodity identification set corresponding to the target interference index set in the commodity recommendation model, acquiring the target commodity recommendation sub-model to be trained from the server according to the target interference index set from the server.
Step S105: and carrying out model training on the target commodity recommendation sub-model to be trained to obtain the target commodity recommendation sub-model to be uploaded.
Model training is carried out on the target commodity recommendation sub-model to be trained to obtain the target commodity recommendation sub-model to be uploaded, and the method comprises the following steps: obtaining a target succinct index set corresponding to a target user terminal according to a target real index set, wherein the target succinct index set is an index set generated according to a union set of the target real index set and real index sets corresponding to a plurality of user terminals and an intersection set of the target real index set, and is used for indicating the position of a target commodity recommendation sub-model in a target commodity recommendation sub-model to be trained; obtaining a target commodity recommendation sub-model from the target commodity recommendation sub-model to be trained according to the target succinct index set; acquiring target training data for training a target commodity recommendation sub-model according to the target succinct index set; according to the target training data, performing model training on the target commodity recommendation sub-model to obtain a target commodity recommendation sub-model to be updated; and replacing the target commodity recommendation sub-model in the target commodity recommendation sub-model to be trained with the target commodity recommendation sub-model to be updated according to the target succinct index set, and obtaining the target commodity recommendation sub-model to be uploaded.
Step S106: and providing the target commodity recommendation sub-model to be uploaded to the server according to the target interference index set.
Providing the target commodity recommendation sub-model to be uploaded to the server according to the target interference index set, wherein the method comprises the following steps: obtaining the position of a target commodity recommendation sub-model to be uploaded in a commodity recommendation model according to the target interference index set; and providing the target commodity recommendation sub-model to be uploaded for the server side aiming at the position of the target commodity recommendation sub-model to be uploaded in the commodity recommendation model.
In an application scenario of the processing method of the object recommendation model, a server obtains a target commodity recommendation model according to a target commodity recommendation sub-model to be uploaded and a plurality of commodity recommendation sub-models to be uploaded provided by a plurality of other target user terminals, wherein the mode of obtaining the target commodity recommendation model is as follows: acquiring weight values of a target commodity recommendation sub-model to be uploaded and commodity recommendation sub-models to be uploaded corresponding to a plurality of user terminals; and carrying out weighted average on the target commodity recommendation sub-model to be uploaded and the commodity recommendation sub-models to be uploaded corresponding to the plurality of user terminals according to the weight values of the target commodity recommendation sub-model to be uploaded and the commodity recommendation sub-models to be uploaded corresponding to the plurality of user terminals, so as to obtain the target commodity recommendation model.
After the server obtains the target commodity recommendation model, a commodity identification set corresponding to the user terminal to be recommended can be further obtained; and acquiring recommended commodities recommended for the user terminal to be recommended according to the commodity identification set corresponding to the user terminal to be recommended.
It should be noted that, because the target real index set is an index set generated according to the target commodity identification set corresponding to the target user terminal and is used for indicating the position of the target commodity recommendation sub-model corresponding to the target commodity identification set in the commodity recommendation model, in the application scenario of the processing method of the object recommendation model provided by the application, the target commodity recommendation sub-model can also be obtained directly through the target real index set to train, the commodity recommendation sub-model to be uploaded is obtained, and the commodity recommendation sub-model to be uploaded is provided to the server according to the target real index set. However, since the target real index set is an index set generated according to the target commodity identification set corresponding to the target user terminal, training is performed by directly obtaining the target commodity recommendation sub-model through the target real index set, and providing the commodity recommendation sub-model to be uploaded to the server, the related personnel can often obtain the target commodity recommendation sub-model or the commodity recommendation sub-model to be uploaded from the server, and the commodity identification set corresponding to the target real index set is reversely deduced, so that the leakage of target user data is caused.
In summary, in the application scenario of the processing method of the object recommendation model provided by the present application, if the target user data already has better confidentiality, or in some application fields, when the target user does not require confidentiality of the user data, the target commodity recommendation sub-model can be obtained for training by the selection of the user through the target real index set, and the commodity recommendation sub-model to be uploaded is provided to the server; and obtaining a target object recommendation sub-model to be trained from the server through the target interference index set, and providing the target object recommendation sub-model to be uploaded to the server. Furthermore, in an application scenario of the method for processing an object recommendation model provided by the present application, two different merchandise recommendation model training modes may be further deduced, where the first mode is: the convenient mode is a mode that a target commodity recommendation sub-model can be obtained directly through a target real index set for training and the commodity recommendation sub-model to be uploaded is provided for a server; the second is: and in the safety mode, firstly, a target object recommendation sub-model to be trained is obtained from the server through a target interference index set, and the target object recommendation sub-model to be uploaded is provided for the server mode. The relevant user can select different modes according to specific application fields and scenes.
In the application scenario of the method for processing the object recommendation model provided by the present application, the object to be targeted may be a short video, a meal, or the like, in addition to being a commodity. The application scenario of the processing method of the object recommendation model provided in the present application is only one embodiment of the application scenario of the processing method of the object recommendation model provided in the present application, and the purpose of the application scenario embodiment is to facilitate understanding of the processing method of the object recommendation model provided in the present application, and is not used to limit the processing method of the object recommendation model provided in the present application. The processing method of the object recommendation model provided by the application can also be applied to other application scenes, and is not described in detail herein.
First embodiment
A method for processing an object recommendation model is provided in a first embodiment of the present application, and is described below with reference to fig. 2 to 4.
Fig. 2 is a flowchart of a processing method of an object recommendation model according to a first embodiment of the present application.
In step S201, a real interference index set corresponding to a target user terminal is obtained, where the target real index set is an index set generated according to a target object identifier set corresponding to the target user terminal, and is used to indicate a position of a target object recommendation sub-model corresponding to the target object identifier set in the object recommendation model.
In the first embodiment of the present application, the objects are generally commodities, dining products, short videos, and the like, and the corresponding target object identification set is generally a target commodity identification set, a dining product identification set target short video identification set, and the like. Specifically, taking a target commodity label set as an example, the target commodity label set is used for indicating various expressions and indications of target commodity information, and the target commodity label set is data extracted from target user behavior data related to user behavior data in a target user terminal.
In the first embodiment of the present application, the target user terminal performs the object recommendation sub-model generally by performing the step after obtaining the request message sent by the obtaining server for training the object recommendation sub-model. The specific process of the real interference index set corresponding to the target user terminal is as follows: obtaining a request message which is sent by the server and is used for requesting training of an object recommendation sub-model, obtaining a target object identification set corresponding to the target user terminal aiming at the request message, and obtaining the target real index set according to the target object identification set.
The target real index set is obtained according to a target object identification set corresponding to the target user terminal, and the target real index set relates to user privacy because the data is extracted from target user behavior data related to the user behavior data in the target user terminal. Therefore, although the method is used for indicating the position of the target object recommendation sub-model corresponding to the target object identification set in the object recommendation model, in order to maintain the data security of the user, in the first embodiment of the present application, the target object identification set is not directly used to obtain the target object recommendation sub-model.
Taking the example that the target object is a target commodity, in the first embodiment of the present application, the target user terminal is typically a smart phone, a PC end, a tablet, and the like, which are installed with an e-commerce platform client. The service end is generally an e-commerce platform server end. The target user terminal is to select a plurality of user terminals meeting preset rules by the server, and each selected user terminal becomes the target user terminal. The user terminal according with the preset rule generally needs to meet the following states: the user terminal is in a standby state, the user terminal is in a wireless network connection state, the user terminal is in a charging state, and the like.
In step S202, according to the obtained real interference index set corresponding to the target user terminal, a target interference index set corresponding to the target user terminal is obtained, where the target interference index set is an index set generated according to a union set of the target real index set and the real index sets corresponding to the plurality of user terminals.
The target interference index set is used for indicating the position of a target object recommendation sub-model to be trained, which comprises the target object recommendation sub-model, in the object recommendation model. In the first embodiment of the present application, fig. 3 is a flowchart of a method for obtaining a target interference index set provided in the first embodiment of the present application, where the process of obtaining the target interference index set corresponding to the target user terminal is shown in the drawing.
Step S202-1: and obtaining the real index sets corresponding to the plurality of user terminals.
The plurality of user terminals are other target user terminals which are selected by the server and participate in the same round of object recommendation model training with the target user terminals.
Step S202-2: and obtaining a union set of the target real index set and the real index sets corresponding to the plurality of user terminals according to the target real index set and the real index sets corresponding to the plurality of user terminals.
In the first embodiment of the present application, according to a target real index set and real index sets corresponding to a plurality of user terminals, a process of obtaining a union set of the target real index set and the real index sets corresponding to the plurality of user terminals is: firstly, according to a target real index set and real index sets corresponding to a plurality of user terminals, a bloom counter corresponding to the target real index set is obtained, and bloom counters corresponding to the plurality of user terminals are obtained. And then, carrying out set union calculation on the bloom counter corresponding to the target real index set and bloom counters corresponding to the plurality of user terminals to obtain the union of the target real index set and the real index sets corresponding to the plurality of user terminals. The bloom filter is a bit vector, a plurality of hash functions are adopted to map elements of a target real index set corresponding to a sparse target interference index set, and the counting bloom filter adopts a plurality of hash functions to map elements of a union set of the target real index set corresponding to the sparse target interference index set and the target real index set corresponding to a plurality of other target user terminals.
Step S202-3: and obtaining a target interference index set according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals.
In a first embodiment of the present application, obtaining a target interference index set according to a union set of a target real index set and real index sets corresponding to a plurality of user terminals includes: obtaining a union set of a target real index set and real index sets corresponding to a plurality of user terminals; generating a specified problem for a plurality of user terminals according to a union set of a target real index set and real index sets corresponding to the plurality of user terminals; obtaining random answers of a plurality of user terminals aiming at specified questions; and obtaining a target interference index set according to random answers of the plurality of user terminals to the specified questions.
In the first embodiment of the present application, the assignment problem is two times, and the two problems are: "do you own a certain index? The index range aimed at by the problem is the union of the real index sets corresponding to all target user terminals which participate in the same round of commodity recommendation model training when the server is selected. When some other target user terminal answers yes with probability 1 and answers no with probability 2 for the first random answer, the target user terminal needs to record an index memo, that is, needs to record an index set corresponding to all answers yes and an index set corresponding to answers no in the first random answer. When some other target user terminal carries out random answer for the second time aiming at the appointed question, if the index aimed at the appointed question is marked as 'yes' in the memo, putting the index into an interference index set with probability of 3; otherwise, the interference index set is put in the probability 4.
In step S203, a target object recommendation sub-model to be trained is obtained from the server according to the target interference index set.
In the first embodiment of the present application, according to a target interference index set, the steps of obtaining a target object recommendation sub-model to be trained from a server are as follows: firstly, obtaining an object identification set corresponding to a target interference index set; then, according to the object identification set corresponding to the target interference index set, the position of the object identification set corresponding to the target interference index set in the object recommendation model is obtained; and finally, obtaining a target object recommendation sub-model to be trained from the server according to the position of the object identification set corresponding to the target interference index set in the object recommendation model. Specifically, obtaining the position of the object identification set corresponding to the target interference index set in the object recommendation model includes: obtaining a parameter matrix set related to an object identification set corresponding to a target interference index set; correspondingly, according to the position of the object identification set corresponding to the target interference index set in the object recommendation model, obtaining a target object recommendation sub-model to be trained from the server side comprises: according to the parameter matrix set related to the object identification set corresponding to the target interference index set, the position of the parameter matrix set related to the object identification set corresponding to the target interference index set in the parameter matrix set related to the object recommendation model is obtained, and the target object recommendation sub-model to be trained is obtained from the server.
In step S204, model training is performed on the target object recommendation sub-model to be trained, and the target object recommendation sub-model to be uploaded is obtained.
In the first embodiment of the present application, a model training is performed for a target object recommendation sub-model to be trained, and fig. 4 is referred to in the process of obtaining a target object recommendation sub-model to be uploaded, which is a flowchart of a method for obtaining a target object recommendation sub-model to be uploaded provided in the first embodiment of the present application.
Step S204-1: and obtaining a target succinct index set corresponding to the target user terminal according to the target real index set, wherein the target succinct index set is an index set generated according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals and the intersection set of the target real index set.
In the first embodiment of the present application, the target succinct index set is used to indicate a position of the target object recommendation sub-model in the target object recommendation sub-model to be trained. The process of obtaining the target succinct index set corresponding to the target user terminal comprises the following steps: obtaining a union set of a target real index set and real index sets corresponding to a plurality of user terminals and an intersection set of the union set and the target real index set; and taking the union set of the target real index set and the real index sets corresponding to the plurality of user terminals and the intersection set of the target real index set as target succinct index sets.
Step S204-2: and obtaining a target object recommendation sub-model from the target object recommendation sub-model to be trained according to the target succinct index set.
After the target succinct index set is obtained, in the first embodiment of the present application, it is further required to obtain an object identifier set corresponding to the target succinct index set according to the target succinct index set, and obtain a position of the object identifier set corresponding to the target succinct index set in the target object recommendation sub-model to be trained according to the object identifier set corresponding to the target succinct index set. At this time, according to the target succinct index set, a target object recommendation sub-model is obtained from the target object recommendation sub-model to be trained, including: and obtaining a target object recommendation sub-model from the target object recommendation sub-model to be trained according to the position of the object identification set corresponding to the target succinct index set in the target object recommendation sub-model to be trained.
Step S204-3: acquiring target training data for training a target object recommendation sub-model according to the target succinct index set;
after the target succinct index set is obtained, in the first embodiment of the present application, it is further required to obtain an object identifier set corresponding to the target succinct index set according to the target succinct index set, and obtain a position of the object identifier set corresponding to the target succinct index set in the target object recommendation sub-model to be trained according to the object identifier set corresponding to the target succinct index set. At this time, according to the target compact index set, the step of obtaining target training data for training the target object recommendation sub-model is as follows: firstly, according to an object identification set corresponding to a target succinct index set, user behavior data related to the object identification set corresponding to the target succinct index set is obtained. Then, user behavior data related to the object identification set corresponding to the target succinct index set is used as target training data.
Step S204-4: and performing model training on the target object recommendation sub-model according to the target training data to obtain the target object recommendation sub-model to be updated.
In a first embodiment of the present application, according to target training data, performing model training on a target object recommendation sub-model to obtain a target object recommendation sub-model to be updated, including: and taking the target training data and the target object recommendation sub-model as input to carry out model training, and obtaining the target object recommendation sub-model to be updated.
Step S204-5: and replacing the target object recommendation sub-model in the target object recommendation sub-model to be trained with the target object recommendation sub-model to be updated according to the target succinct index set, and obtaining the target object recommendation sub-model to be uploaded.
After the target succinct index set is obtained, in the first embodiment of the present application, it is further required to obtain an object identifier set corresponding to the target succinct index set according to the target succinct index set, and obtain a position of the object identifier set corresponding to the target succinct index set in the target object recommendation sub-model to be trained according to the object identifier set corresponding to the target succinct index set. At this time, according to the target succinct index set, replacing the target object recommendation sub-model in the target object recommendation sub-model to be trained with the target object recommendation sub-model to be updated to obtain the target object recommendation sub-model to be uploaded, including: and replacing the target object recommendation sub-model in the target object recommendation sub-model to be updated with the target object recommendation sub-model to be updated according to the position of the object identification set corresponding to the target succinct index set in the target object recommendation sub-model to be trained, so as to obtain the target object recommendation sub-model to be uploaded.
In step S205, the target object recommendation sub-model to be uploaded is provided to the server according to the target interference index set.
In a first embodiment of the present application, providing a target object recommendation sub-model to be uploaded to a server according to a target interference index set includes: obtaining the position of a target object recommendation sub-model to be uploaded in an object recommendation model according to a target interference index set; and providing the target object recommendation sub-model to be uploaded for the server side according to the position of the target object recommendation sub-model to be uploaded in the object recommendation model.
In the first embodiment of the present application, the object recommendation sub-model is generally performed after obtaining the request message sent by the server for requesting training of the object recommendation sub-model, so according to the target interference index set, the providing the target object recommendation sub-model to be uploaded to the server specifically includes: and for the request message, providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set.
According to the processing method of the object recommendation model, firstly, a real interference index set for indicating the position of a target object recommendation sub-model corresponding to a target object identification set in the object recommendation model is obtained, and further, according to the obtained real interference index set, a target interference index set for indicating the position of a target object recommendation sub-model to be trained containing the target object recommendation sub-model in the object recommendation model is obtained; secondly, according to the target interference index set, a target object recommendation sub-model to be trained is obtained from the server; thirdly, performing model training on the target object recommendation sub-model to be trained to obtain a target object recommendation sub-model to be uploaded; and finally, providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set. According to the processing method of the object recommendation model, the target object recommendation sub-model to be trained is obtained from the server to conduct model training according to the index set generated by the union set of the target real index set and the real index sets corresponding to the plurality of user terminals, the target object recommendation sub-model to be uploaded is obtained, and the target object recommendation sub-model to be uploaded is provided for the server, so that the efficiency of recommendation model training can be improved and the complexity of recommendation model training can be reduced on the basis of ensuring the safety of user terminal data.
Second embodiment
Corresponding to the processing method of the recommendation model provided in the first embodiment of the present application, the second embodiment of the present application provides a processing device of the recommendation model. Since the apparatus embodiment is substantially similar to the first embodiment of the method, the description is relatively simple, and reference is made to the description of the method embodiment for relevant points. The device embodiments described below are merely illustrative.
Fig. 5 is a schematic diagram of a processing device of a recommendation model according to a second embodiment of the present application.
The processing device of the recommendation model comprises:
a real interference index set obtaining unit 501, configured to obtain a real interference index set corresponding to a target user terminal, where the target real index set is an index set generated according to a target object identifier set corresponding to the target user terminal, and is configured to indicate a position of a target object recommendation sub-model corresponding to the target object identifier set in the object recommendation model;
a real interference index set obtaining unit 502, configured to obtain a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal, where the target interference index set is an index set generated according to a union set of the target real index set and real index sets corresponding to a plurality of user terminals, and is configured to indicate a position of a target object recommendation sub-model to be trained including the target object recommendation sub-model in the object recommendation model;
An object to be trained recommending sub-model obtaining unit 503, configured to obtain the object to be trained recommending sub-model from a server according to the target interference index set;
the object recommendation sub-model to be uploaded obtaining unit 504 is configured to perform model training on the target object recommendation sub-model to be trained to obtain a target object recommendation sub-model to be uploaded;
and the object recommendation sub-model to be uploaded providing unit 505 is configured to provide the object recommendation sub-model to be uploaded to the server according to the target interference index set.
Optionally, the real interference index set obtaining unit 501 is specifically configured to obtain a target object identifier set corresponding to the target user terminal; and obtaining the target real index set according to the target object identification set.
Optionally, the real interference index set obtaining unit 502 is specifically configured to obtain real index sets corresponding to the plurality of user terminals; obtaining a union set of the target real index set and the real index sets corresponding to the plurality of user terminals according to the target real index set and the real index sets corresponding to the plurality of user terminals; and obtaining the target interference index set according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals.
Optionally, the obtaining the target interference index set according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals includes:
obtaining a union set of the target real index set and real index sets corresponding to the plurality of user terminals;
generating a specified problem for the plurality of user terminals according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals;
obtaining random answers of the plurality of user terminals to the specified questions;
and obtaining the target interference index set according to random answers of the plurality of user terminals to the specified questions.
Optionally, the obtaining, according to the target real index set and the real index sets corresponding to the plurality of user terminals, a union set of the target real index set and the real index sets corresponding to the plurality of user terminals includes:
according to the target real index set and the real index sets corresponding to the plurality of user terminals, a bloom counter corresponding to the target real index set is obtained, and bloom counters corresponding to the plurality of user terminals are obtained;
and performing set union calculation on bloom counters corresponding to the target real index set and bloom counters corresponding to the plurality of user terminals to obtain a union of the target real index set and real index sets corresponding to the plurality of user terminals.
Optionally, the to-be-trained object recommendation sub-model obtaining unit 503 is specifically configured to obtain an object identifier set corresponding to the target interference index set; obtaining the position of the object identification set corresponding to the target interference index set in the object recommendation model according to the object identification set corresponding to the target interference index set; and obtaining the target object recommendation sub-model to be trained from a server according to the position of the object identification set corresponding to the target interference index set in the object recommendation model.
Optionally, the object recommendation sub-model to be uploaded obtaining unit 504 is specifically configured to obtain, according to the target real index set, a target succinct index set corresponding to the target user terminal, where the target succinct index set is an index set generated according to a union set of the target real index set and real index sets corresponding to the plurality of user terminals and an intersection set of the target real index sets, and is used to indicate a position of the target object recommendation sub-model in the target object recommendation sub-model to be trained; according to the target succinct index set, obtaining a target object recommendation sub-model from the target object recommendation sub-model to be trained; acquiring target training data for training the target object recommendation sub-model according to the target succinct index set; according to the target training data, performing model training on the target object recommendation sub-model to obtain a target object recommendation sub-model to be updated; and replacing the target object recommendation sub-model in the target object recommendation sub-model to be trained with the target object recommendation sub-model to be updated according to the target succinct index set, and obtaining the target object recommendation sub-model to be uploaded.
Optionally, the obtaining, according to the target real index set, a target succinct index set corresponding to the target user terminal includes:
obtaining a union set of the target real index set and real index sets corresponding to the plurality of user terminals and an intersection set of the union set and the target real index set;
and taking the union set of the target real index set and the real index sets corresponding to the plurality of user terminals and the intersection set of the target real index set as the target succinct index set.
Optionally, the obtaining, according to the target succinct index set, the target object recommendation sub-model from the target object recommendation sub-model to be trained includes: and obtaining the target object recommendation sub-model from the target object recommendation sub-model to be trained according to the position of the object identification set corresponding to the target succinct index set in the target object recommendation sub-model to be trained.
Optionally, the obtaining, according to the target succinct index set, target training data for training the target object recommendation sub-model includes:
obtaining user behavior data related to the object identification set corresponding to the target succinct index set according to the object identification set corresponding to the target succinct index set;
And taking the user behavior data related to the object identification set corresponding to the target succinct index set as target training data.
Optionally, performing model training on the target object recommendation sub-model according to the target training data to obtain a target object recommendation sub-model to be updated, including: and carrying out model training by taking the target training data and the target object recommendation sub-model as input to obtain the target object recommendation sub-model to be updated.
Optionally, replacing the target object recommendation sub-model in the target object recommendation sub-model to be trained with the target object recommendation sub-model to be updated according to the target succinct index set, to obtain the target object recommendation sub-model to be uploaded includes: and replacing the target object recommendation sub-model in the target object recommendation sub-model to be updated with the target object recommendation sub-model to be updated according to the position of the object identification set corresponding to the target succinct index set in the target object recommendation sub-model to be trained, so as to obtain the target object recommendation sub-model to be uploaded.
Optionally, the method further comprises:
according to the target succinct index set, an object identification set corresponding to the target succinct index set is obtained;
And obtaining the position of the object identification set corresponding to the target succinct index set in the target object recommendation sub-model to be trained according to the object identification set corresponding to the target succinct index set.
Optionally, the object recommendation sub-model to be uploaded provides unit 505, which is specifically configured to obtain, according to the target interference index set, a position of the object recommendation sub-model to be uploaded in the object recommendation model; and providing the target object recommendation sub-model to be uploaded for the server side according to the position of the target object recommendation sub-model to be uploaded in the object recommendation model.
Optionally, the processing device for a recommendation model provided in the second embodiment of the present application further includes: a request message unit, configured to obtain a request message sent by the server side and requesting a training object recommendation sub-model;
the object recommendation sub-model to be uploaded providing unit 505 is specifically configured to provide, for the request message, the object recommendation sub-model to be uploaded to the server according to the target interference index set.
The processing device for an object recommendation model provided in the second embodiment of the present application first obtains a real interference index set for indicating a position of a target object recommendation sub-model corresponding to a target object identification set in the object recommendation model, and further obtains a target interference index set for indicating a position of a target object recommendation sub-model to be trained including the target object recommendation sub-model in the object recommendation model according to the obtained real interference index set; secondly, according to the target interference index set, a target object recommendation sub-model to be trained is obtained from the server; thirdly, performing model training on the target object recommendation sub-model to be trained to obtain a target object recommendation sub-model to be uploaded; and finally, providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set. According to the processing device for the object recommendation model, the target object recommendation sub-model to be trained is obtained from the server to conduct model training through the index set generated according to the target real index set and the union set of the real index sets corresponding to the plurality of user terminals, the target object recommendation sub-model to be uploaded is obtained, and the target object recommendation sub-model to be uploaded is provided for the server, so that the efficiency of recommendation model training can be improved on the basis of ensuring the safety of user terminal data, and the complexity of recommendation model training is reduced.
Third embodiment
Corresponding to the processing method of the recommendation model provided in the first embodiment of the present application, a third embodiment of the present application provides an electronic device.
Fig. 6 is a schematic diagram of an electronic device according to an embodiment of the present application, as shown in fig. 6.
The electronic device includes:
a processor 601; and
a memory 602, configured to store a program of a processing method for an object recommendation model, and after the device is powered on and the processor executes the program of the processing method for an object recommendation model, perform the following steps:
obtaining a real interference index set corresponding to a target user terminal, wherein the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of a target object recommendation sub-model corresponding to the target object identification set in the object recommendation model;
obtaining a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to the union of the target real index set and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained, which comprises the target object recommendation sub-model, in the object recommendation model;
According to the target interference index set, obtaining a target object recommendation sub-model to be trained from a server;
model training is carried out on the target object recommendation sub-model to be trained, and a target object recommendation sub-model to be uploaded is obtained;
and providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set.
It should be noted that, for the detailed description of the processing method of the recommendation model performed by the electronic device provided in the third embodiment of the present application, reference may be made to the related description of the first embodiment of the present application, which is not repeated here.
Fourth embodiment
In correspondence with the processing method of the recommended model provided in the first embodiment of the present application, the fourth embodiment of the present application provides a storage medium storing a program of the processing method of the recommended model, the program being executed by a processor to perform the steps of:
obtaining a real interference index set corresponding to a target user terminal, wherein the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of a target object recommendation sub-model corresponding to the target object identification set in the object recommendation model;
Obtaining a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to the union of the target real index set and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained, which comprises the target object recommendation sub-model, in the object recommendation model;
according to the target interference index set, obtaining a target object recommendation sub-model to be trained from a server;
model training is carried out on the target object recommendation sub-model to be trained, and a target object recommendation sub-model to be uploaded is obtained;
and providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set.
It should be noted that, for the detailed description of the storage medium provided in the fourth embodiment of the present application, reference may be made to the related description of the first embodiment of the present application, which is not repeated here.
Fifth embodiment
The fifth embodiment of the present application provides another method for processing an object recommendation model, corresponding to the method for processing an object recommendation model provided in the first embodiment of the present application. Since some of the content of this method embodiment is substantially similar to the first method embodiment, the description is relatively simple, and reference will be made to some explanation of the method embodiment for that matter. The embodiments described below are merely illustrative.
Fig. 7 is a flowchart of a processing method of an object recommendation model according to a fifth embodiment of the present application.
In step S701, a target object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, is obtained, where the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals.
In a fifth embodiment of the present application, a target interference index set is used to indicate a position of a target object recommendation sub-model to be trained, which includes a target object recommendation sub-model, in an object recommendation model, and a target real index set is an index set generated according to a target object identifier set corresponding to a target user terminal, and is used to indicate a position of a target object recommendation sub-model corresponding to the target object identifier set in the object recommendation model, where the target object recommendation sub-model to be uploaded is an object sub-model obtained after the target user terminal performs model training on the target object recommendation sub-model to be trained.
Before the target object recommendation sub-model to be uploaded, the server side sends a request message for requesting training of the target object recommendation sub-model to the target user terminal, and at this time, obtains the target object recommendation sub-model to be uploaded, which is uploaded by the target user terminal according to the target interference index set corresponding to the target user terminal, including: and aiming at the request message, obtaining a target object recommendation sub-model to be uploaded, which is uploaded by the target user terminal according to the target interference index set corresponding to the target user terminal.
In step S702, a target object recommendation model is obtained according to the target object recommendation sub-model to be uploaded, where the target object recommendation model is used for performing object recommendation on the user terminal according to the object identifier set corresponding to the user terminal.
In the fifth embodiment of the present application, there is also a need to obtain an object recommendation sub-model to be uploaded corresponding to a plurality of user terminals, and the object recommendation sub-model to be uploaded for the plurality of user terminals according to an interference index set corresponding to the plurality of user terminals, and correspondingly, the object recommendation model to be uploaded according to the object recommendation sub-model to be uploaded, including: obtaining a target object recommendation sub-model to be uploaded and weight values of the target object recommendation sub-models to be uploaded corresponding to a plurality of user terminals; and according to the weight values of the object recommendation sub-model to be uploaded and the object recommendation sub-models to be uploaded corresponding to the plurality of user terminals, aggregating the object recommendation sub-model to be uploaded and the object recommendation sub-models to be uploaded corresponding to the plurality of user terminals to obtain the object recommendation model to be uploaded.
In the fifth embodiment of the present application, after the target object recommendation model is obtained, an object identifier set corresponding to the user terminal to be recommended may be further obtained, and the object identifier set corresponding to the user terminal to be recommended is input into the target object recommendation model, so as to obtain a recommended object recommended for the user terminal to be recommended.
It should be noted that, for a detailed description of the processing method of the another object recommendation model provided in the fifth embodiment of the present application, reference may be made to the related description of the first embodiment of the present application, which is not repeated here.
Sixth embodiment
The sixth embodiment of the present application provides a processing apparatus for a recommendation model, corresponding to the processing method for a recommendation model provided in the fifth embodiment of the present application. Since the apparatus embodiment is substantially similar to the method of the fifth embodiment, the description is relatively simple, and reference is made to the description of the method embodiment for relevant points. The device embodiments described below are merely illustrative.
Fig. 8 is a schematic diagram of a processing apparatus of a recommendation model according to a sixth embodiment of the present application.
The processing device of the recommendation model comprises:
an object recommendation sub-model to be uploaded obtaining unit 801, configured to obtain an object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, where the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals, and is used to indicate a position of a target object recommendation sub-model to be trained including the target object recommendation sub-model in the object recommendation model, and the target real index set is an index set generated according to a target object identifier set corresponding to the target user terminal, and is used to indicate a position of a target object recommendation sub-model corresponding to the target object identifier set in the object recommendation model, and the object recommendation sub-model to be uploaded is an object sub-model obtained after model training is performed by the target user terminal on the target object recommendation sub-model to be trained;
The target object recommendation model obtaining unit 802 is configured to obtain a target object recommendation model according to the target object recommendation sub-model to be uploaded, where the target object recommendation model is used for performing object recommendation on a user terminal according to an object identifier set corresponding to the user terminal.
Optionally, the processing device for a recommendation model provided in the sixth embodiment of the present application further includes: the second object recommendation sub-model obtaining unit obtains object recommendation sub-models to be uploaded corresponding to a plurality of user terminals, wherein the object recommendation sub-models are uploaded for the plurality of user terminals according to interference index sets corresponding to the plurality of user terminals;
the object recommendation sub-model to be uploaded obtaining unit 801 is specifically configured to obtain weight values of the object recommendation sub-model to be uploaded and object recommendation sub-models to be uploaded corresponding to the plurality of user terminals; and aggregating the target object recommendation sub-model to be uploaded and the object recommendation sub-model to be uploaded corresponding to the plurality of user terminals according to the weight values of the target object recommendation sub-model to be uploaded and the object recommendation sub-models to be uploaded corresponding to the plurality of user terminals, so as to obtain the target object recommendation model.
Optionally, the processing device for a recommendation model provided in the sixth embodiment of the present application further includes: a request message providing unit, configured to send a request message for requesting a training object recommendation sub-model to the target user terminal;
the target object recommendation model obtaining unit 802 is specifically configured to obtain, for the request message, a target object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal.
Optionally, the processing device for a recommendation model provided in the sixth embodiment of the present application further includes:
the object identification set obtaining unit is used for obtaining an object identification set corresponding to the user terminal to be recommended;
the recommended object obtaining unit is used for inputting the object identification set corresponding to the user terminal to be recommended into the target object recommendation model by the object to obtain a recommended object recommended for the user terminal to be recommended.
Seventh embodiment
The seventh embodiment of the present application provides an electronic device, corresponding to the processing method of the recommendation model provided in the fifth embodiment of the present application.
Fig. 6 is a schematic diagram of an electronic device according to an embodiment of the present application, as shown in fig. 6.
The electronic device includes:
a processor 601; and
a memory 602, configured to store a program of a processing method for an object recommendation model, and after the device is powered on and the processor executes the program of the processing method for an object recommendation model, perform the following steps:
obtaining a target object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained comprising the target object recommendation sub-model in the object recommendation model, the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of the target object recommendation sub-model corresponding to the target object identification set in the object recommendation model, and the target object recommendation sub-model to be uploaded is an object sub-model obtained after model training is carried out on the target object recommendation sub-model by the target user terminal;
And obtaining a target object recommendation model according to the target object recommendation sub-model to be uploaded, wherein the target object recommendation model is used for recommending the object for the user terminal according to the object identification set corresponding to the user terminal.
It should be noted that, for a detailed description of the processing method of the recommendation model performed by the electronic device provided in the seventh embodiment of the present application, reference may be made to a related description of the fifth embodiment of the present application, which is not repeated here.
Eighth embodiment
In correspondence with the processing method of the recommended model provided in the fifth embodiment of the present application, the eighth embodiment of the present application provides a storage medium storing a program of the processing method of the recommended model, the program being executed by a processor to perform the steps of:
obtaining a target object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained comprising the target object recommendation sub-model in the object recommendation model, the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of the target object recommendation sub-model corresponding to the target object identification set in the object recommendation model, and the target object recommendation sub-model to be uploaded is an object sub-model obtained after model training is carried out on the target object recommendation sub-model by the target user terminal;
And obtaining a target object recommendation model according to the target object recommendation sub-model to be uploaded, wherein the target object recommendation model is used for recommending the object for the user terminal according to the object identification set corresponding to the user terminal.
It should be noted that, for the detailed description of the storage medium provided in the eighth embodiment of the present application, reference may be made to the related description of the fifth embodiment of the present application, which is not repeated here.
While the preferred embodiment has been described, it is not intended to limit the invention thereto, and any person skilled in the art may make variations and modifications without departing from the spirit and scope of the invention, so that the scope of the invention shall be defined by the claims.
In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include volatile memory, random Access Memory (RAM), and/or nonvolatile memory, such as Read Only Memory (ROM) or Flash memory (Flash RAM), among others, in a computer readable medium. Memory is an example of computer-readable media.
1. Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable Media, as defined herein, does not include non-Transitory computer-readable Media (transmission Media), such as modulated data signals and carrier waves.
2. It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
Claims (25)
1. A method of processing an object recommendation model, comprising:
obtaining a real interference index set corresponding to a target user terminal, wherein the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of a target object recommendation sub-model corresponding to the target object identification set in the object recommendation model;
obtaining a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to the union of the target real index set and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained, which comprises the target object recommendation sub-model, in the object recommendation model;
According to the target interference index set, obtaining a target object recommendation sub-model to be trained from a server;
model training is carried out on the target object recommendation sub-model to be trained, and a target object recommendation sub-model to be uploaded is obtained;
and providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set.
2. The method for processing the object recommendation model according to claim 1, wherein the obtaining the real interference index set corresponding to the target user terminal includes:
obtaining a target object identification set corresponding to the target user terminal;
and obtaining the target real index set according to the target object identification set.
3. The method for processing the object recommendation model according to claim 2, wherein the obtaining the target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal includes:
obtaining real index sets corresponding to the plurality of user terminals;
obtaining a union set of the target real index set and the real index sets corresponding to the plurality of user terminals according to the target real index set and the real index sets corresponding to the plurality of user terminals;
And obtaining the target interference index set according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals.
4. The method for processing the object recommendation model according to claim 3, wherein the obtaining the target interference index set according to the union of the target real index set and the real index sets corresponding to the plurality of user terminals includes:
obtaining a union set of the target real index set and real index sets corresponding to the plurality of user terminals;
generating a specified problem for the plurality of user terminals according to the union set of the target real index set and the real index sets corresponding to the plurality of user terminals;
obtaining random answers of the plurality of user terminals to the specified questions;
and obtaining the target interference index set according to random answers of the plurality of user terminals to the specified questions.
5. The method for processing the object recommendation model according to claim 3, wherein the obtaining, according to the target real index set and the real index sets corresponding to the plurality of user terminals, a union of the target real index set and the real index sets corresponding to the plurality of user terminals includes:
According to the target real index set and the real index sets corresponding to the plurality of user terminals, a bloom counter corresponding to the target real index set is obtained, and bloom counters corresponding to the plurality of user terminals are obtained;
and performing set union calculation on bloom counters corresponding to the target real index set and bloom counters corresponding to the plurality of user terminals to obtain a union of the target real index set and real index sets corresponding to the plurality of user terminals.
6. The method for processing the object recommendation model according to claim 1, wherein the obtaining the object recommendation sub-model to be trained from the server according to the target interference index set includes:
obtaining an object identification set corresponding to the target interference index set;
obtaining the position of the object identification set corresponding to the target interference index set in the object recommendation model according to the object identification set corresponding to the target interference index set;
and obtaining the target object recommendation sub-model to be trained from a server according to the position of the object identification set corresponding to the target interference index set in the object recommendation model.
7. The method for processing the object recommendation model according to claim 1, wherein the model training is performed on the target object recommendation sub-model to be trained to obtain the target object recommendation sub-model to be uploaded, comprising:
Obtaining a target succinct index set corresponding to the target user terminal according to the target real index set, wherein the target succinct index set is an index set generated according to the union of the target real index set and the real index sets corresponding to the plurality of user terminals and the intersection of the target real index set, and is used for indicating the position of the target object recommendation sub-model in the target object recommendation sub-model to be trained;
according to the target succinct index set, obtaining a target object recommendation sub-model from the target object recommendation sub-model to be trained;
acquiring target training data for training the target object recommendation sub-model according to the target succinct index set;
according to the target training data, performing model training on the target object recommendation sub-model to obtain a target object recommendation sub-model to be updated;
and replacing the target object recommendation sub-model in the target object recommendation sub-model to be trained with the target object recommendation sub-model to be updated according to the target succinct index set, and obtaining the target object recommendation sub-model to be uploaded.
8. The method for processing the object recommendation model according to claim 7, wherein the obtaining, according to the target real index set, the target succinct index set corresponding to the target user terminal includes:
Obtaining a union set of the target real index set and real index sets corresponding to the plurality of user terminals and an intersection set of the union set and the target real index set;
and taking the union set of the target real index set and the real index sets corresponding to the plurality of user terminals and the intersection set of the target real index set as the target succinct index set.
9. The method for processing the object recommendation model according to claim 7, wherein the obtaining the target object recommendation sub-model from the target object recommendation sub-model to be trained according to the target succinct index set includes: and obtaining the target object recommendation sub-model from the target object recommendation sub-model to be trained according to the position of the object identification set corresponding to the target succinct index set in the target object recommendation sub-model to be trained.
10. The method for processing the object recommendation model according to claim 7, wherein the obtaining target training data for training the target object recommendation sub-model according to the target brevity index set includes:
obtaining user behavior data related to the object identification set corresponding to the target succinct index set according to the object identification set corresponding to the target succinct index set;
And taking the user behavior data related to the object identification set corresponding to the target succinct index set as target training data.
11. The method for processing the object recommendation model according to claim 7, wherein the performing model training on the object recommendation sub-model according to the target training data to obtain the object recommendation sub-model to be updated includes: and carrying out model training by taking the target training data and the target object recommendation sub-model as input to obtain the target object recommendation sub-model to be updated.
12. The method for processing the object recommendation model according to claim 7, wherein the replacing the target object recommendation sub-model in the target object recommendation sub-model to be trained with the target object recommendation sub-model to be updated according to the target succinct index set, to obtain the target object recommendation sub-model to be uploaded, includes: and replacing the target object recommendation sub-model in the target object recommendation sub-model to be updated with the target object recommendation sub-model to be updated according to the position of the object identification set corresponding to the target succinct index set in the target object recommendation sub-model to be trained, so as to obtain the target object recommendation sub-model to be uploaded.
13. The processing method of an object recommendation model according to claim 9, 10, 12, further comprising:
according to the target succinct index set, an object identification set corresponding to the target succinct index set is obtained;
and obtaining the position of the object identification set corresponding to the target succinct index set in the target object recommendation sub-model to be trained according to the object identification set corresponding to the target succinct index set.
14. The method for processing the object recommendation model according to claim 1, wherein the providing the object recommendation sub-model to be uploaded to the server according to the target interference index set includes:
obtaining the position of the target object recommendation sub-model to be uploaded in the object recommendation model according to the target interference index set;
and providing the target object recommendation sub-model to be uploaded for the server side according to the position of the target object recommendation sub-model to be uploaded in the object recommendation model.
15. The method of processing an object recommendation model according to claim 1, further comprising: acquiring a request message which is sent by the server and is used for requesting a training object recommendation sub-model;
the providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set includes: and for the request message, providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set.
16. A processing apparatus of an object recommendation model, comprising:
the real interference index set obtaining unit is used for obtaining a real interference index set corresponding to a target user terminal, wherein the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of a target object recommendation sub-model corresponding to the target object identification set in the object recommendation model;
a real interference index set obtaining unit, configured to obtain a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal, where the target interference index set is an index set generated according to a union set of the target real index set and real index sets corresponding to a plurality of user terminals, and is configured to indicate a position of a target object recommendation sub-model to be trained including the target object recommendation sub-model in the object recommendation model;
the object recommendation sub-model obtaining unit is used for obtaining the object recommendation sub-model to be trained from the server according to the object interference index set;
the object recommendation sub-model to be uploaded is used for carrying out model training on the target object recommendation sub-model to be trained to obtain the target object recommendation sub-model to be uploaded;
And the object recommendation sub-model to be uploaded is provided for providing the object recommendation sub-model to be uploaded to the server according to the target interference index set.
17. An electronic device, comprising:
a processor; and
a memory for storing a program of a processing method for an object recommendation model, the apparatus being powered on and executing the program of the processing method for an object recommendation model by the processor, and executing the steps of:
obtaining a real interference index set corresponding to a target user terminal, wherein the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of a target object recommendation sub-model corresponding to the target object identification set in the object recommendation model;
obtaining a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to the union of the target real index set and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained, which comprises the target object recommendation sub-model, in the object recommendation model;
According to the target interference index set, obtaining a target object recommendation sub-model to be trained from a server;
model training is carried out on the target object recommendation sub-model to be trained, and a target object recommendation sub-model to be uploaded is obtained;
and providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set.
18. A storage medium storing a program of a processing method for an object recommendation model, the program being executed by a processor to perform the steps of:
obtaining a real interference index set corresponding to a target user terminal, wherein the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of a target object recommendation sub-model corresponding to the target object identification set in the object recommendation model;
obtaining a target interference index set corresponding to the target user terminal according to the obtained real interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to the union of the target real index set and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained, which comprises the target object recommendation sub-model, in the object recommendation model;
According to the target interference index set, obtaining a target object recommendation sub-model to be trained from a server;
model training is carried out on the target object recommendation sub-model to be trained, and a target object recommendation sub-model to be uploaded is obtained;
and providing the target object recommendation sub-model to be uploaded to the server according to the target interference index set.
19. A method of processing an object recommendation model, comprising:
obtaining a target object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained comprising the target object recommendation sub-model in the object recommendation model, the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of the target object recommendation sub-model corresponding to the target object identification set in the object recommendation model, and the target object recommendation sub-model to be uploaded is an object sub-model obtained after model training is carried out on the target object recommendation sub-model by the target user terminal;
And obtaining a target object recommendation model according to the target object recommendation sub-model to be uploaded, wherein the target object recommendation model is used for recommending the object for the user terminal according to the object identification set corresponding to the user terminal.
20. The method of processing an object recommendation model according to claim 19, further comprising: obtaining object recommendation sub-models to be uploaded corresponding to a plurality of user terminals, wherein the object recommendation sub-models are uploaded for the plurality of user terminals according to interference index sets corresponding to the plurality of user terminals;
the obtaining the target object recommendation model according to the target object recommendation sub-model to be uploaded comprises the following steps:
acquiring weight values of the target object recommendation sub-model to be uploaded and the object recommendation sub-models to be uploaded corresponding to the plurality of user terminals;
and aggregating the target object recommendation sub-model to be uploaded and the object recommendation sub-model to be uploaded corresponding to the plurality of user terminals according to the weight values of the target object recommendation sub-model to be uploaded and the object recommendation sub-models to be uploaded corresponding to the plurality of user terminals, so as to obtain the target object recommendation model.
21. The method of processing an object recommendation model according to claim 19, further comprising: transmitting a request message for requesting a training object recommendation sub-model to the target user terminal;
The obtaining the target object recommendation sub-model to be uploaded, which is uploaded by the target user terminal according to the target interference index set corresponding to the target user terminal, includes: and aiming at the request message, obtaining a target object recommendation sub-model to be uploaded, which is uploaded by the target user terminal according to the target interference index set corresponding to the target user terminal.
22. The method of processing an object recommendation model according to claim 19, further comprising:
obtaining an object identification set corresponding to a user terminal to be recommended;
and inputting the object identification set corresponding to the user terminal to be recommended into the target object recommendation model to obtain a recommended object recommended for the user terminal to be recommended.
23. A processing apparatus of an object recommendation model, comprising:
the object recommendation sub-model to be uploaded is used for obtaining an object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals, the index set is used for indicating the position of a target object recommendation sub-model to be trained containing the object recommendation sub-model in the object recommendation model, the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal, the index set is used for indicating the position of the target object recommendation sub-model corresponding to the target object identification set in the object recommendation model, and the object recommendation sub-model to be uploaded is an object sub-model obtained after model training is carried out on the target object recommendation sub-model by the target user terminal;
The target object recommendation model obtaining unit is used for obtaining a target object recommendation model according to the target object recommendation sub-model to be uploaded, and the target object recommendation model is used for carrying out object recommendation on the user terminal according to an object identification set corresponding to the user terminal.
24. An electronic device, comprising:
a processor; and
a memory for storing a program of a processing method for an object recommendation model, the apparatus being powered on and executing the program of the processing method for an object recommendation model by the processor, and executing the steps of:
obtaining a target object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained comprising the target object recommendation sub-model in the object recommendation model, the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of the target object recommendation sub-model corresponding to the target object identification set in the object recommendation model, and the target object recommendation sub-model to be uploaded is an object sub-model obtained after model training is carried out on the target object recommendation sub-model by the target user terminal;
And obtaining a target object recommendation model according to the target object recommendation sub-model to be uploaded, wherein the target object recommendation model is used for recommending the object for the user terminal according to the object identification set corresponding to the user terminal.
25. A storage medium storing a program of a processing method for an object recommendation model, the program being executed by a processor to perform the steps of:
obtaining a target object recommendation sub-model to be uploaded, which is uploaded by a target user terminal according to a target interference index set corresponding to the target user terminal, wherein the target interference index set is an index set generated according to a union set of a target real index set corresponding to the target user terminal and real index sets corresponding to a plurality of user terminals and is used for indicating the position of a target object recommendation sub-model to be trained comprising the target object recommendation sub-model in the object recommendation model, the target real index set is an index set generated according to a target object identification set corresponding to the target user terminal and is used for indicating the position of the target object recommendation sub-model corresponding to the target object identification set in the object recommendation model, and the target object recommendation sub-model to be uploaded is an object sub-model obtained after model training is carried out on the target object recommendation sub-model by the target user terminal;
And obtaining a target object recommendation model according to the target object recommendation sub-model to be uploaded, wherein the target object recommendation model is used for recommending the object for the user terminal according to the object identification set corresponding to the user terminal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010120167.7A CN113312542B (en) | 2020-02-26 | 2020-02-26 | Processing method and device of object recommendation model and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010120167.7A CN113312542B (en) | 2020-02-26 | 2020-02-26 | Processing method and device of object recommendation model and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113312542A CN113312542A (en) | 2021-08-27 |
| CN113312542B true CN113312542B (en) | 2023-12-22 |
Family
ID=77370071
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010120167.7A Active CN113312542B (en) | 2020-02-26 | 2020-02-26 | Processing method and device of object recommendation model and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113312542B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019128552A1 (en) * | 2017-12-29 | 2019-07-04 | Oppo广东移动通信有限公司 | Information pushing method, apparatus, terminal, and storage medium |
| CN110297848A (en) * | 2019-07-09 | 2019-10-01 | 深圳前海微众银行股份有限公司 | Recommended models training method, terminal and storage medium based on federation's study |
| CN110659920A (en) * | 2018-06-28 | 2020-01-07 | 阿里巴巴集团控股有限公司 | Method and device for recommending service object |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11514333B2 (en) * | 2018-04-30 | 2022-11-29 | Meta Platforms, Inc. | Combining machine-learning and social data to generate personalized recommendations |
-
2020
- 2020-02-26 CN CN202010120167.7A patent/CN113312542B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019128552A1 (en) * | 2017-12-29 | 2019-07-04 | Oppo广东移动通信有限公司 | Information pushing method, apparatus, terminal, and storage medium |
| CN110659920A (en) * | 2018-06-28 | 2020-01-07 | 阿里巴巴集团控股有限公司 | Method and device for recommending service object |
| CN110297848A (en) * | 2019-07-09 | 2019-10-01 | 深圳前海微众银行股份有限公司 | Recommended models training method, terminal and storage medium based on federation's study |
Non-Patent Citations (2)
| Title |
|---|
| 基于SVD++与标签的跨域推荐模型;邢长征;杨晓婷;;计算机工程(第04期);全文 * |
| 大数据下基于Spark的电商实时推荐系统的设计与实现;岑凯伦;于红岩;杨腾霄;;现代计算机(专业版)(第24期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113312542A (en) | 2021-08-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110392155B (en) | Notification message display and processing method, device and equipment | |
| CN109472677A (en) | Information-pushing method, device, electronic equipment and computer readable storage medium | |
| CN107808346B (en) | Evaluation method and evaluation device for potential target object | |
| US20140051464A1 (en) | Service prequalification | |
| CN107733786B (en) | Friend recommendation method and device | |
| CN110688574A (en) | Learning content recommendation method, system, device and readable storage medium | |
| CN110796269A (en) | Method and device for generating model, and method and device for processing information | |
| CN109451334B (en) | User portrait generation processing method and device and electronic equipment | |
| CN109993562B (en) | Satisfaction degree simulation method and device and terminal equipment | |
| CN113312542B (en) | Processing method and device of object recommendation model and electronic equipment | |
| CN111177093A (en) | Method, device and medium for sharing scientific and technological resources | |
| CN106804056A (en) | A kind of method and apparatus for providing the access information of WAP | |
| CN107547221A (en) | A kind of method and apparatus for being used to provide log information | |
| CN112668723A (en) | Machine learning method and system | |
| CN110336731B (en) | User matching method and device in group | |
| CN109582834B (en) | Data risk prediction method and device | |
| CN116680480A (en) | Product recommendation method and device, electronic equipment and readable storage medium | |
| CN115564486A (en) | Data pushing method, device, equipment and medium | |
| CN113821617A (en) | Questionnaire survey method, equipment and storage medium | |
| CN108460475A (en) | Poor student's prediction technique and device based on network playing by students behavior | |
| CN110337074B (en) | Interactive information transmission method, system and terminal equipment | |
| US20160225112A1 (en) | Research system and method | |
| CN113487389A (en) | Information recommendation method and device | |
| CN107423982A (en) | Service implementation method and device based on account | |
| CN107548064A (en) | For the method and apparatus for the security information for providing WAP |
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 |