CN111079022B - Personalized recommendation method, device, equipment and medium based on federal learning - Google Patents
Personalized recommendation method, device, equipment and medium based on federal learning Download PDFInfo
- Publication number
- CN111079022B CN111079022B CN201911326853.3A CN201911326853A CN111079022B CN 111079022 B CN111079022 B CN 111079022B CN 201911326853 A CN201911326853 A CN 201911326853A CN 111079022 B CN111079022 B CN 111079022B
- Authority
- CN
- China
- Prior art keywords
- data
- personalized recommendation
- model
- gradient
- federal learning
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- 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
- G06F16/9538—Presentation of query results
Abstract
The application discloses a personalized recommendation method, a device, equipment and a medium based on federal learning, wherein the personalized recommendation method based on federal learning comprises the following steps: receiving uploading data, extracting a target recall set corresponding to the uploading data from a preset recall set storage database, obtaining data to be predicted which are commonly corresponding to the uploading data and the target recall set, inputting the data to be predicted into a personalized recommendation model obtained based on federal learning, obtaining a model output result, and screening the model output result to obtain the personalized recommendation result. The application solves the technical problem of poor personalized recommendation effect in the prior art.
Description
Technical Field
The application relates to the technical field of artificial intelligence of financial science and technology (Fintech), in particular to a personalized recommendation method, device, equipment and medium based on federal learning.
Background
With the continuous development of financial technology, especially internet technology finance, more and more technologies (such as distributed, blockchain, artificial intelligence, etc.) are applied in the finance field, but the finance industry also puts forward higher requirements on technologies, such as distribution of corresponding backlog in the finance industry.
Along with the continuous development of computer software and artificial intelligence, the personalized recommendation technology is also widely applied, at present, a personalized recommendation provider predicts personalized behaviors or articles of a user through data such as user attribute data, user behavior context data and the like obtained by a user, for example, a mobile phone favored by the user and a webpage click rate of the user are predicted, however, in the method, the feature richness of the user data often has great influence on a prediction result, and along with the strictness of a data privacy protection legislation, the user data cannot be subjected to plaintext sharing in all parties of different data, so the feature richness of the user data of a single personalized recommendation provider is often lower, the prediction precision of the personalized behaviors or articles of the user is further lower, and the personalized recommendation effect is further poor, and therefore, the technical problem of poor personalized recommendation effect exists in the prior art.
Disclosure of Invention
The application mainly aims to provide a personalized recommendation method, device, equipment and medium based on federal learning, and aims to solve the technical problem of poor personalized recommendation effect in the prior art.
In order to achieve the above object, an embodiment of the present application provides a personalized recommendation method based on federal learning, where the personalized recommendation method based on federal learning is applied to a personalized recommendation device based on federal learning, and the personalized recommendation method based on federal learning includes:
receiving uploading data, and extracting a target recall set corresponding to the uploading data from a preset recall set storage database;
obtaining to-be-predicted data which corresponds to the uploading data and the target recall set together, inputting the to-be-predicted data into a personalized recommendation model obtained based on the federal learning, and obtaining a model output result;
and screening the model output result to obtain a personalized recommendation result.
Optionally, the target recall set includes a list of items to be recommended, the upload data includes user data, item data and behavioral data,
the step of obtaining the data to be predicted, which corresponds to the uploading data and the target recall set together, and inputting the data to be predicted into the personalized recommendation model obtained based on the federal learning, and obtaining a model output result comprises the following steps:
screening the item data based on the item list to be recommended to obtain item data to be recommended, and setting the user data, the item data to be recommended and the behavior data as the data to be predicted;
And inputting the data to be predicted into the personalized recommendation model to score and sort the articles to be recommended in the article list to be recommended, so as to obtain a model output result.
Optionally, the step of inputting the data to be predicted into the personalized recommendation model to score and sort the articles to be recommended in the article list to be recommended, and the step of obtaining a model output result includes:
inputting the data to be predicted into the personalized recommendation model to score the articles to be recommended based on the user data and the behavior data, and obtaining a scoring result;
and sorting the articles to be recommended based on the scoring result to obtain a model output result.
Optionally, the personalized recommendation method based on federal learning is applied to a first device performing federal learning,
the step of obtaining the data to be predicted, which corresponds to the uploading data and the target recall set together, and inputting the data to be predicted into the personalized recommendation model obtained based on the federal learning, and the step of obtaining the model output result comprises the following steps:
sample matching is carried out on second equipment associated with the first equipment, and a public training sample ID (Identity document, identity card identification number) is obtained;
Based on the public training sample ID, the personalized recommendation model is obtained by interacting with the second device to perform federal learning.
Optionally, the personalized recommendation model comprises a logistic regression model,
the step of obtaining the personalized recommendation model by interacting with the second device for federal learning based on the common training sample ID comprises:
based on the public training sample ID, extracting first sample data corresponding to the public training sample ID, and calculating a first weight corresponding to the first sample data;
receiving a second weight sent by the second equipment, calculating a gradient auxiliary variable which corresponds to the first weight and the second weight together through a preset intermediate parameter formula, and sending the gradient auxiliary variable to the second equipment, wherein the second equipment is used for calculating a second gradient which corresponds to the gradient auxiliary variable;
calculating a first gradient based on the gradient auxiliary variable, and sending the first gradient to a preset federal server, wherein the preset federal server is used for calculating a federal model total gradient based on the first gradient and a second gradient sent by the second device;
And receiving the federal model total gradient fed back by the federal server, and iteratively updating the local training model of the first device based on the federal model total gradient to obtain the logistic regression model.
Optionally, the step of receiving the upload data and extracting the target recall corresponding to the upload data from the preset recall storage database includes:
receiving uploading data and extracting a sample ID in the uploading data;
and inquiring a corresponding target recall set in the preset recall set storage database based on the sample ID.
Optionally, the personalized recommendation method based on federal learning is applied to a first device performing federal learning,
the step of receiving the uploading data and extracting the target recall corresponding to the uploading data from the preset recall storage database comprises the following steps:
performing federal learning by second equipment associated with the first equipment to obtain a federal recall algorithm model;
sample uploading data are obtained, the sample uploading data are input into the federation recall algorithm model, a target recall set is obtained, and the target recall set is stored in the preset recall set storage database.
The application also provides a personalized recommendation device based on federal learning, which is applied to personalized recommendation equipment based on federal learning, and comprises:
the extraction module is used for receiving the uploading data and extracting a target recall set corresponding to the uploading data from a preset recall set storage database;
the prediction module is used for acquiring data to be predicted which corresponds to the uploading data and the target recall together, inputting the data to be predicted into a personalized recommendation model acquired based on the federal learning, and acquiring a model output result;
and the screening module is used for screening the model output result to obtain a personalized recommendation result.
Optionally, the prediction module includes:
the screening unit is used for screening the item data based on the item list to be recommended to obtain item data to be recommended, and setting the user data, the item data to be recommended and the behavior data as the data to be predicted;
the predicting unit is used for inputting the data to be predicted into the personalized recommendation model so as to score and sort the articles to be recommended in the article list to be recommended, and a model output result is obtained.
Optionally, the prediction unit includes:
the scoring subunit is used for inputting the data to be predicted into the personalized recommendation model so as to score the articles to be recommended based on the user data and the behavior data and obtain a scoring result;
and the sorting subunit is used for sorting the articles to be recommended based on the scoring result to obtain a model output result.
Optionally, the personalized recommendation device based on federal learning further comprises:
the sample matching module is used for carrying out sample matching on the second equipment associated with the first equipment to obtain a public training sample ID;
and the first federal learning module is used for obtaining the personalized recommendation model by interacting with the second equipment to perform federal learning based on the public training sample ID.
Optionally, the federal learning module includes:
the first calculating unit is used for extracting first sample data corresponding to the public training sample ID based on the public training sample ID and calculating a first weight corresponding to the first sample data;
the second computing unit is used for receiving the second weight sent by the second device, computing a gradient auxiliary variable which corresponds to the first weight and the second weight together through a preset intermediate parameter formula, and sending the gradient auxiliary variable to the second device, wherein the second device is used for computing a second gradient which corresponds to the gradient auxiliary variable;
A third calculation unit, configured to calculate a first gradient based on the gradient auxiliary variable, and send the first gradient to a preset federal server, where the preset federal server is configured to calculate a federal model total gradient based on the first gradient and a second gradient sent by the second device;
and the iteration updating unit is used for receiving the total federal model gradient fed back by the federal server, and carrying out iteration updating on the local training model of the first equipment based on the total federal model gradient to obtain the logistic regression model.
Optionally, the extracting module includes:
the extraction unit is used for receiving the uploading data and extracting a sample ID in the uploading data;
and the inquiring unit is used for inquiring the corresponding target recall set in the preset recall set storage database based on the sample ID.
Optionally, the personalized recommendation device based on federal learning includes:
the second federation learning module is used for performing federation learning on the second equipment associated with the first equipment to obtain a federation recall algorithm model;
the storage module is used for acquiring sample uploading data, inputting the sample uploading data into the federal recall algorithm model, obtaining a target recall set, and storing the target recall set in the preset recall set storage database.
The application also provides personalized recommendation equipment based on federal learning, which comprises: the personalized recommendation method based on the federal learning comprises a memory, a processor and a program of the personalized recommendation method based on the federal learning, wherein the program of the personalized recommendation method based on the federal learning is stored in the memory and can run on the processor, and the program of the personalized recommendation method based on the federal learning can realize the steps of the personalized recommendation method based on the federal learning when being executed by the processor.
The application also provides a medium which is a readable storage medium and is stored with a program for realizing the personalized recommendation method based on federal learning, and the program for realizing the personalized recommendation method based on federal learning realizes the steps of the personalized recommendation method based on federal learning when being executed by a processor.
According to the method, the uploading data are received, the target recall set corresponding to the uploading data is extracted from the preset recall set storage database, the uploading data and the target recall set are obtained to be predicted together, the to-be-predicted data are input into the personalized recommendation model obtained based on federal learning, the model output result is obtained, and the model output result is screened to obtain the personalized recommendation result. The method comprises the steps of firstly receiving uploading data, extracting a target recall set corresponding to the uploading data from a preset recall set storage database, further obtaining to-be-predicted data commonly corresponding to the uploading data and the target recall set, inputting the to-be-predicted data into a personalized recommendation model obtained based on federal learning, obtaining a model output result, and further screening the model output result to obtain a personalized recommendation result. That is, the method and the device for predicting the personalized recommendation result of the user based on the federal learning acquire the data to be predicted based on the uploaded data and the target recall, and then the personalized recommendation result of the user is predicted by inputting the data to be predicted into the personalized recommendation model acquired based on the federal learning, that is, the personalized recommendation model can be trained by combining the multiparty data through the federal learning, so that the feature richness of a training sample of the personalized recommendation model is improved, the data privacy of each data provider and each data user cannot be revealed, the robustness and the breadth of the personalized recommendation model are improved, the prediction accuracy of the personalized recommendation model is improved, the situation that the personalized recommendation effect is poor due to the prediction accuracy of the personalized recommendation model is avoided, the model prediction effect of combining the multiparty data to predict is achieved under the condition that the personalized recommendation is performed by only using the local data, the calculation amount of the personalized recommendation model is reduced, the personalized recommendation response speed is improved, and the technical problem is solved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application.
In order to more clearly illustrate the embodiments of the application or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, and it will be obvious to a person skilled in the art that other drawings can be obtained from these drawings without inventive effort.
FIG. 1 is a schematic flow chart of a first embodiment of a personalized recommendation method based on federal learning;
FIG. 2 is a flow chart of a second embodiment of a personalized recommendation method based on federal learning according to the present application;
fig. 3 is a schematic device structure diagram of a hardware running environment according to an embodiment of the present application.
The achievement of the objects, functional features and advantages of the present application will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
The embodiment of the application provides a personalized recommendation method based on federal learning, which is applied to personalized recommendation equipment based on federal learning, and in a first embodiment of the personalized recommendation method based on federal learning, referring to fig. 1, the personalized recommendation method based on federal learning comprises:
Step S10, receiving uploading data, and extracting a target recall set corresponding to the uploading data from a preset recall set storage database;
in this embodiment, it should be noted that, the upload data includes a sample ID, user data, article data and behavior data, where the upload data is uploaded by a client, the user data includes user natural attribute data and user interest attribute data, the article data includes data such as an article name and an article attribute of an article to be personalized recommended, the behavior data includes behavior data of the user on the article and context data when the behavior occurs, for example, behavior data of the user on the article includes browsing, clicking, and the like, and the context data when the behavior occurs includes a geographic location, a network type, and the like, and the sample ID includes a user name, a user identification number, a user phone number, and the like.
Receiving uploading data, extracting a target recall set corresponding to the uploading data from a preset recall set storage database, specifically receiving the uploading data, and inquiring the target recall set corresponding to the user data in the preset recall set storage database based on the user data in the uploading data, wherein the target recall set comprises one or more initial articles to be recommended, and the initial articles to be recommended refer to the articles possibly interested by the user.
The step of receiving the uploading data and extracting the target recall set corresponding to the uploading data from a preset recall set storage database comprises the following steps:
step S11, receiving uploading data and extracting a sample ID in the uploading data;
in this embodiment, it should be noted that, the sample ID includes an identity tag such as a user ID, a character string, etc., the uploaded data includes one or more data samples, and the data samples are represented by a preset data sample format, for example, "ID, label, user_a_feature_i, item_feature_i, action_i, other_a_i" is a preset data sample format, where ID is the sample ID, label is a sample tag, where the sample tag identifies a type of a user, for example, the user is a good client or a bad client, etc., user_a_feature_i is user data, item_feature_i is item data, action_i is behavior data, and other_a_i is other data.
Step S12, based on the sample ID, inquiring a corresponding target recall set in the preset recall set storage database.
In this embodiment, the sample ID includes an identity tag such as a user ID, a character string, and a phone number, that is, the sample ID may be represented by a character string or may be directly represented by a phone number.
And inquiring a corresponding target recall set in the preset recall set storage database based on the sample ID, specifically, searching in the preset recall set storage database by taking the sample ID as a keyword to obtain the target recall set.
Wherein the personalized recommendation method based on federal learning is applied to first equipment for federal learning,
the step of receiving the uploading data and extracting the target recall corresponding to the uploading data from the preset recall storage database comprises the following steps:
step A10, performing federal learning by a second device associated with the first device to obtain a federal recall algorithm model;
in this embodiment, the federal learning includes longitudinal federal learning and transverse federal learning, and the first device and the second device may be communicatively connected.
Performing federal learning on a second device associated with the first device to obtain a federal recall algorithm model, specifically, performing sample alignment on the second device to align a sample ID of the first device with a sample ID of the second device to obtain a common training sample ID, training a first training model of the first device based on uploading data corresponding to the common training sample ID to obtain a first training result, calculating an error between the first training result and a theoretical training result, further performing bias on a preset objective function based on the error and a network weight of the first training model to obtain a first gradient, wherein the preset objective function is a function of the error and the network weight, further transmitting the first gradient to a preset federal server to obtain a federal gradient based on a preset federal rule by the preset federal server, further receiving a gradient fed back by the preset server, further performing equal iteration on the federal error, and the first gradient is updated based on the preset federal rule, and the federal error is equal, and the iteration is stopped, and the iteration is equal.
And step A20, obtaining sample uploading data, inputting the sample uploading data into the federation recall algorithm model, obtaining a target recall set, and storing the target recall set in the preset recall set storage database.
In this embodiment, it should be noted that the target recall set includes one or more target recalls, where each sample ID corresponds to a target recall set.
Obtaining sample uploading data, inputting the sample uploading data into the federation recall algorithm model to obtain a target recall set, storing the target recall set in the preset recall set storage database, specifically obtaining sample uploading data, wherein the sample uploading data comprises uploading data of each user stored in the first equipment, each user corresponds to a sample ID, the uploading data comprises user data, article data, behavior data and the like, further inputting the sample uploading data into the federation recall algorithm model to predict articles which the user of the first equipment may be interested in, obtaining one or more articles to be recommended, dividing the articles to be recommended which belong to the same sample ID into one target recall set, obtaining the target recall set, storing the target recall set in the preset recall set storage database, wherein the target recall set is stored in the preset recall set storage database in an article storage list, and the like, and the target recall set comprises a target recall set in a keyword storage form, and the target recall set comprises a sample ID.
Step S20, obtaining the data to be predicted, which corresponds to the uploading data and the target recall set together, and inputting the data to be predicted into a personalized recommendation model obtained based on the federal learning to obtain a model output result;
in this embodiment, it should be noted that two or more participants of federal learning are provided, where the participants include a first device and a second device, the personalized recommendation model obtained based on federal learning includes a logistic regression model already trained based on federal learning, and the model output result includes items to be recommended and scores and ranks thereof, and the data to be predicted includes user data, behavior data, item data corresponding to the target recall, and the like.
And obtaining the data to be predicted, which corresponds to the uploading data and the target recall set together, and inputting the data to be predicted into a personalized recommendation model obtained based on the federal learning to obtain a model output result, specifically inputting the data to be predicted, which corresponds to the uploading data and the target recall set together, into a personalized recommendation model obtained based on the federal learning to score and sort the articles to be recommended, which correspond to the target recall set, to obtain an article sorting list, which corresponds to the articles to be recommended, and a score, which corresponds to the articles to be recommended, namely, to obtain the model output result.
Wherein the target recall includes a list of items to be recommended, the upload data includes user data, item data, and behavioral data,
the step of obtaining the data to be predicted, which corresponds to the uploading data and the target recall set together, and inputting the data to be predicted into the personalized recommendation model obtained based on the federal learning, and obtaining a model output result comprises the following steps:
step S21, screening the item data based on the item list to be recommended to obtain item data to be recommended, and setting the user data, the item data to be recommended and the behavior data as the data to be predicted;
in this embodiment, it should be noted that the items in the item to be recommended list are associated with the item to be recommended data.
And S22, inputting the data to be predicted into the personalized recommendation model to score and sort the articles to be recommended in the article list to be recommended, and obtaining a model output result.
In this embodiment, the data to be predicted is input into the personalized recommendation model to score and sort the articles to be recommended in the article list to obtain a model output result, specifically, the data to be predicted is input into the personalized recommendation model to score the articles in the article list to be recommended based on the user data, the behavior data and the article data to be recommended to obtain a scoring result, and then the articles in the article list to be recommended are sorted based on the scoring result to obtain a model output result.
In step S22, the step of inputting the data to be predicted into the personalized recommendation model to score and sort the articles to be recommended in the article list to be recommended, and the step of obtaining a model output result includes:
step S221, inputting the data to be predicted into the personalized recommendation model to score the articles to be recommended based on the user data and the behavior data, and obtaining a scoring result;
in this embodiment, the data to be predicted is input into the personalized recommendation model to score the article to be recommended based on the user data and the behavior data, so as to obtain a scoring result, specifically, the data to be predicted is input into the personalized recommendation model to score the article to be recommended based on the user data and the behavior data, so as to obtain a first scoring result and a second scoring result, and the scoring result is calculated by a preset calculation rule based on the first scoring result and the second scoring result, where the preset calculation rule includes weighted average, summation, product seeking, and the like, for example, assuming that the calculation rule is summation, the first scoring result is 1 score, the second scoring result is 2 scores, and the scoring result is 3 scores.
And step S222, sorting the articles to be recommended based on the scoring result to obtain a model output result.
In this embodiment, the articles to be recommended are ranked based on the scoring result to obtain a model output result, and specifically, the scored articles to be recommended are ranked in a preset ranking manner based on the scoring result, where the preset ranking manner includes ranking from small to large, ranking from large to small, and the like, so as to obtain the model output result.
And step S30, screening the model output result to obtain a personalized recommendation result.
In this embodiment, the model output result is screened to obtain a personalized recommendation result, specifically, the model output result is screened based on a preset business logic, and the articles to be recommended in the model output result are extracted as the personalized recommendation result, for example, the articles with the highest scores in the model output result and the preset number are extracted as the personalized recommendation result.
According to the method, the target recall set corresponding to the uploading data is extracted from the preset recall set storage database, the uploading data and the target recall set are obtained to be predicted data which corresponds to the target recall set together, the to-be-predicted data are input into the personalized recommendation model obtained based on federal learning, the model output result is obtained, and the model output result is screened to obtain the personalized recommendation result. That is, in this embodiment, the receiving of the uploading data is performed first, then the extracting of the target recall set corresponding to the uploading data from the preset recall set storage database is performed, then the obtaining of the data to be predicted, which corresponds to the uploading data and the target recall set together, is performed, the data to be predicted is input into the personalized recommendation model obtained based on the federal learning, the model output result is obtained, and then the screening of the model output result is performed, so as to obtain the personalized recommendation result. That is, in this embodiment, after data to be predicted is obtained based on the uploading data and the target recall set, the data to be predicted is input into the personalized recommendation model obtained based on the federal learning, so as to predict the personalized recommendation result of the user, where the personalized recommendation model is obtained based on the federal learning, that is, the personalized recommendation model can be trained by combining the multiparty data through the federal learning, so that the feature richness of the training sample of the personalized recommendation model is improved, the data privacy of each data provider and the data user cannot be revealed, the robustness and the breadth of the personalized recommendation model are improved, the prediction accuracy of the personalized recommendation model is improved, the situation that the personalized recommendation effect is poor due to the prediction accuracy of the personalized recommendation model is avoided, and the model prediction effect of combining the multiparty data to predict is achieved under the condition that only the local data is used for prediction, so that the calculation amount of the personalized recommendation is reduced, the personalized recommendation response speed is improved, and the technical problem is solved.
Further, referring to fig. 2, in another embodiment of the personalized recommendation method based on federal learning according to the first embodiment of the present application, the personalized recommendation method based on federal learning is applied to a first device performing federal learning,
the step of obtaining the data to be predicted, which corresponds to the uploading data and the target recall set together, and inputting the data to be predicted into the personalized recommendation model obtained based on the federal learning, and the step of obtaining the model output result comprises the following steps:
step B10, performing sample matching with second equipment associated with the first equipment to obtain a public training sample ID;
in this embodiment, it should be noted that the public training sample ID includes an identity tag such as a user ID and an identification string.
Sample matching is carried out on second equipment associated with the first equipment to obtain a public training sample ID, and specifically, intersection processing is carried out on the first training sample ID in the first equipment and the second training sample ID in the second equipment to obtain the public training sample ID.
And step B20, based on the public training sample ID, obtaining the personalized recommendation model by interacting with the second equipment to perform federal learning.
In this embodiment, based on the common training sample ID, performing federal learning by interacting with the second device to obtain the personalized recommendation model, specifically, based on the common training sample ID, extracting first sample data corresponding to the common training sample ID, and based on the first sample data, further, obtaining a first gradient, by interacting with the second device, assisting the second device to obtain a second gradient, and sending the first gradient to a preset federal server, further, receiving a federal model total gradient fed back by the preset federal server, where the federal model total gradient is obtained by the preset federal server federally using a preset federal rule to obtain the personalized recommendation model by performing weighted average, summation, and the like on the first training model in the first device based on the federal model total gradient.
Wherein the personalized recommendation model comprises a logistic regression model,
the step of obtaining the personalized recommendation model by interacting with the second device for federal learning based on the common training sample ID comprises:
Step B21, extracting first sample data corresponding to the public training sample ID based on the public training sample ID, and calculating a first weight corresponding to the first sample data;
in this embodiment, it should be noted that, the first sample data is upload data of a user client, the upload data includes one or more data samples, the first weight is a product of a first sample feature corresponding to the first sample data and a first weight in the first training model, where the first weight is a network weight of the first training model, for example, assume that the first weight is W B The first sample is characterized by X B The first weight is W B X B 。
Step B22, receiving a second weight sent by the second device, calculating a gradient auxiliary variable which corresponds to the first weight and the second weight together through a preset intermediate parameter formula, and sending the gradient auxiliary variable to the second device, wherein the second device is used for calculating a second gradient which corresponds to the gradient auxiliary variable;
in this embodiment, a second weight sent by the second device is received, and a gradient auxiliary variable corresponding to the first weight and the second weight is calculated through a preset intermediate parameter formula, and the gradient auxiliary variable is sent to the second device, where the second device is configured to calculate a second gradient corresponding to the gradient auxiliary variable, specifically, receive the second weight sent by the second device in an encrypted manner, decrypt the encrypted second weight to obtain a second weight, where the second weight is a product of a second sample feature corresponding to the second sample data and a second weight in a second training model in the second device, where the second weight is a network weight of the second training model, further, perform a union process on the first weight and the second weight to obtain a total weight corresponding to the first device and the second device, and further, substitute the total weight variable into the preset intermediate parameter to obtain the auxiliary gradient variable, and send the auxiliary variable to the auxiliary gradient formula as follows, where the auxiliary variable is associated with the intermediate parameter formula,
Wherein [ (d)]]As the gradient auxiliary variable, w T x is the total weight, y is the sample label, y may take a value of 1 or-1, for example, when y=1, it may indicate that the customer is a good customer, and when y= -1, it may indicate that the customer is a bad customer.
Step B23, calculating a first gradient based on the gradient auxiliary variable, and sending the first gradient to a preset federal server, wherein the preset federal server is used for calculating a federal model total gradient based on the first gradient and a second gradient sent by the second device;
in this embodiment, a first gradient is calculated based on the gradient auxiliary variable, and the first gradient is sent to a preset federal server, where the preset federal server is configured to calculate a total federal model gradient based on the first gradient and a second gradient sent by the second device, specifically, calculate a product of the gradient auxiliary variable and a first sample feature, obtain the first gradient, where the first gradient may be calculated by the following formula,
wherein g B For the first sampling gradient, w T x is the aboveThe total weight variable, y, is the sample tag, y may take a value of 1 or-1, e.g., when y=1, it may represent that the customer is a good customer, and when y= -1, it may represent that the customer is a bad customer, x B For the first sample feature of the first device, further, the first gradient is sent to a preset federal server, so that a federal model total gradient corresponding to the first gradient and a second gradient sent by the second device is calculated by the preset federal server based on preset federal rules, wherein the preset federal rules comprise weighted average, summation and the like, and the second gradient is the product of the gradient auxiliary variable and the second sample feature.
And step B24, receiving the total gradient of the federation model fed back by the federation server, and carrying out iterative updating on the local training model of the first equipment based on the total gradient of the federation model to obtain the logistic regression model.
In this embodiment, it should be noted that the local training model is the first training model.
And receiving the total gradient of the federal model fed back by the federal server, iteratively updating the local training model of the first device based on the total gradient of the federal model to obtain the logistic regression model, specifically, receiving the total gradient of the federal model fed back by the federal server, performing training updating on the local training model of the first device based on the total gradient of the federal model, judging whether the updated local training device reaches a preset training completion condition, taking the updated local training device as the logistic regression model if the updated local training device reaches the preset training completion condition, and performing next federal learning to perform training updating on the local training device if the updated local training device does not reach the preset training completion condition until the local training device reaches the preset training completion condition, wherein the training completion condition comprises that the model reaches the maximum iteration times, the model error is smaller than a preset error threshold value and converges, and the like.
In this embodiment, sample matching is performed by a second device associated with the first device, so as to obtain a public training sample ID, and further based on the public training sample ID, federal learning is performed by interacting with the second device, so as to obtain the personalized recommendation model. That is, in this embodiment, first, sample matching of a second device associated with the first device is performed to obtain a public training sample ID, and then based on the public training sample ID, and interaction with the second device is performed to perform federal learning, so as to obtain the personalized recommendation model. That is, the embodiment provides a method for obtaining a personalized recommendation model through federal learning, that is, by interacting with the second device, federal learning can be performed to obtain the personalized recommendation model in combination with training sample data of the second device, so that feature richness of a training sample of the personalized recommendation model is improved, further, the personalized recommendation model is obtained through training of a training sample with higher feature richness, robustness and breadth of the personalized recommendation model are improved, further, prediction accuracy of the personalized recommendation model is improved, further, a model prediction effect of predicting in combination with multiparty data can be achieved under the condition that only local data is used for prediction through inputting the personalized recommendation model obtained through federal learning, calculation amount during personalized recommendation is reduced, further, response speed during personalized recommendation is improved, and the effect of personalized recommendation is enhanced.
Referring to fig. 3, fig. 3 is a schematic device structure diagram of a hardware running environment according to an embodiment of the present application.
As shown in fig. 3, the personalized recommendation device based on federal learning may include: a processor 1001, such as a CPU, memory 1005, and a communication bus 1002. Wherein a communication bus 1002 is used to enable connected communication between the processor 1001 and a memory 1005. The memory 1005 may be a high-speed RAM memory or a stable memory (non-volatile memory), such as a disk memory. The memory 1005 may also optionally be a storage device separate from the processor 1001 described above.
Optionally, the personalized recommendation device based on federal learning may further include a rectangular user interface, a network interface, a camera, an RF (Radio Frequency) circuit, a sensor, an audio circuit, a WiFi module, and the like. The rectangular user interface may include a Display screen (Display), an input sub-module such as a Keyboard (Keyboard), and the optional rectangular user interface may also include a standard wired interface, a wireless interface. The network interface may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface).
It will be appreciated by those skilled in the art that the federal learning-based personalized recommendation device architecture shown in fig. 3 does not constitute a limitation of federal learning-based personalized recommendation devices, and may include more or fewer components than illustrated, or may combine certain components, or may be a different arrangement of components.
As shown in fig. 3, an operating system, a network communication module, and a federal learning-based personalized recommendation program may be included in a memory 1005, which is a type of computer storage medium. The operating system is a program that manages and controls the personalized recommendation device hardware and software resources based on federal learning, supporting the operation of personalized recommendation programs based on federal learning, as well as other software and/or programs. The network communication module is used to enable communication between components within the memory 1005 and other hardware and software in the federal learning-based personalized recommendation system.
In the personalized recommendation device based on federal learning shown in fig. 3, the processor 1001 is configured to execute a personalized recommendation program based on federal learning stored in the memory 1005, to implement the steps of the personalized recommendation method based on federal learning described above.
The specific implementation mode of the personalized recommendation device based on the federal learning is basically the same as the above-mentioned personalized recommendation method based on the federal learning, and is not repeated here.
The embodiment of the application also provides a personalized recommendation device based on federal learning, which comprises:
The extraction module is used for receiving the uploading data and extracting a target recall set corresponding to the uploading data from a preset recall set storage database;
the prediction module is used for acquiring data to be predicted which corresponds to the uploading data and the target recall together, inputting the data to be predicted into a personalized recommendation model acquired based on the federal learning, and acquiring a model output result;
and the screening module is used for screening the model output result to obtain a personalized recommendation result.
Optionally, the prediction module includes:
the screening unit is used for screening the item data based on the item list to be recommended to obtain item data to be recommended, and setting the user data, the item data to be recommended and the behavior data as the data to be predicted;
the predicting unit is used for inputting the data to be predicted into the personalized recommendation model so as to score and sort the articles to be recommended in the article list to be recommended, and a model output result is obtained.
Optionally, the prediction unit includes:
the scoring subunit is used for inputting the data to be predicted into the personalized recommendation model so as to score the articles to be recommended based on the user data and the behavior data and obtain a scoring result;
And the sorting subunit is used for sorting the articles to be recommended based on the scoring result to obtain a model output result.
Optionally, the personalized recommendation device based on federal learning further comprises:
the sample matching module is used for carrying out sample matching on the second equipment associated with the first equipment to obtain a public training sample ID;
and the first federal learning module is used for obtaining the personalized recommendation model by interacting with the second equipment to perform federal learning based on the public training sample ID.
Optionally, the federal learning module includes:
the first calculating unit is used for extracting first sample data corresponding to the public training sample ID based on the public training sample ID and calculating a first weight corresponding to the first sample data;
the second computing unit is used for receiving the second weight sent by the second device, computing a gradient auxiliary variable which corresponds to the first weight and the second weight together through a preset intermediate parameter formula, and sending the gradient auxiliary variable to the second device, wherein the second device is used for computing a second gradient which corresponds to the gradient auxiliary variable;
A third calculation unit, configured to calculate a first gradient based on the gradient auxiliary variable, and send the first gradient to a preset federal server, where the preset federal server is configured to calculate a federal model total gradient based on the first gradient and a second gradient sent by the second device;
and the iteration updating unit is used for receiving the total federal model gradient fed back by the federal server, and carrying out iteration updating on the local training model of the first equipment based on the total federal model gradient to obtain the logistic regression model.
Optionally, the extracting module includes:
the extraction unit is used for receiving the uploading data and extracting a sample ID in the uploading data;
and the inquiring unit is used for inquiring the corresponding target recall set in the preset recall set storage database based on the sample ID.
Optionally, the personalized recommendation device based on federal learning includes:
the second federation learning module is used for performing federation learning on the second equipment associated with the first equipment to obtain a federation recall algorithm model;
the storage module is used for acquiring sample uploading data, inputting the sample uploading data into the federal recall algorithm model, obtaining a target recall set, and storing the target recall set in the preset recall set storage database.
The specific implementation manner of the personalized recommendation device based on federal learning is basically the same as that of each embodiment of the personalized recommendation method based on federal learning, and is not repeated here.
Embodiments of the present application provide a medium, which is a readable storage medium, and which stores one or more programs, and the one or more programs are further executable by one or more processors to implement the steps of the personalized recommendation method based on federal learning as described in any one of the above.
The specific implementation manner of the medium is basically the same as that of each embodiment of the personalized recommendation method based on federal learning, and is not repeated here.
The foregoing description is only of the preferred embodiments of the present application, and is not intended to limit the scope of the application, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein, or any application, directly or indirectly, within the scope of the application.
Claims (8)
1. The personalized recommendation method based on the federal learning is characterized by being applied to first equipment for performing the federal learning, and comprises the following steps of:
Sample matching is carried out on second equipment associated with the first equipment, and a public training sample ID is obtained;
based on the public training sample ID, extracting first sample data corresponding to the public training sample ID, and calculating a first weight corresponding to the first sample data;
receiving a second weight sent by the second equipment, calculating a gradient auxiliary variable which corresponds to the first weight and the second weight together through a preset intermediate parameter formula, and sending the gradient auxiliary variable to the second equipment, wherein the second equipment is used for calculating a second gradient which corresponds to the gradient auxiliary variable;
calculating a first gradient based on the gradient auxiliary variable, and sending the first gradient to a preset federal server, wherein the preset federal server is used for calculating a federal model total gradient based on the first gradient and a second gradient sent by the second device;
receiving the federal model total gradient fed back by the federal server, and iteratively updating a local training model of the first device based on the federal model total gradient to obtain a logistic regression model;
receiving uploading data, and extracting a target recall set corresponding to the uploading data from a preset recall set storage database;
Obtaining to-be-predicted data which corresponds to the uploading data and the target recall set together, inputting the to-be-predicted data into a personalized recommendation model obtained based on the federal learning, and obtaining a model output result, wherein the personalized recommendation model comprises the logistic regression model;
and screening the model output result to obtain a personalized recommendation result.
2. The federally learned personalized recommendation method according to claim 1, wherein the target recall set comprises a list of items to be recommended, the upload data comprises user data, item data, and behavior data,
the step of obtaining the data to be predicted, which corresponds to the uploading data and the target recall set together, and inputting the data to be predicted into the personalized recommendation model obtained based on the federal learning, and obtaining a model output result comprises the following steps:
screening the item data based on the item list to be recommended to obtain item data to be recommended, and setting the user data, the item data to be recommended and the behavior data as the data to be predicted;
and inputting the data to be predicted into the personalized recommendation model to score and sort the articles to be recommended in the article list to be recommended, so as to obtain a model output result.
3. The personalized recommendation method based on federal learning according to claim 2, wherein the step of inputting the data to be predicted into the personalized recommendation model to score and sort the items to be recommended in the item list to be recommended, and obtaining a model output result comprises:
inputting the data to be predicted into the personalized recommendation model to score the articles to be recommended based on the user data and the behavior data, and obtaining a scoring result;
and sorting the articles to be recommended based on the scoring result to obtain a model output result.
4. The personalized recommendation method based on federal learning of claim 1, wherein the step of receiving the uploaded data and extracting the target recall corresponding to the uploaded data from a preset recall storage database comprises:
receiving uploading data and extracting a sample ID in the uploading data;
and inquiring a corresponding target recall set in the preset recall set storage database based on the sample ID.
5. The personalized recommendation method based on federal learning according to claim 1, wherein the personalized recommendation method based on federal learning is applied to a first device performing federal learning,
The step of receiving the uploading data and extracting the target recall corresponding to the uploading data from the preset recall storage database comprises the following steps:
performing federal learning by second equipment associated with the first equipment to obtain a federal recall algorithm model;
sample uploading data are obtained, the sample uploading data are input into the federation recall algorithm model, a target recall set is obtained, and the target recall set is stored in the preset recall set storage database.
6. The personalized recommendation device based on the federal learning is applied to first equipment for performing the federal learning, and comprises:
the extraction module is used for receiving the uploading data and extracting a target recall set corresponding to the uploading data from a preset recall set storage database;
the prediction module is used for acquiring data to be predicted, which corresponds to the uploading data and the target recall set together, and inputting the data to be predicted into the personalized recommendation model acquired based on the federal learning to acquire a model output result;
The screening module is used for screening the model output result to obtain a personalized recommendation result;
the sample matching module is used for carrying out sample matching with second equipment associated with the first equipment to obtain a public training sample ID;
the first federal learning module is used for obtaining the personalized recommendation model by interacting with the second equipment to perform federal learning based on the public training sample ID;
the personalized recommendation model includes a logistic regression model, and the first federal learning module includes:
the first calculating unit is used for extracting first sample data corresponding to the public training sample ID based on the public training sample ID and calculating a first weight corresponding to the first sample data;
the second computing unit is used for receiving a second weight sent by the second equipment, computing a gradient auxiliary variable which corresponds to the first weight and the second weight together through a preset intermediate parameter formula, and sending the gradient auxiliary variable to the second equipment, wherein the second equipment is used for computing a second gradient which corresponds to the gradient auxiliary variable;
a third calculation unit, configured to calculate a first gradient based on the gradient auxiliary variable, and send the first gradient to a preset federal server, where the preset federal server is configured to calculate a federal model total gradient based on the first gradient and a second gradient sent by the second device;
And the iteration updating unit is used for receiving the total gradient of the federation model fed back by the federation server, and carrying out iteration updating on the local training model of the first equipment based on the total gradient of the federation model to obtain the logistic regression model.
7. A personalized recommendation device based on federal learning, the personalized recommendation device based on federal learning comprising: a memory, a processor and a program stored on the memory for implementing the federal learning-based personalized recommendation method,
the memory is used for storing a program for realizing a personalized recommendation method based on federal learning;
the processor is configured to execute a program implementing the federal learning-based personalized recommendation method to implement the steps of the federal learning-based personalized recommendation method according to any one of claims 1 to 5.
8. A medium having stored thereon a program for implementing a federal learning-based personalized recommendation method, the program for implementing the federal learning-based personalized recommendation method being executed by a processor to implement the steps of the federal learning-based personalized recommendation method according to any one of claims 1 to 5.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911326853.3A CN111079022B (en) | 2019-12-20 | 2019-12-20 | Personalized recommendation method, device, equipment and medium based on federal learning |
| PCT/CN2020/135030 WO2021121106A1 (en) | 2019-12-20 | 2020-12-09 | Federated learning-based personalized recommendation method, apparatus and device, and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911326853.3A CN111079022B (en) | 2019-12-20 | 2019-12-20 | Personalized recommendation method, device, equipment and medium based on federal learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111079022A CN111079022A (en) | 2020-04-28 |
| CN111079022B true CN111079022B (en) | 2023-10-03 |
Family
ID=70316309
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911326853.3A Active CN111079022B (en) | 2019-12-20 | 2019-12-20 | Personalized recommendation method, device, equipment and medium based on federal learning |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111079022B (en) |
| WO (1) | WO2021121106A1 (en) |
Families Citing this family (33)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111079022B (en) * | 2019-12-20 | 2023-10-03 | 深圳前海微众银行股份有限公司 | Personalized recommendation method, device, equipment and medium based on federal learning |
| CN111291417B (en) * | 2020-05-09 | 2020-08-28 | 支付宝(杭州)信息技术有限公司 | Method and device for protecting data privacy of multi-party combined training object recommendation model |
| CN111859136B (en) * | 2020-07-23 | 2024-03-15 | 深圳前海微众银行股份有限公司 | Personalized recommendation method, device, equipment and readable storage medium |
| CN113781134A (en) * | 2020-07-28 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Item recommendation method and device and computer-readable storage medium |
| CN111897800B (en) * | 2020-08-05 | 2021-06-22 | 全球能源互联网研究院有限公司 | Electric vehicle charging facility recommendation method and system based on federal learning |
| CN112287231B (en) * | 2020-11-05 | 2024-04-05 | 深圳大学 | Federal recommendation gradient acquisition method and device, intelligent terminal and storage medium |
| CN112449009B (en) * | 2020-11-12 | 2023-01-10 | 深圳大学 | SVD-based communication compression method and device for Federal learning recommendation system |
| CN112836830B (en) * | 2021-02-01 | 2022-05-06 | 广西师范大学 | Method for voting and training in parallel by using federated gradient boosting decision tree |
| CN112966182B (en) * | 2021-03-09 | 2024-02-09 | 中国民航信息网络股份有限公司 | Project recommendation method and related equipment |
| CN113313268A (en) * | 2021-06-11 | 2021-08-27 | 杭州煋辰数智科技有限公司 | Federal learning-based prediction method and device, storage medium and remote sensing equipment |
| CN113762532B (en) * | 2021-06-28 | 2024-04-05 | 京东科技控股股份有限公司 | Training method and device of federal learning model, electronic equipment and storage medium |
| CN113656681B (en) * | 2021-07-08 | 2023-08-11 | 北京奇艺世纪科技有限公司 | Object evaluation method, device, equipment and storage medium |
| CN113704555B (en) * | 2021-07-16 | 2023-11-07 | 杭州医康慧联科技股份有限公司 | Feature management method based on medical direction federal learning |
| CN113627496A (en) * | 2021-07-27 | 2021-11-09 | 交控科技股份有限公司 | Method, device, electronic equipment and readable storage medium for predicting fault of turnout switch machine |
| CN113642707B (en) * | 2021-08-12 | 2023-08-18 | 深圳平安智汇企业信息管理有限公司 | Model training method, device, equipment and storage medium based on federal learning |
| CN113722738B (en) * | 2021-09-02 | 2023-08-08 | 脸萌有限公司 | Data protection method, device, medium and electronic equipment |
| CN113569151B (en) * | 2021-09-18 | 2021-12-17 | 平安科技（深圳）有限公司 | Data recommendation method, device, equipment and medium based on artificial intelligence |
| CN113902473A (en) * | 2021-09-29 | 2022-01-07 | 支付宝(杭州)信息技术有限公司 | Training method and device of business prediction system |
| CN114363176B (en) * | 2021-12-20 | 2023-08-08 | 中山大学 | Network identification method, device, terminal and medium based on federal learning |
| CN114969503B (en) * | 2022-03-30 | 2024-04-02 | 贵州大学 | Multi-data user portrait realizing method based on federal learning |
| CN114676851B (en) * | 2022-04-08 | 2024-03-29 | 中国科学技术大学 | Combined training method, equipment and storage medium for recall and sequence model |
| CN114741611B (en) * | 2022-06-08 | 2022-10-14 | 杭州金智塔科技有限公司 | Federal recommendation model training method and system |
| CN115169576B (en) * | 2022-06-24 | 2024-02-09 | 上海富数科技有限公司 | Model training method and device based on federal learning and electronic equipment |
| CN114863499B (en) * | 2022-06-30 | 2022-12-13 | 广州脉泽科技有限公司 | Finger vein and palm vein identification method based on federal learning |
| CN114996317B (en) * | 2022-07-05 | 2024-02-23 | 中国电信股份有限公司 | Asynchronous optimization method and device based on longitudinal federal learning and storage medium |
| CN114969486B (en) * | 2022-08-02 | 2022-11-04 | 平安科技（深圳）有限公司 | Corpus recommendation method, apparatus, device and storage medium |
| CN115238065B (en) * | 2022-09-22 | 2022-12-20 | 太极计算机股份有限公司 | Intelligent document recommendation method based on federal learning |
| CN116629374B (en) * | 2023-04-13 | 2023-12-26 | 南京大学 | Searching and multiplexing method for heterogeneous feature space learning piece |
| CN117493662A (en) * | 2023-10-09 | 2024-02-02 | 上海及未科技有限公司 | Personalized recommendation method and system based on federal learning |
| CN117076783B (en) * | 2023-10-16 | 2023-12-26 | 广东省科技基础条件平台中心 | Scientific and technological information recommendation method, device, medium and equipment based on data analysis |
| CN117421486B (en) * | 2023-12-18 | 2024-03-19 | 杭州金智塔科技有限公司 | Recommendation model updating system and method based on spherical tree algorithm and federal learning |
| CN117454185B (en) * | 2023-12-22 | 2024-03-12 | 深圳市移卡科技有限公司 | Federal model training method, federal model training device, federal model training computer device, and federal model training storage medium |
| CN117592042B (en) * | 2024-01-17 | 2024-04-05 | 杭州海康威视数字技术股份有限公司 | Privacy disclosure detection method and device for federal recommendation system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019090954A1 (en) * | 2017-11-07 | 2019-05-16 | 华为技术有限公司 | Prediction method, and terminal and server |
| CN110134868A (en) * | 2019-05-14 | 2019-08-16 | 辽宁工程技术大学 | A kind of recommended method based on the analysis of user preference isomerism |
| 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 |
| CN110428058A (en) * | 2019-08-08 | 2019-11-08 | 深圳前海微众银行股份有限公司 | Federal learning model training method, device, terminal device and storage medium |
| CN110443063A (en) * | 2019-06-26 | 2019-11-12 | 电子科技大学 | The method of the federal deep learning of self adaptive protection privacy |
| CN110458663A (en) * | 2019-08-06 | 2019-11-15 | 上海新共赢信息科技有限公司 | A kind of vehicle recommended method, device, equipment and storage medium |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10102478B2 (en) * | 2015-06-26 | 2018-10-16 | Conduent Business Services, Inc. | Distributed and privacy-preserving prediction method |
| US20190182059A1 (en) * | 2017-12-12 | 2019-06-13 | Facebook, Inc. | Utilizing machine learning from exposed and non-exposed user recall to improve digital content distribution |
| US11093873B2 (en) * | 2018-03-30 | 2021-08-17 | Atlassian Pty Ltd. | Using a productivity index and collaboration index for validation of recommendation models in federated collaboration systems |
| CN110110229B (en) * | 2019-04-25 | 2021-06-04 | 深圳前海微众银行股份有限公司 | Information recommendation method and device |
| CN111079022B (en) * | 2019-12-20 | 2023-10-03 | 深圳前海微众银行股份有限公司 | Personalized recommendation method, device, equipment and medium based on federal learning |
-
2019
- 2019-12-20 CN CN201911326853.3A patent/CN111079022B/en active Active
-
2020
- 2020-12-09 WO PCT/CN2020/135030 patent/WO2021121106A1/en active Application Filing
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019090954A1 (en) * | 2017-11-07 | 2019-05-16 | 华为技术有限公司 | Prediction method, and terminal and server |
| CN110134868A (en) * | 2019-05-14 | 2019-08-16 | 辽宁工程技术大学 | A kind of recommended method based on the analysis of user preference isomerism |
| CN110443063A (en) * | 2019-06-26 | 2019-11-12 | 电子科技大学 | The method of the federal deep learning of self adaptive protection privacy |
| 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 |
| CN110458663A (en) * | 2019-08-06 | 2019-11-15 | 上海新共赢信息科技有限公司 | A kind of vehicle recommended method, device, equipment and storage medium |
| CN110428058A (en) * | 2019-08-08 | 2019-11-08 | 深圳前海微众银行股份有限公司 | Federal learning model training method, device, terminal device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2021121106A1 (en) | 2021-06-24 |
| CN111079022A (en) | 2020-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111079022B (en) | Personalized recommendation method, device, equipment and medium based on federal learning | |
| US10958748B2 (en) | Resource push method and apparatus | |
| US8374914B2 (en) | Advertising using image comparison | |
| CN107577807B (en) | Method and device for pushing information | |
| CN110825956A (en) | Information flow recommendation method and device, computer equipment and storage medium | |
| CN109189931B (en) | Target statement screening method and device | |
| US10606910B2 (en) | Ranking search results using machine learning based models | |
| US11748452B2 (en) | Method for data processing by performing different non-linear combination processing | |
| CN111967914A (en) | User portrait based recommendation method and device, computer equipment and storage medium | |
| CN111914159A (en) | Information recommendation method and terminal | |
| CN111160624A (en) | User intention prediction method, user intention prediction device and terminal equipment | |
| CN110598084A (en) | Object sorting method, commodity sorting device and electronic equipment | |
| CN113378067B (en) | Message recommendation method, device and medium based on user mining | |
| CN115619448A (en) | User loss prediction method and device, computer equipment and storage medium | |
| CN112330382A (en) | Item recommendation method and device, computing equipment and medium | |
| KR102205686B1 (en) | Method and apparatus for ranking candiate character and method and device for inputting character | |
| CN112381236A (en) | Data processing method, device, equipment and storage medium for federal transfer learning | |
| US11062224B2 (en) | Prediction using fusion of heterogeneous unstructured data | |
| JP6680663B2 (en) | Information processing apparatus, information processing method, prediction model generation apparatus, prediction model generation method, and program | |
| US20160124580A1 (en) | Method and system for providing content with a user interface | |
| CN111078998A (en) | Information retrieval method, information retrieval device, storage medium and server | |
| CN116204709A (en) | Data processing method and related device | |
| CN116049536A (en) | Recommendation method and related device | |
| CN113515701A (en) | Information recommendation method and device | |
| US9805097B2 (en) | Method and system for providing a search result |
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 |