CN115936112B - Client portrait model training method and system based on federal learning - Google Patents
Client portrait model training method and system based on federal learning Download PDFInfo
- Publication number
- CN115936112B CN115936112B CN202310018596.7A CN202310018596A CN115936112B CN 115936112 B CN115936112 B CN 115936112B CN 202310018596 A CN202310018596 A CN 202310018596A CN 115936112 B CN115936112 B CN 115936112B
- Authority
- CN
- China
- Prior art keywords
- client
- sample set
- privacy
- parameter information
- module
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application provides a client portrait model training method and system based on federal learning, wherein in the method, a first object and a second object respectively determine training data of the first object and the second object from local data of the first object, and the local model is jointly trained by utilizing different training data to learn more features, so that the accuracy of model training is ensured, and the first object and the second object determine the same sample after privacy processing is carried out on the respective data, so that the data of each party is ensured not to be leaked. And the second object processes the third class feature (such as the feature requiring authorization) into a client class label, realizes indirect utilization of the third class feature without client authorization, further utilizes the client class label and the second class feature (such as the public feature) to perform model training together with the first object, meets the prediction requirement of the high-quality client portrait, and achieves the purpose that the high-quality client portrait can be obtained while meeting the large-scale prediction requirement without client authorization during prediction.
Description
Technical Field
The application relates to the technical field of federal learning, in particular to a client portrait model training method and system based on federal learning.
Background
Currently, some institutions often utilize their own customer data to image (e.g., categorize, rank, score, etc.) customers by modeling them and provide services to the customers based on the image results. However, since the model is mostly built based on business data of clients in institutions, it is difficult to introduce data outside institutions, resulting in low accuracy of the built model.
The federal learning technology can realize multiparty data fusion, and solves the problems to a certain extent. However, at present, the training stage and the prediction stage of the federal learning model generally adopt low-value data without authorization, and the model is low in quality although the model can be applied to a large-scale new customer portrait; or, the training stage and the prediction stage both adopt high-value data requiring client authorization, but the client scope is limited to a small number of authorized new clients, and the large-scale requirements cannot be met.
Disclosure of Invention
The application provides the following technical scheme:
in one aspect, the present application provides a client portrait model training method based on federal learning, including:
The first object determining a first class of characteristics of each sample in the first set of customer samples that it owns, while sending risk preference information to the second object;
the second object determines second class characteristics and third class characteristics of a second client sample in a second client sample set owned by the second object, and determines a client class label corresponding to the second client sample based on the risk preference information and the third class characteristics, wherein the second class characteristics and the third class characteristics are different from the first class characteristics;
the first object performs privacy processing on the first client sample set by adopting a target privacy processing mode to obtain a first privacy sample set, the second object performs privacy processing on the second client sample set by adopting the target privacy processing mode to obtain a second privacy sample set, the same sample part between the first privacy sample set and the second privacy sample set is determined, and the same sample part is sent to the first object;
the first object and the second object respectively acquire the to-be-used characteristics corresponding to the same sample part from the characteristics of the first object and the second object, training data is determined, the training data of the first object comprises a first type of characteristics to be used, and the training data of the second object comprises a second type of characteristics to be used and a client type label;
The first object and the second object train the local model based on training data respectively, obtain parameter information of the local model, encrypt the parameter information of the local model to obtain parameter information ciphertext, and send the parameter information ciphertext to the other party;
the first object and the second object update parameters of the local model based on the parameter information ciphertext from the other party respectively;
and returning to execute the step of training the local models of the first object and the second object based on the training data of the first object and the second object respectively under the condition that the local models of the first object and the second object do not reach the training ending condition.
Optionally, the first object performs privacy processing on the first client sample set by using a target privacy processing mode to obtain a first privacy sample set, the second object performs privacy processing on a second client sample set owned by the first object by using the target privacy processing mode to obtain a second privacy sample set, and determining the same sample part between the first privacy sample set and the second privacy sample set includes:
the second object obtains an encryption key generated by the first object, compresses the second client sample set based on a hash function, adds noise data, and encrypts the noise data by using the encryption key to obtain a second compressed sample set;
The second object sends the second compressed sample set to the first object, so that the first object decrypts the second compressed sample set based on a decryption key corresponding to the encryption key to obtain a second decrypted sample set;
the first object compresses a first customer sample set through a hash function, decrypts the first customer sample set, performs secondary hash to obtain a first privacy sample set, and sends the second decrypted sample set and the first privacy sample set to the second object;
the second object deletes the noise data in the second decryption sample set, and performs secondary hash calculation on the second decryption sample set after deleting the noise data to obtain a second privacy sample set;
the second object determines the same portion of the first and second privacy sample sets.
Optionally, the local model of the first object and the local model of the second object are federal XGBoost models with the same structure, the first object and the second object train their local models based on their training data respectively, obtain parameter information of the local model, encrypt the parameter information of the local model to obtain parameter information ciphertext, and send the parameter information ciphertext to the other party, including:
The second object determines a root node based on training data of the second object, splits the root node based on the training data, and initially determines a local model of the second object;
the second object calculates a first derivative and a second derivative of training data in the local model, performs additive homomorphic encryption on the first derivative and the second derivative of the training data in the local model to obtain a first parameter information ciphertext, and sends the first parameter information ciphertext to the first object;
the first object determines all possible values of all features of training data of the first object as first-order derivatives and second-order derivatives of split points based on the first parameter information ciphertext, performs homomorphic encryption on all possible values of all features of the training data of the first object as first-order derivatives and second-order derivatives of the split points to obtain second parameter information ciphertext, and sends the second parameter information ciphertext to the second object;
the second object determines a global optimal partitioning point based on the second parameter information ciphertext, if the global optimal partitioning point is in the first object, the information of the global optimal partitioning point is sent to the first object, the first object partitions training data of the first object based on the information of the global optimal partitioning point, and a partitioning result is sent to the second object; if the global optimal partitioning point is in the second object, the second object partitions the training data of the second object and does not send a partitioning result;
The second object updates the local model based on the result of the division;
and returning to execute the step of training the local models of the first object and the second object based on the training data of the first object and the second object respectively under the condition that the local models of the first object and the second object do not reach the training ending condition, wherein the step comprises the following steps:
and in the case that the local models of the first object and the second object do not reach the training ending condition, returning to execute the second object to calculate the first derivative and the second derivative of the data in the local model.
Optionally, the training ending condition includes:
the loss function values of the local models of the first object and the second object converge;
or, the maximum tree depth of the local models of the first object and the second object reaches a set depth.
Optionally, the privacy processing of the first client sample set by the first object in a target privacy processing mode includes:
the first object randomly selects M groups of client samples from a first client sample set, wherein the total amount of the M groups of client samples is more than N;
the first object sends codes of the M groups of client samples to the second object, so that the second object randomly extracts M client samples from each group of client samples in the M groups of client samples, determines client class labels corresponding to each client sample in the M client samples based on the risk preference information, and determines distribution rationality of the M client samples based on the client class labels, wherein M is not less than N;
If the distribution rationality of the M client samples meets the rationality requirement, determining the M groups of client samples as a first client sample set to be used;
and the first object adopts a target privacy processing mode to carry out privacy processing on the first client sample set to be used.
Another aspect of the present application provides a federally learning-based customer representation model training system, comprising:
a first object module for determining a first class characteristic of each first client sample in a set of first client samples owned by the first object module, and simultaneously transmitting risk preference information to a second object module;
the second object module is configured to determine a second class feature and a third class feature of a second client sample in a second client sample set owned by the second object module, and determine a client class label corresponding to the second client sample based on the risk preference information and the third class feature, where the second class feature and the third class feature are different from the first class feature;
the first object module is further configured to perform privacy processing on the first client sample set by using a target privacy processing manner, so as to obtain a first privacy sample set;
the second object module is further configured to perform privacy processing on the second client sample set by using the target privacy processing manner, obtain a second privacy sample set, determine a same sample portion between the first privacy sample set and the second privacy sample set, and send the same sample portion to the first object;
The first object module and the second object module are also respectively used for acquiring the to-be-used characteristics corresponding to the same sample part from the characteristics of the first object module and the second object module, determining training data, wherein the training data of the first object comprises a first type of characteristics to be used, and the training data of the second object comprises a second type of characteristics to be used and a client type label;
the first object module and the second object module are also respectively used for training the local model based on training data thereof, obtaining parameter information of the local model, encrypting the parameter information of the local model to obtain a parameter information ciphertext, and sending the parameter information ciphertext to the other party;
the first object module and the second object module are also respectively used for updating the parameters of the local model based on the parameter information ciphertext from the other party;
and under the condition that the local models of the first object module and the second object module do not reach the training ending condition, respectively returning and executing the steps of training the local models based on the training data of the first object module and the second object module.
Optionally, the first object module performs privacy processing on the first client sample set by using a target privacy processing mode to obtain a first privacy sample set, and the second object module performs privacy processing on a second client sample set owned by the first object module by using the target privacy processing mode to obtain a second privacy sample set, so as to determine the same sample part between the first privacy sample set and the second privacy sample set, which specifically includes:
the second object module obtains an encryption key generated by the first object, compresses the second client sample set based on a hash function, adds noise data, and encrypts the noise data by using the encryption key to obtain a second compressed sample set;
the second object module sends the second compressed sample set to the first object module, so that the first object module decrypts the second compressed sample set based on a decryption key corresponding to the encryption key to obtain a second decrypted sample set;
the first object module compresses a first customer sample set through a hash function, decrypts the first customer sample set, performs secondary hash to obtain a first privacy sample set, and sends the second decrypted sample set and the first privacy sample set to the second object module;
The second object module deletes the noise data in the second decryption sample set, and performs secondary hash calculation on the second decryption sample set after deleting the noise data to obtain a second privacy sample set;
the second object module determines the same portion of the first and second privacy sample sets.
Optionally, the local model of the first object module and the local model of the second object module are federal XGBoost models with the same structure, the first object module and the second object module train their local models based on their training data respectively, obtain parameter information of the local model, encrypt the parameter information of the local model to obtain parameter information ciphertext, and send the parameter information ciphertext to the other party, including:
the second object module determines a root node based on training data of the second object module, splits the root node based on the training data and initially determines a local model of the second object;
the second object module calculates a first derivative and a second derivative of training data in the local model, performs additive homomorphic encryption on the first derivative and the second derivative of the training data in the local model to obtain a first parameter information ciphertext, and sends the first parameter information ciphertext to the first object;
The first object module determines all possible values of all features of training data of the first object as first-order derivatives and second-order derivatives of split points based on the first parameter information ciphertext, performs homomorphic encryption on all possible values of all features of the training data of the first object as first-order derivatives and second-order derivatives of the split points to obtain second parameter information ciphertext, and sends the second parameter information ciphertext to the second object module;
the second object module determines a global optimal partitioning point based on the second parameter information ciphertext, if the global optimal partitioning point is in the first object module, the information of the global optimal partitioning point is sent to the first object module, so that the first object module partitions training data of the first object module based on the information of the global optimal partitioning point, and sends a partitioning result to the second object module;
the second object module updates the local model based on the division result;
and in the case that the local models of the first object module and the second object module do not reach the training ending condition, returning to execute the second object module to calculate the first derivative and the second derivative of the data in the local model.
Optionally, the training ending condition includes:
the loss function values of the local models of the first object module and the second object module converge;
or, the maximum depth of the tree of the local model of the first object module and the second object module reaches the set depth.
Optionally, the privacy processing of the first client sample set by the first object module in the target privacy processing mode specifically includes:
randomly selecting M groups of customer samples from a first customer sample set, wherein the total amount of the M groups of customer samples is greater than N;
transmitting the codes of the M groups of client samples to the second object, so that the second object randomly extracts M client samples from each group of client samples in the M groups of client samples, determines client class labels corresponding to each client sample in the M client samples based on the risk preference information, and determines distribution rationality of the M client samples based on the client class labels, wherein M is not less than N;
if the distribution rationality of the M client samples meets the rationality requirement, determining the M groups of client samples as a first client sample set to be used;
And carrying out privacy processing on the first customer sample set to be used by adopting a target privacy processing mode.
In the method, the training data of the first object and the second object are determined from the local data of the first object and the second object respectively, and the local model is jointly trained by utilizing different training data, so that the local model of the first object and the local model of the second object can learn more characteristics, the accuracy of model training is ensured, and the first object and the second object determine the same part after privacy processing is carried out on the respective data, and the data of each party is ensured not to be leaked.
And the second object can combine the parameters of the local model of the first object, train the model based on the client class label and the second class feature (such as the public feature), train with the first object together without directly utilizing the third class feature (such as the authorization feature), meet the prediction requirement of the high-quality client portrait, meet the large-scale prediction requirement without client authorization during prediction, and simultaneously obtain the purpose of the high-quality client portrait.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort to a person skilled in the art.
FIG. 1 is a flow chart of a client representation model training method based on federal learning according to a first embodiment of the present application;
FIG. 2 is a schematic flow chart of a client representation model training method based on federal learning according to a second embodiment of the present application;
FIG. 3 is a schematic flow chart of a client representation model training method based on federal learning according to a third embodiment of the present application;
FIG. 4 is a schematic diagram of a client representation model training system based on federal learning.
Detailed Description
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
Referring to fig. 1, a flowchart of a client portrait model training method based on federal learning according to a first embodiment of the present application is shown in fig. 1, where the method may include, but is not limited to, the following steps:
In step S11, the first object determines a first class feature of each first client sample in the first client sample set owned by the first object, and simultaneously sends risk preference information to the second object.
The first object, can be understood as: an object with portrait forecast requirements, such as a banking institution.
It should be noted that, the risk preference information provided by the first object is determined in advance by the first object and the second object according to the third type of features of the second object, and the risk preference information is irrelevant to the first type of features.
Step S12, the second object determines a second class feature and a third class feature of a second client sample in a second client sample set owned by the second object, and determines a client class label corresponding to the second client sample based on the risk preference information and the third class feature, where the second class feature and the third class feature are different from the first class feature.
The second class of characteristics of the second customer sample may include, but is not limited to: the second customer sample has a public feature.
A third class of features of the second customer sample may include, but is not limited to: high value characteristics of the second customer sample.
It will be appreciated that the overt feature is a feature that can overt acquired data and the high value feature is a feature that requires customer authorization to overt.
The second object may be, but is not limited to: a data provider, such as a government agency.
In this embodiment, determining, based on the risk preference information and the third class feature, a client class label corresponding to the second client sample may include, but is not limited to: and determining a client rating rule based on the risk preference information, and determining a client category label corresponding to the second client sample based on the client rating rule and the third class feature. The customer category label may characterize a rating of the customer. Specifically, the second object groups the second client samples in the second client sample set according to the industry to obtain S groups of client samples. Customer patterns for a groupPresent q, according to its high value feature A i Ranking is evenly divided into 10 grades (1-10, 1 grade represents the highest category, 10 grade represents the lowest category, and the number of enterprises in each grade is the same) to obtain the grade D of each customer sample under the characteristic q (A i ) Other features do so as well. The final grade of the customer sample is the minimum grade D in all the characteristics of the group q =Min(D q (A 1 ),D q (A 2 ),...,D q (A Na ))，N a Is the feature number. And then mixing the labels of the S groups of customer samples to form a customer category label corresponding to one second customer sample.
Step S13, the first object performs privacy processing on the first client sample set by using a target privacy processing mode to obtain a first privacy sample set, the second object performs privacy processing on the second client sample set by using the target privacy processing mode to obtain a second privacy sample set, determines the same sample part between the first privacy sample set and the second privacy sample set, and sends the same sample part to the first object.
Step S14, the first object and the second object respectively acquire the features to be used corresponding to the same sample portion from the features owned by the first object and the second object, and determine training data, where the training data of the first object includes a first type of features to be used, and the training data of the second object includes a second type of features to be used and a client type tag.
The first object takes the first type of to-be-used characteristics in the to-be-used characteristics corresponding to the same sample part as training data of the first object.
And the second object takes the second type of characteristics to be used and the client type label in the acquired characteristics to be used corresponding to the same sample part as training data of the second object.
And step S15, the first object and the second object train the local model based on training data respectively, obtain parameter information of the local model, encrypt the parameter information of the local model to obtain a parameter information ciphertext, and send the parameter information ciphertext to the other party.
In this embodiment, the local model of the first object and the local model of the second object may be, but are not limited to, federal XGBoost models with the same structure. Accordingly, this step may include, but is not limited to:
s151, the second object determines a root node based on training data of the second object, splits the root node based on the training data, and initially determines a local model of the second object.
S152, the second object calculates a first derivative and a second derivative of training data in the local model, performs additive homomorphic encryption on the first derivative and the second derivative of the training data in the local model to obtain a first parameter information ciphertext, and sends the first parameter information ciphertext to the first object.
The first derivative of one of the training data in the local model of the second object may be expressed asThe second derivative of the training data may be expressed as. The second object can calculate the sum of the derivatives of the training data in its local model to obtain the total derivativeAndthen willAndcarrying out additive homomorphic encryption to obtainAndwill beAndand sending the first object to each first object.
S153, the first object determines all possible values of all features of training data of the first object as a first derivative and a second derivative of a splitting point based on the first parameter information ciphertext, homomorphic encryption is performed on all possible values of all features of the training data of the first object as the first derivative and the second derivative of the splitting point, a second parameter information ciphertext is obtained, and the second parameter information ciphertext is sent to the second object.
The homomorphic encrypted first derivative in the second parameter information ciphertext can be expressed as
The homomorphic encrypted second derivative in the second parameter information ciphertext can be expressed as. Wherein, the liquid crystal display device comprises a liquid crystal display device,represents the mth first objectAll possible values of each feature k of the training data are taken as the first derivative of the split point,represents the mth first object All possible values of each feature k of the training data are taken as the second derivatives of the split points,represents the mth first objectThe ith possible value of each feature k of the training data is taken as the first derivative of the split point,represents the mth first objectThe ith possible value of each feature k of the training data is taken as the second derivative of the split point,representing all possible values for each feature k.
It will be appreciated that the additive homomorphic encryption property isThat is, the addition result of the data in the encryption state is consistent with the result of the data addition and then encryption, thereby ensuring the correct resultAnd the derivative is not decrypted by the first object.
S154, the second object determines a global optimal partitioning point based on the second parameter information ciphertext, if the global optimal partitioning point is in the first object, the information of the global optimal partitioning point is sent to the first object, the first object partitions training data of the first object based on the information of the global optimal partitioning point, and a partition result is sent to the second object; if the global optimal partitioning point is in the second object, the second object partitions the training data of the second object and does not send a partitioning result;
Specifically, the second object decrypts the second parameter information text to obtainAndall possible gains are calculated, and for a certain feature a certain valued gain is calculated as:
，，，
and comparing all the gains, and selecting the characteristics and the values corresponding to the optimal gain as global optimal dividing points.
Based onObtaining [ place, feature and value of global optimal partition point ]]。
For the iterative formula to be a solution,to the right of the medium number is the historical optimal gain of the previous iteration round, gain is the current iteration roundThe max () function is used to take the maximum value of both, and assign the maximum value toTo the left of the medium number. Through iteration, the [ place party, feature and value of the global optimal partition point are obtained]。
Specifically, if the global optimal partition point is at the second object, no transmission is made. If the global optimal partition point is in the first object, the second object sends the [ feature value ] to the corresponding first object.
The object with the characteristic divides the data according to the global optimal dividing point, and sends the dividing result to the second object, and meanwhile, a lookup table is built in the second object, and the [ characteristic, value ] is recorded.
And S155, updating the local model of the second object based on the division result.
Specifically, the second object splits the current node according to the received division result, and records the object to which the characteristics of the current node belong.
And step S16, the first object and the second object update the parameters of the local model based on the parameter information ciphertext from the other party respectively.
And step S17, returning to the step of executing the local model of the first object and the second object to train the local model of the first object and the second object respectively based on the training data of the first object and the second object when the local model of the first object and the second object does not reach the training ending condition.
In this embodiment, the training end condition may include, but is not limited to:
the loss function values of the local models of the first object and the second object converge;
or, the maximum tree depth of the local models of the first object and the second object reaches a set depth.
And ending training when the local models of the first object and the second object reach the training ending condition, and taking the trained local model as a prediction model.
When the prediction model of the first object is applied, the first type characteristics of the client to be predicted can be input into the prediction model of the first object to obtain the client type label of the client to be predicted, which is predicted by the prediction model of the first object, as a rating result.
When the prediction model of the second object is applied, the second class feature of the client to be predicted can be input into the prediction model of the second object to obtain the client class label of the client to be predicted by the prediction model of the second object, and the client class label is used as a rating result.
In this embodiment, the first object and the second object determine their training data from their local data, and jointly train the local model using different training data, so that the local models of the first object and the second object can learn more features, and ensure accuracy of model training.
And the second object can combine the parameters of the local model of the first object, train the model based on the client class label and the second class feature (such as the public feature), train with the first object together without directly utilizing the third class feature (such as the authorization feature), meet the prediction requirement of the high-quality client portrait, meet the large-scale prediction requirement without client authorization during prediction, and simultaneously obtain the purpose of the high-quality client portrait.
As another optional embodiment of the present application, referring to fig. 2, a flowchart of a client portrait model training method based on federal learning according to a second embodiment of the present application is mainly a refinement of the model training method based on federal learning described in the first embodiment, as shown in fig. 2, where the method may include, but is not limited to, the following steps:
in step S21, the first object determines a first class feature of each first client sample in the first client sample set owned by the first object, and simultaneously sends risk preference information to the second object.
Step S22, the second object determines a second class feature and a third class feature of a second client sample in a second client sample set owned by the second object, and determines a client class label corresponding to the second client sample based on the risk preference information and the third class feature, where the second class feature and the third class feature are different from the first class feature.
The detailed process of steps S21 to S22 can be referred to the related description of steps S11 to S12 in the first embodiment, and will not be described herein.
Step S23, the second object obtains an encryption key generated by the first object, compresses the second client sample set based on a hash function, adds noise data, and encrypts the noise data by using the encryption key to obtain a second compressed sample set.
In this embodiment, the first object may generate an encryption key and a decryption key, which are in one-to-one correspondence, and may be interchanged. In particular, the first object may, but is not limited to, generate an encryption key and a decryption key based on the RSA algorithm.
The first object sends the encryption key to the second object. Accordingly, the second object receives the encryption key generated by the first object.
Specifically, the second compressed sample set may be derived based on the following formula:
where Z1 represents a second compressed sample set, R represents noise data, en represents an encryption key, hash () represents a Hash function, U represents a second client sample, and U represents a set containing the second client sample.
Step S24, the second object sends the second compressed sample set to the first object, so that the first object decrypts the second compressed sample set based on the decryption key corresponding to the encryption key, and obtains a second decrypted sample set.
In this embodiment, the first object may obtain the second decrypted sample set based on the following formula:
wherein Z2 represents a second set of decrypted samples, R represents the noise data, hash () represents a Hash function, de represents the decryption key, U represents a second set of client samples, and U represents a set of client samples containing the second set of client samples.
Step S25, the first object compresses the first client sample set through a hash function, decrypts the first client sample set, hashes the first client sample set for a second time to obtain a first privacy sample set, and sends the second decrypted sample set and the first privacy sample set to the second object.
In this embodiment, the first object may determine the first privacy sample set based on the following formula:
where B represents the first set of privacy samples, hash () represents a Hash function, de represents the decryption key, V represents the first set of customer samples, and V represents the first set of customer samples.
And S26, deleting the noise data in the second decryption sample set by the second object, and performing secondary hash calculation on the second decryption sample set after deleting the noise data to obtain a second privacy sample set.
In this embodiment, the second privacy sample set may be obtained based on the following formula:
where Z3 represents the second set of privacy samples, hash () represents the Hash function, de represents the decryption key, U represents the second set of client samples, and U represents the second set of client samples.
Step S27, the second object determines the same portion of the first and second privacy sample sets.
In this embodiment, the whole process of steps S23 to S27 does not have plaintext transfer, and the original customer sample cannot be analyzed. The two parties obtain the same part without knowing all customer samples of the other party, and the difference parts of the two parties are well protected.
Steps S23 to S27 are a specific embodiment of step S13 in example 1.
Step S28, the first object and the second object respectively acquire the features to be used corresponding to the same sample portion from the features owned by the first object and the second object, and determine training data, where the training data of the first object includes a first type of features to be used, and the training data of the second object includes a second type of features to be used and a client type tag.
Step 29, the first object and the second object train the local model based on the training data respectively, obtain the parameter information of the local model, encrypt the parameter information of the local model to obtain the parameter information ciphertext, and send the parameter information ciphertext to the other party.
Step S210, the first object and the second object update the parameters of the local model based on the parameter information ciphertext from the other party, respectively.
Step S211, when the local models of the first object and the second object do not reach the training end condition, returning to execute the step of training the local models of the first object and the second object based on the training data of the first object and the second object, respectively.
The detailed process of steps S28 to S211 can be referred to the description of steps S14 to S17 in the first embodiment, and will not be repeated here.
As another optional embodiment of the present application, referring to fig. 3, a flowchart of a client portrait model training method based on federal learning according to a third embodiment of the present application is provided, where this embodiment is mainly a refinement of the model training method based on federal learning described in the foregoing first embodiment, as shown in fig. 3, and the method may include, but is not limited to, the following steps:
in step S31, the first object determines a first class feature of each first client sample in the first client sample set owned by the first object, and simultaneously sends risk preference information to the second object.
Step S32, the second object determines a second class feature and a third class feature of a second client sample in a second client sample set owned by the second object, and determines a client class label corresponding to the second client sample based on the risk preference information and the third class feature, where the second class feature and the third class feature are different from the first class feature.
The detailed process of steps S31 to S32 can be referred to the related description of steps S11 to S12 in the first embodiment, and will not be described herein.
Step S33, the first object randomly selects M groups of customer samples from the first customer sample set, wherein the total amount of the M groups of customer samples is greater than N.
Step S34, the first object sends the codes of the M groups of client samples to the second object, so that the second object randomly extracts M client samples from each group of client samples in the M groups of client samples, determines a client class label corresponding to each client sample in the M client samples based on the risk preference information, and determines distribution rationality of the M client samples based on the client class label, wherein M is not less than N.
In this embodiment, the encoding of the M sets of client samples may be, but is not limited to: the enterprises are numbered in a unified way.
And step S35, if the distribution rationality of the M client samples meets the rationality requirement, determining the M groups of client samples as a first client sample set to be used.
Step S36, the first object performs privacy processing on the first client sample set to be used by adopting a target privacy processing mode, so as to obtain a first privacy sample set.
Step S37, the second object performs privacy processing on the second client sample set by using the target privacy processing mode, so as to obtain a second privacy sample set, determine the same sample portion between the first client sample set and the second client sample set, and send the same sample portion to the first object.
Steps S33 to S37 are a specific embodiment of step S13 in example 1.
Step S38, the first object and the second object respectively acquire the features to be used corresponding to the same sample part from the features owned by the first object and the second object, and training data is determined, wherein the training data of the first object comprises the first type of features to be used, and the training data of the second object comprises the second type of features to be used and customer type labels.
Step S39, the first object and the second object train the local model based on training data respectively, obtain parameter information of the local model, encrypt the parameter information of the local model to obtain parameter information ciphertext, and send the parameter information ciphertext to the other party.
Step S310, the first object and the second object update the parameters of the local model based on the parameter information ciphertext from the other party, respectively.
Step S311, when the local models of the first object and the second object do not reach the training end condition, returns to execute the step of training the local models of the first object and the second object based on the training data thereof, respectively.
The detailed process of steps S38 to S311 can be referred to the related description of steps S14 to S17 in the first embodiment, and will not be described herein.
In this embodiment, the first object and the second object determine their training data from their local data, and jointly train the local model by using different training data, so that the local models of the first object and the second object can learn more features, and the accuracy of model training is ensured.
And the second object can combine the parameters of the local model of the first object, train the model based on the client class label and the second class feature (such as the public feature), not directly train by using the third class feature (such as the authorization feature), so as to meet the prediction requirement of the high-quality client portrait, and achieve the purpose of meeting the large-scale prediction requirement without client authorization during prediction and simultaneously obtaining the high-quality client portrait.
Further, the first object randomly selects M groups of client samples, the total amount of the M groups of client samples is greater than N, the first object sends first class features of the M groups of client samples to the second object, so that the second object randomly extracts first class features of M client samples from the first class features of each group of client samples in the M groups of client samples, determines client class labels corresponding to each client sample in the M client samples based on the first class features of the M client samples, determines distribution rationality of the M client samples based on the client class labels corresponding to each client sample in the M client samples, and further improves model training accuracy if the distribution rationality of the M client samples meets rationality requirements.
Corresponding to the embodiment of the client portrait model training method based on federal learning provided by the application, the application also provides an embodiment of the client portrait model training system based on federal learning.
In this embodiment, as shown in fig. 4, the client portrait model training system based on federal learning includes: a first object module 100 and a second object module 200.
A first object module 100 for determining a first type of characteristic of each first client sample in a set of first client samples it owns, while transmitting risk preference information to a second object module 200;
the second object module 200 is configured to determine a second class feature and a third class feature of a second client sample in a second client sample set owned by the second object module, and determine a client class label corresponding to the second client sample based on the risk preference information and the third class feature, where the second class feature and the third class feature are different from the first class feature;
the first object module 100 is further configured to perform privacy processing on the first client sample set by using a target privacy processing manner, so as to obtain a first privacy sample set;
the second object module 200 is further configured to perform privacy processing on the second client sample set by using the target privacy processing manner, obtain a second privacy sample set, determine a same sample portion between the first privacy sample set and the second privacy sample set, and send the same sample portion to the first object;
the first object module 100 and the second object module 200 are further configured to obtain, from the features owned by each object module, the features to be used corresponding to the same sample portion, and determine training data, where the training data of the first object includes a first type of feature to be used, and the training data of the second object includes a second type of feature to be used and a client type tag;
The first object module 100 and the second object module 200 are further configured to train a local model based on training data thereof, obtain parameter information of the local model, encrypt the parameter information of the local model to obtain a parameter information ciphertext, and send the parameter information ciphertext to a counterpart;
the first object module 100 and the second object module 200 are further configured to update parameters of the local model based on the parameter information ciphertext from the other party, respectively;
in the case where the local models of the first object module 100 and the second object module 200 do not reach the training end condition, the first object module 100 and the second object module 200 return to perform the step of training the local models thereof based on the training data thereof, respectively.
In this embodiment, the first object module 100 performs privacy processing on the first client sample set by using a target privacy processing manner to obtain a first privacy sample set, the second object module 200 performs privacy processing on a second client sample set owned by the first object module by using the target privacy processing manner to obtain a second privacy sample set, and a process of determining the same sample portion between the first privacy sample set and the second privacy sample set specifically includes
The second object module 200 obtains an encryption key generated by the first object, compresses the second client sample set based on a hash function, adds noise data, and encrypts the noise data by using the encryption key to obtain a second compressed sample set;
the second object module 200 sends the second compressed sample set to the first object module 100, so that the first object module 100 decrypts the second compressed sample set based on a decryption key corresponding to the encryption key to obtain a second decrypted sample set;
the first object module 100 compresses the first client sample set through a hash function, decrypts the first client sample set, performs a second hash to obtain a first privacy sample set, and sends the second decrypted sample set and the first privacy sample set to the second object module 200;
the second object module 200 deletes the noise data in the second decrypted sample set, and performs a second hash calculation on the second decrypted sample set after deleting the noise data to obtain a second privacy sample set;
the second object module 200 determines the same portion of the first and second privacy sample sets.
In this embodiment, the local model of the first object module 100 and the local model of the second object module 200 are federal XGBoost models with the same structure, the first object module 100 and the second object module 200 respectively train the local models based on training data thereof, obtain parameter information of the local model, encrypt the parameter information of the local model to obtain parameter information ciphertext, and send the parameter information ciphertext to the other party, including:
the second object module 200 determines a root node based on training data thereof, splits the root node based on the training data, and initially determines a local model of the second object;
the second object module 200 calculates a first derivative and a second derivative of the training data in the local model, performs additive homomorphic encryption on the first derivative and the second derivative of the training data in the local model to obtain a first parameter information ciphertext, and sends the first parameter information ciphertext to the first object;
the first object module 100 determines all possible values of each feature of the training data of the first object as a first derivative and a second derivative of the split point based on the first parameter information ciphertext, performs homomorphic encryption on all possible values of each feature of the training data of the first object as the first derivative and the second derivative of the split point to obtain a second parameter information ciphertext, and sends the second parameter information ciphertext to the second object module 200;
The second object module 200 determines a global optimal partition point based on the second parameter information ciphertext, if the global optimal partition point is in the first object module 100, sends information of the global optimal partition point to the first object module 100, so that the first object module 100 divides training data of the first object module 100 based on the information of the global optimal partition point, and sends a division result to the second object module 200;
the second object module 200 updates the local model thereof based on the result of the division;
in case the local model of the first object module 100 and the second object module 200 does not reach the training end condition, the second object module 200 is executed back to calculate the first and second derivatives of the data in the local model thereof.
In this embodiment, the training end condition includes:
the loss function values of the local models of the first object module 100 and the second object module 200 converge;
or, the maximum depth of the tree of the local model of the first object module 100 and the second object module 200 reaches a set depth.
In this embodiment, the process of privacy processing of the first client sample set by the first object module 100 in the target privacy processing manner specifically includes:
randomly selecting M groups of customer samples from a first customer sample set, wherein the total amount of the M groups of customer samples is greater than N;
transmitting the codes of the M groups of client samples to the second object, so that the second object randomly extracts M client samples from each group of client samples in the M groups of client samples, determines a client class label corresponding to each client sample in the M client samples based on the risk preference information, and determines the distribution rationality of the M client samples based on the client class labels, wherein M is not less than N;
if the distribution rationality of the M client samples meets the rationality requirement, determining the M groups of client samples as a first client sample set to be used;
and carrying out privacy processing on the first customer sample set to be used by adopting a target privacy processing mode.
It should be noted that, in each embodiment, the differences from the other embodiments are emphasized, and the same similar parts between the embodiments are referred to each other. For the apparatus class embodiments, the description is relatively simple as it is substantially similar to the method embodiments, and reference is made to the description of the method embodiments for relevant points.
Finally, it is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
For convenience of description, the above devices are described as being functionally divided into various units, respectively. Of course, the functions of each element may be implemented in one or more software and/or hardware elements when implemented in the present application.
From the above description of embodiments, it will be apparent to those skilled in the art that the present application may be implemented in software plus a necessary general purpose hardware platform. Based on such understanding, the technical solutions of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a storage medium, such as a ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform the methods described in the embodiments or some parts of the embodiments of the present application.
The foregoing has described in detail a federal learning-based client image model training method and system, and specific examples have been used herein to illustrate the principles and embodiments of the present application, the above examples being provided only to assist in understanding the method and core ideas of the present application; meanwhile, as those skilled in the art will have modifications in the specific embodiments and application scope in accordance with the ideas of the present application, the present description should not be construed as limiting the present application in view of the above.
Claims (10)
1. A federal learning-based customer representation model training method, comprising:
the first object determining a first class of characteristics of each sample in the first set of customer samples that it owns, while sending risk preference information to the second object;
the second object determines second class characteristics and third class characteristics of a second client sample in a second client sample set owned by the second object, and determines a client class label corresponding to the second client sample based on the risk preference information and the third class characteristics, wherein the second class characteristics and the third class characteristics are different from the first class characteristics, the second class characteristics comprise public characteristics of the second client sample, and the public characteristics are data which can be obtained in a public way; the third class of features includes high-value features of the second customer sample, the high-value features being features that require customer authorization to be disclosed;
The first object performs privacy processing on the first client sample set by adopting a target privacy processing mode to obtain a first privacy sample set, the second object performs privacy processing on the second client sample set by adopting the target privacy processing mode to obtain a second privacy sample set, the same sample part between the first privacy sample set and the second privacy sample set is determined, and the same sample part is sent to the first object;
the first object and the second object respectively acquire the to-be-used characteristics corresponding to the same sample part from the characteristics of the first object and the second object, training data is determined, the training data of the first object comprises a first type of characteristics to be used, and the training data of the second object comprises a second type of characteristics to be used and a client type label;
the first object and the second object train the local model based on training data respectively, obtain parameter information of the local model, encrypt the parameter information of the local model to obtain parameter information ciphertext, and send the parameter information ciphertext to the other party;
the first object and the second object update parameters of the local model based on the parameter information ciphertext from the other party respectively;
And returning to execute the step of training the local models of the first object and the second object based on the training data of the first object and the second object respectively under the condition that the local models of the first object and the second object do not reach the training ending condition.
2. The method according to claim 1, wherein the first object performs privacy processing on the first client sample set by using a target privacy processing manner to obtain a first privacy sample set, the second object performs privacy processing on a second client sample set owned by the second object by using the target privacy processing manner to obtain a second privacy sample set, and determining the same sample portion between the first privacy sample set and the second privacy sample set includes:
the second object obtains an encryption key generated by the first object, compresses the second client sample set based on a hash function, adds noise data, and encrypts the noise data by using the encryption key to obtain a second compressed sample set;
the second object sends the second compressed sample set to the first object, so that the first object decrypts the second compressed sample set based on a decryption key corresponding to the encryption key to obtain a second decrypted sample set;
The first object compresses a first customer sample set through a hash function, decrypts the first customer sample set, performs secondary hash to obtain a first privacy sample set, and sends the second decrypted sample set and the first privacy sample set to the second object;
the second object deletes the noise data in the second decryption sample set, and performs secondary hash calculation on the second decryption sample set after deleting the noise data to obtain a second privacy sample set;
the second object determines the same portion of the first and second privacy sample sets.
3. The method according to claim 1, wherein the local model of the first object and the local model of the second object are federal XGBoost models with the same structure, the first object and the second object train their local models based on training data thereof, respectively, and obtain parameter information of the local model, encrypt the parameter information of the local model to obtain parameter information ciphertext, and send the parameter information ciphertext to the other party, including:
the second object determines a root node based on training data of the second object, splits the root node based on the training data, and initially determines a local model of the second object;
The second object calculates a first derivative and a second derivative of training data in the local model, performs additive homomorphic encryption on the first derivative and the second derivative of the training data in the local model to obtain a first parameter information ciphertext, and sends the first parameter information ciphertext to the first object;
the first object determines all possible values of all features of training data of the first object as first-order derivatives and second-order derivatives of split points based on the first parameter information ciphertext, performs homomorphic encryption on all possible values of all features of the training data of the first object as first-order derivatives and second-order derivatives of the split points to obtain second parameter information ciphertext, and sends the second parameter information ciphertext to the second object;
the second object determines a global optimal partitioning point based on the second parameter information ciphertext, if the global optimal partitioning point is in the first object, the information of the global optimal partitioning point is sent to the first object, the first object partitions training data of the first object based on the information of the global optimal partitioning point, and a partitioning result is sent to the second object; if the global optimal partitioning point is in the second object, the second object partitions the training data of the second object and does not send a partitioning result;
The second object updates the local model based on the result of the division;
and returning to execute the step of training the local models of the first object and the second object based on the training data of the first object and the second object respectively under the condition that the local models of the first object and the second object do not reach the training ending condition, wherein the step comprises the following steps:
and in the case that the local models of the first object and the second object do not reach the training ending condition, returning to execute the second object to calculate the first derivative and the second derivative of the data in the local model.
4. A method according to claim 3, wherein the training end condition comprises:
the loss function values of the local models of the first object and the second object converge;
or, the maximum tree depth of the local models of the first object and the second object reaches a set depth.
5. The method of claim 1, wherein the first object privacy-processes the first set of customer samples using a target privacy process, comprising:
the first object randomly selects M groups of client samples from a first client sample set, wherein the total amount of the M groups of client samples is more than N;
The first object sends codes of the M groups of client samples to the second object, so that the second object randomly extracts M client samples from each group of client samples in the M groups of client samples, determines client class labels corresponding to each client sample in the M client samples based on the risk preference information, and determines distribution rationality of the M client samples based on the client class labels, wherein M is not less than N;
if the distribution rationality of the M client samples meets the rationality requirement, determining the M groups of client samples as a first client sample set to be used;
and the first object adopts a target privacy processing mode to carry out privacy processing on the first client sample set to be used.
6. A federal learning-based customer representation model training system, comprising:
a first object module for determining a first class characteristic of each first client sample in a set of first client samples owned by the first object module, and simultaneously transmitting risk preference information to a second object module;
the second object module is configured to determine a second class feature and a third class feature of a second client sample in a second client sample set owned by the second object module, determine a client class label corresponding to the second client sample based on the risk preference information and the third class feature, where the second class feature and the third class feature are different from the first class feature, and the second class feature includes a public feature of the second client sample, where the public feature is data that can be obtained in a public manner; the third class of features includes high-value features of the second customer sample, the high-value features being features that require customer authorization to be disclosed;
The first object module is further configured to perform privacy processing on the first client sample set by using a target privacy processing manner, so as to obtain a first privacy sample set;
the second object module is further configured to perform privacy processing on the second client sample set by using the target privacy processing manner, obtain a second privacy sample set, determine a same sample portion between the first privacy sample set and the second privacy sample set, and send the same sample portion to the first object;
the first object module and the second object module are also respectively used for acquiring the to-be-used characteristics corresponding to the same sample part from the characteristics of the first object module and the second object module, determining training data, wherein the training data of the first object comprises a first type of characteristics to be used, and the training data of the second object comprises a second type of characteristics to be used and a client type label;
the first object module and the second object module are also respectively used for training the local model based on training data thereof, obtaining parameter information of the local model, encrypting the parameter information of the local model to obtain a parameter information ciphertext, and sending the parameter information ciphertext to the other party;
The first object module and the second object module are also respectively used for updating the parameters of the local model based on the parameter information ciphertext from the other party;
and under the condition that the local models of the first object module and the second object module do not reach the training ending condition, respectively returning and executing the steps of training the local models based on the training data of the first object module and the second object module.
7. The system of claim 6, wherein the first object module performs privacy processing on the first client sample set by using a target privacy processing manner to obtain a first privacy sample set, and the second object module performs privacy processing on a second client sample set owned by the first object module by using the target privacy processing manner to obtain a second privacy sample set, and the process of determining the same sample portion between the first privacy sample set and the second privacy sample set specifically includes:
the second object module obtains an encryption key generated by the first object, compresses the second client sample set based on a hash function, adds noise data, and encrypts the noise data by using the encryption key to obtain a second compressed sample set;
The second object module sends the second compressed sample set to the first object module, so that the first object module decrypts the second compressed sample set based on a decryption key corresponding to the encryption key to obtain a second decrypted sample set;
the first object module compresses a first customer sample set through a hash function, decrypts the first customer sample set, performs secondary hash to obtain a first privacy sample set, and sends the second decrypted sample set and the first privacy sample set to the second object module;
the second object module deletes the noise data in the second decryption sample set, and performs secondary hash calculation on the second decryption sample set after deleting the noise data to obtain a second privacy sample set;
the second object module determines the same portion of the first and second privacy sample sets.
8. The system of claim 6, wherein the local model of the first object module and the local model of the second object module are federal XGBoost models with the same structure, the first object module and the second object module train their local models based on their training data, respectively, and obtain parameter information of the local model, encrypt the parameter information of the local model to obtain parameter information ciphertext, and send the parameter information ciphertext to the other party, including:
The second object module determines a root node based on training data of the second object module, splits the root node based on the training data and initially determines a local model of the second object;
the second object module calculates a first derivative and a second derivative of training data in the local model, performs additive homomorphic encryption on the first derivative and the second derivative of the training data in the local model to obtain a first parameter information ciphertext, and sends the first parameter information ciphertext to the first object;
the first object module determines all possible values of all features of training data of the first object as first-order derivatives and second-order derivatives of split points based on the first parameter information ciphertext, performs homomorphic encryption on all possible values of all features of the training data of the first object as first-order derivatives and second-order derivatives of the split points to obtain second parameter information ciphertext, and sends the second parameter information ciphertext to the second object module;
the second object module determines a global optimal partitioning point based on the second parameter information ciphertext, if the global optimal partitioning point is in the first object module, the information of the global optimal partitioning point is sent to the first object module, so that the first object module partitions training data of the first object module based on the information of the global optimal partitioning point, and sends a partitioning result to the second object module;
The second object module updates the local model based on the division result;
and in the case that the local models of the first object module and the second object module do not reach the training ending condition, returning to execute the second object module to calculate the first derivative and the second derivative of the data in the local model.
9. The system of claim 8, wherein the training end condition comprises:
the loss function values of the local models of the first object module and the second object module converge;
or, the maximum depth of the tree of the local model of the first object module and the second object module reaches the set depth.
10. The system of claim 6, wherein the first object module performs privacy processing on the first set of customer samples by using a target privacy processing method, specifically comprising:
randomly selecting M groups of customer samples from a first customer sample set, wherein the total amount of the M groups of customer samples is greater than N;
transmitting the codes of the M groups of client samples to the second object, so that the second object randomly extracts M client samples from each group of client samples in the M groups of client samples, determines client class labels corresponding to each client sample in the M client samples based on the risk preference information, and determines distribution rationality of the M client samples based on the client class labels, wherein M is not less than N;
If the distribution rationality of the M client samples meets the rationality requirement, determining the M groups of client samples as a first client sample set to be used;
and carrying out privacy processing on the first customer sample set to be used by adopting a target privacy processing mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310018596.7A CN115936112B (en) | 2023-01-06 | 2023-01-06 | Client portrait model training method and system based on federal learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310018596.7A CN115936112B (en) | 2023-01-06 | 2023-01-06 | Client portrait model training method and system based on federal learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115936112A CN115936112A (en) | 2023-04-07 |
| CN115936112B true CN115936112B (en) | 2023-07-18 |
Family
ID=86699208
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310018596.7A Active CN115936112B (en) | 2023-01-06 | 2023-01-06 | Client portrait model training method and system based on federal learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115936112B (en) |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109299728B (en) * | 2018-08-10 | 2023-06-27 | 深圳前海微众银行股份有限公司 | Sample joint prediction method, system and medium based on construction of gradient tree model |
| US11588621B2 (en) * | 2019-12-06 | 2023-02-21 | International Business Machines Corporation | Efficient private vertical federated learning |
| CN112085159B (en) * | 2020-07-24 | 2023-08-15 | 西安电子科技大学 | User tag data prediction system, method and device and electronic equipment |
| CN112750043B (en) * | 2021-01-14 | 2024-02-02 | 中国工商银行股份有限公司 | Service data pushing method, device and server |
| CN113283185B (en) * | 2021-07-23 | 2021-11-12 | 平安科技（深圳）有限公司 | Federal model training and client imaging method, device, equipment and medium |
| CN113722987B (en) * | 2021-08-16 | 2023-11-03 | 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 |
| CN114329527A (en) * | 2021-12-17 | 2022-04-12 | 阿里巴巴（中国）有限公司 | Intersection data acquisition method, equipment and system |
| CN114969503B (en) * | 2022-03-30 | 2024-04-02 | 贵州大学 | Multi-data user portrait realizing method based on federal learning |
| CN115269983A (en) * | 2022-07-28 | 2022-11-01 | 上海光之树科技有限公司 | Target sample recommendation method based on two-party data privacy protection |
-
2023
- 2023-01-06 CN CN202310018596.7A patent/CN115936112B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115936112A (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20200394518A1 (en) | Method for collaborative learning of an artificial neural network without disclosing training data | |
| US20230078061A1 (en) | Model training method and apparatus for federated learning, device, and storage medium | |
| CN110572253B (en) | Method and system for enhancing privacy of federated learning training data | |
| CN110189192B (en) | Information recommendation model generation method and device | |
| CN114401079B (en) | Multi-party united information value calculation method, related equipment and storage medium | |
| WO2020015478A1 (en) | Model-based prediction method and device | |
| EP3075098B1 (en) | Server-aided private set intersection (psi) with data transfer | |
| CN110633806A (en) | Longitudinal federated learning system optimization method, device, equipment and readable storage medium | |
| CN111210003B (en) | Longitudinal federated learning system optimization method, device, equipment and readable storage medium | |
| CN110084377A (en) | Method and apparatus for constructing decision tree | |
| CN112989399B (en) | Data processing system and method | |
| CN114696990B (en) | Multi-party computing method, system and related equipment based on fully homomorphic encryption | |
| CN113947211A (en) | Federal learning model training method and device, electronic equipment and storage medium | |
| CN113542228A (en) | Data transmission method and device based on federal learning and readable storage medium | |
| CN113221153B (en) | Graph neural network training method and device, computing equipment and storage medium | |
| CN113609781A (en) | Automobile production mold optimization method, system, equipment and medium based on federal learning | |
| CN112818369A (en) | Combined modeling method and device | |
| CN113409134A (en) | Enterprise financing trust method and device based on federal learning | |
| CN112000988A (en) | Factorization machine regression model construction method and device and readable storage medium | |
| CN115965093A (en) | Model training method and device, storage medium and electronic equipment | |
| CN115936112B (en) | Client portrait model training method and system based on federal learning | |
| CN117094773A (en) | Online migration learning method and system based on blockchain privacy calculation | |
| CN111914281A (en) | Bayes model training method and device based on block chain and homomorphic encryption | |
| Elmisery | Private personalized social recommendations in an IPTV system | |
| CN114723068A (en) | Federal model training method and device |
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 |