Model training method, medium, electronic equipment and program product

Abstract

The invention provides a model training method, a medium, electronic equipment and a program product, and relates to the technical field of machine learning, and the method comprises the steps: sending a business model to a client, enabling the client to train the business model according to the business data of the client, and obtaining a trained first business model, according to the method and the device, the first business model is acquired, the vector representation corresponding to the business data is acquired from the first business model, the vector representation is returned to the server side, the server side trains the business model based on the vector representation, and the trained second business model is acquired, so that the business data can be more effectively acquired on the premise of ensuring the privacy of the business data of the client side. The training of the service model of the server is completed, so that the service model of the server can better provide services to the outside, and the model structures of the service models, used for federated learning, of the server and the client can be kept consistent, so that the cognition of the server and the client on the service models is kept consistent, and the service experience of the client is improved. Therefore, the online service model of the server can have a better effect.

Description

Technical Field

[0001] The present disclosure relates to the field of machine learning technology, and more particularly, to a model training method, a medium, an electronic device, and a program product. Background Art

[0002] As a distributed machine learning method, federated learning has significant advantages in privacy protection and data cooperation. The core idea of federated learning is that when multiple data sources jointly participate in model training, without the need for raw data transfer, the model is jointly trained only by interacting with the intermediate parameters of the model, so that the raw data can stay local.

[0003] In the related art of federated learning, it is necessary for each participating party to have relatively high technical capabilities and be able to establish an effective communication mechanism in order to achieve successful federated learning. This means that all participating parties need to have certain technical capabilities to understand and implement the necessary interface specifications and be able to overcome the problems that occur in federated learning. Summary of the Invention

[0004] This Summary of the Invention section is provided to introduce concepts in a brief form that will be described in detail in the subsequent Detailed Description section. This Summary of the Invention section is not intended to identify key features or essential features of the claimed technical solution, nor is it intended to be used to limit the scope of the claimed technical solution.

[0005] In a first aspect, the present disclosure provides a model training method, which is executed by a server, and the method includes: Sending a business model deployed on the server to a client; the client is used to train the business model according to the business data of the client to obtain a first trained business model, and obtain a vector representation corresponding to the business data from the first trained business model and return the vector representation to the server; the business data is data related to the business model; Based on the vector representation, training the business model deployed on the server to obtain a second trained business model, so as to provide services through the second trained business model.

[0006] In a second aspect, the present disclosure provides a model training method, which is executed by a client, and the method includes: Receiving a business model deployed on the server sent by the server; Training the business model according to the business data of the client to obtain a first trained business model; the business data is data related to the business model; Obtaining a vector representation corresponding to the business data from the first trained business model and sending the vector representation to the server; Among them, the server is used to train the business model deployed on the server based on the vector representation to obtain a second trained business model, so as to provide services externally through the second business model.

[0007] In a third aspect, the present disclosure provides a computer-readable medium having a computer program stored thereon, and when the computer program is executed by a processing device, the steps of the method described in the first aspect are implemented, or the steps of the method described in the second aspect are implemented.

[0008] In a fourth aspect, the present disclosure provides an electronic device, including: a storage device having a computer program stored thereon; a processing device configured to execute the computer program in the storage device to implement the steps of the method described in the first aspect, or the steps of the method described in the second aspect.

[0009] In a fifth aspect, the present disclosure provides a computer program product including a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented, or the steps of the method described in the second aspect are implemented.

[0010] Based on the above technical solutions, by sending the business model deployed on the server to the client, the client can train the business model according to the business data of the client to obtain a first trained business model, and obtain the vector representation corresponding to the business data from the first business model and return the vector representation to the server. Then, the server trains the business model based on the vector representation to obtain a second trained business model. This can not only complete the training of the business model on the server while ensuring the privacy of the business data of the client, enabling the business model deployed on the server to better provide services externally, but also make the model structures of the business models used for federated learning by the server and the client consistent, so that the server and the client have the same understanding of the business model, thereby enabling the business model on the server line to have better effects.

[0011] Other features and advantages of the present disclosure will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In combination with the drawings and with reference to the following specific implementation, the above and other features, advantages, and aspects of the embodiments of the present disclosure will become more obvious. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic, and the original elements and elements are not necessarily drawn to scale. In the drawings: Figure 1It is a flowchart of a model training method shown according to an exemplary embodiment.

[0013] Figure 2 It is a flowchart of a model training method shown according to another exemplary embodiment.

[0014] Figure 3 It is a schematic diagram of a model training method shown according to an exemplary embodiment.

[0015] Figure 4 It is a schematic structural diagram of a model training device shown according to an exemplary embodiment.

[0016] Figure 5 It is a schematic structural diagram of a model training device shown according to another exemplary embodiment.

[0017] Figure 6 It is a schematic structural diagram of an electronic device shown according to an exemplary embodiment. Detailed implementation manners

[0018] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Instead, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0019] It should be understood that the various steps recited in the method embodiments of the present disclosure can be executed in different orders and / or executed in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this regard.

[0020] As used herein, the term "including" and its variations are open-ended, that is, "including but not limited to". The term "based on" is "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0021] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or the interdependent relationship.

[0022] It should be noted that the modifiers "a" and "multiple" mentioned in this disclosure are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly specified in the context, it should be understood as "one or more".

[0023] The names of the messages or information exchanged between multiple devices in the embodiments of this disclosure are only for illustrative purposes and are not used to limit the scope of these messages or information.

[0024] It can be understood that before using the technical solutions disclosed in the embodiments of this disclosure, the types, usage scopes, usage scenarios, etc. of the personal information involved in this disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0025] For example, when responding to receiving an active request from a user, a prompt message is sent to the user to clearly prompt the user that the operation requested by the user will require obtaining and using the user's personal information (such as business data in the client). Thus, the user can autonomously choose whether to provide personal information to software or hardware such as an electronic device, an application program, a server, or a storage medium that performs the operations of the technical solutions of this disclosure according to the prompt message.

[0026] As an optional but non-limiting implementation manner, the manner of sending a prompt message to the user in response to receiving an active request from the user can be, for example, in the form of a pop-up window, and the prompt message can be presented in text in the pop-up window. In addition, the pop-up window can also carry selection controls for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0027] It can be understood that the above process of notifying and obtaining the user's authorization is only illustrative and does not constitute a limitation on the implementation manner of this disclosure. Other manners that meet relevant laws and regulations can also be applied to the implementation manner of this disclosure.

[0028] At the same time, it can be understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) should comply with the requirements of corresponding laws, regulations and related provisions.

[0029] Figure 1 is a flowchart of a model training method shown according to an exemplary embodiment. As Figure 1 shown, the embodiments of this disclosure provide a model training method, and this method can specifically be executed by a model training device, and this device can be implemented in a software and / or hardware manner and is configured in the server. As Figure 1 shown, this method can include the following steps.

[0030] In step 110, a business model deployed on the server is sent to the client; the client is used to train the business model according to the business data of the client to obtain a first trained business model, and obtain the vector representation corresponding to the business data from the first trained business model and return the vector representation to the server.

[0031] Here, the business model deployed on the server refers to a business model deployed on the server for providing services externally. The server may refer to an application server. For example, the application server may be a server of an application program providing multimedia content services.

[0032] It should be noted that the business model may be a business model that has been trained to a certain extent or a business model that has not been trained. By sending the business model deployed on the server to the client, the business data of the client can be used to optimize the business model deployed on the server.

[0033] In the embodiments of the present disclosure, the business model may be a machine learning model for pushing multimedia content to users, and the multimedia content may be content such as videos, pictures and texts, audio, etc. Of course, in other application scenarios, the business model may also be a machine learning model for providing other services. For example, the business model may be a machine learning model for providing risk prediction services.

[0034] The business data of the client is data related to the business model. Taking the business model as a machine learning model for pushing multimedia content to users as an example, the business data is the conversion data of the user for the multimedia content. For example, the client may be the end of the advertiser, and the business model is used to push the promotion content published by the advertiser to the user. Correspondingly, the client can obtain the corresponding business data through the detection link corresponding to the promotion content (such as ClickID, which is used to uniquely identify the user's click action in the promotion content).

[0035] The business model sent to the client enables the client to train the business model according to the business data of the client to obtain a first trained business model, extract the vector representation (embedding) corresponding to the business data from the first trained business model, and return the vector representation corresponding to the business data obtained from the first trained business model to the server.

[0036] It should be noted that the client uses the business data held by the client to train the received business model. After obtaining the first trained business model, the client extracts the vector representation corresponding to the business data from the first trained business model and sends the vector representation corresponding to the business data to the server.

[0037] It should be understood that the client can be regarded as an electronic device in the customer domain, and the customer domain holds the business data of user privacy. To protect the business data from being transmitted outside the customer domain, the client trains the business model by using the business data, converts the business data into a vector representation, and provides the vector representation to the server. Since the vector representation cannot be converted back into the original business data, but the gradient information of the business model is obtained by using the vector representation, the business data in the client will not be transmitted outside the customer domain. Thus, while realizing the privacy protection of the business data, the training of the business model in the server is achieved, so that the trained second business model can provide services more accurately to the outside world.

[0038] In step 120, based on the vector representation, the business model deployed on the server is trained to obtain the trained second business model, so as to provide services to the outside world through the second business model.

[0039] Here, the server receives the vector representation sent by the client and trains the business model deployed on the server through the vector representation to obtain the trained second business model, so as to provide services to the outside world through the trained second business model.

[0040] It is worth noting that since the vector representation will not be converted into the original business data but is used to train the business model, the business data in the client will not be spread outside the client, greatly protecting the privacy of the business data. Moreover, since the server provides the business model to the client, for the client, the client only needs to provide the business data and computing resources to obtain the vector representation for training the business model. Therefore, the training of the business model in the server can be completed on the premise of ensuring the privacy of the business data, enabling the business model deployed on the server to provide better services to the outside world.

[0041] Taking the business model as a machine learning model for pushing multimedia content to users and the client as the terminal of the advertiser as an example, after training the business model in the server with the vector representation corresponding to the business data, the business model can significantly improve the promotion effect of the promotion content published by the client. Through experiments, it is found that after training the business model in the server with the vector representation corresponding to the business data, the AUC (the area under the ROC curve (Receiver Operating Characteristic Curve), and the ROC curve is a performance measurement tool used to show the performance of the model at different thresholds) of the trained second business model can be greatly improved, thereby improving the promotion effect of the promotion content.

[0042] It should be noted that for the server side, the developers of the server side generally have relatively high technical capabilities. For the client side, the business personnel of the client side generally do not have relatively high technical capabilities. Through the model training method provided by the embodiments of the present disclosure, the server side sends a business model to the client side, which can make the model structures of the business models for federated learning on the server side and the client side consistent, so that the understandings of the business models on the server side and the client side are consistent, thereby enabling the business model on the server side to have better effects.

[0043] Therefore, by sending the business model deployed on the server side to the client side, enabling the client side to train the business model according to the business data of the client side to obtain the first trained business model, and obtaining the vector representation corresponding to the business data from the first business model and returning the vector representation to the server side, and then the server side training the business model based on the vector representation to obtain the second trained business model, not only can the training of the business model on the server side be completed on the premise of ensuring the privacy of the business data of the client side, enabling the business model deployed on the server side to better provide services externally, but also by the server side sending the business model to the client side, the model structures of the business models for federated learning on the server side and the client side can be made consistent, so that the understandings of the business models on the server side and the client side are consistent, thereby enabling the business model on the server side to have better effects.

[0044] In some implementable embodiments, in step 110, a business model and a training method for training the business model are sent to the client side; the client side is used to train the business model according to the business data of the client side in combination with the training method to obtain the first trained business model.

[0045] Here, following the above embodiments, the server side can not only send a business model to the client side, but also send a training method for training the business model. Among them, the training method is used to instruct the client side to train the sent business model based on the sent training method.

[0046] The client side receives the business model and the training method sent by the server side, uses the business data in the client side, and trains the received business model through the sent training method to obtain the first trained business model, and returns the vector representation corresponding to the business data obtained from the first trained business model to the server side.

[0047] It should be noted that by the server sending the service model and the training method to the client, the service model and the training method between the server and the client can be kept consistent, so that the server and the client have the same understanding of the service model and the training method, thus enabling the service model on the server side to have better effects. Moreover, by the server sending the service model and the training method to the client, it is not necessary for the business personnel on the client side to have high technical capabilities. They only need to train the service model using the business data in the client according to the sent training method.

[0048] In some embodiments, the training method includes data description information for screening business data for model training and / or a loss function for training the service model.

[0049] Among them, the data description information is used to instruct the client to select target business data matching the data description information from the business data of the client according to the data description information, and train the service model according to the target business data; the loss function is used to instruct the client to train the service model according to the loss function.

[0050] The data description information is information used to instruct the client on what kind of business data to use for training the service model. For example, the data description information can be used to instruct to use specific types of business data to train the service model. The client screens the business data matching the data description information from the business data held by the client through the data description information to train the service model.

[0051] The loss function is a function used to evaluate the difference between the predicted value and the true value of the service model. The smaller the value of the loss function, the closer the predicted value of the service model is to the true value, that is, the better the performance of the service model. Exemplarily, the loss function can be a cross-entropy loss function, etc.

[0052] It is worth noting that in the embodiments of the present disclosure, the training method may include data description information and / or a loss function. Of course, in other embodiments, the training method may also include parameters other than data description information and / or a loss function. For example, the training method may also include batch information for indicating the training batches of the service model.

[0053] Correspondingly, when the training method includes data description information and a loss function, the client is used to select target business data matching the data description information from the business data of the client according to the data description information, and train the service model according to the target business data and the loss function to obtain the first trained service model.

[0054] It should be noted that in the server, the same training method is also used to train the business model deployed on the server.

[0055] It should be understood that for the client side, the client does not need to have technical capabilities. It only needs to screen the target business data from the business data held in the client according to the issued data description information, input the target business data into the business model, and adjust the parameters of the business model based on the issued loss function.

[0056] That is to say, for the client, the characteristics of the business data for training the business model, the loss function, and the structure of the business model are all consistent with those of the server. The client only needs to provide business data and computing resources, which can make the server and the client consistent in the model structure dimension, data feature dimension, and spatial representation dimension, greatly ensuring the consistency between the client and the server.

[0057] It should be noted that for each client participating in federated learning, since the training method and model structure are sent from the server to the client, the quality of the business data provided by each client can be kept consistent. Therefore, the business model finally trained by the server can have good generalization ability and can have good performance in terms of the quality of business data and the consistency of samples.

[0058] Thus, through the above implementation methods, the training of the business model on the server can be completed on the premise of ensuring the privacy of the business data of the client. Moreover, by sending the business model and training method from the server to the client, the business model can have high generalization ability. In addition, since both the business model and the training method are provided by the server, for the client, only business data and computing power need to be provided, greatly reducing the technical ability requirements of federated learning for users.

[0059] In some implementable implementation methods, the vector representation corresponding to the business data is the vector representation of the business data in the middle layer of the first trained business model.

[0060] Here, the first layer of the business model is the input layer of the business model, the last layer of the business model is the output layer of the business model, and the middle layer of the business model can be the model structure layer between the input layer and the output layer of the business model. For example, the middle layer of the business model can be the hidden layer of the business model.

[0061] The client can extract the vector representation corresponding to the business data from the middle layer of the first trained business model in the client and send the vector representation to the server. Then, the server uses the vector representation of the middle layer sent by the client to train the business model deployed on the server.

[0062] It should be noted that for the vector representation at the bottom layer of the business model, the changes in the vector representation are relatively large, while the spatial dimension of the vector representation at the top layer of the business model is not rich enough. The vector representation of the business data obtained in the middle layer of the business model is both stable and rich in spatial dimension. Therefore, the vector representation of the middle layer can be used to train the business model on the server side.

[0063] In some implementable embodiments, in step 110, the server can send a preset application program to the client.

[0064] Here, the client is used to run the preset application program, obtain the business model from the server through the preset application program, obtain business data from the client, and train the business model according to the business data to obtain the first trained business model.

[0065] Among them, when the business model deployed in the server needs to be updated, the server can send a preset application program to the client. It should be noted that the preset application program can be understood as an update package.

[0066] The client receives the preset application program sent by the server and runs the preset application program to obtain the business model from the server through the preset application program, obtain business data from the client, and train the business model according to the business data to obtain the first trained business model, and return the vector representation corresponding to the business data obtained by the first trained business model to the server.

[0067] That is to say, for the client, the technical personnel on the client side do not need to perform any operations, but only need to provide computing resources to run the preset application program and provide the business data in the client.

[0068] It should be noted that the client can use a container (docker) in the cloud service to run the preset application program.

[0069] Thus, through the preset application program, the client can automatically complete the training of the business model by running the preset application program and obtain the vector representation of the business data, thereby greatly reducing the requirements for the technical capabilities of users in federated learning.

[0070] Figure 2 It is a flowchart of a model training method shown according to another exemplary embodiment. As Figure 2 shown, the embodiments of the present disclosure provide a model training method. Specifically, this method can be executed by a model training device, which can be implemented in a software and / or hardware manner and configured in the client. As Figure 2 shown, this method can include the following steps.

[0071] S210. Receive the business model deployed on the server sent by the server. S220. Train the business model according to the business data of the client to obtain the first trained business model. S230. Obtain the vector representation corresponding to the business data from the first business model and send the vector representation to the server.

[0072] Here, the server is used to train the business model deployed on the server based on the vector representation to obtain the second trained business model, so as to provide services externally through the second business model.

[0073] For the detailed description of the above steps S210 to S230, reference can be made to the relevant description on the server side, which will not be elaborated here.

[0074] Thus, through the above steps S210 to S230, not only can the training of the business model on the server be completed on the premise of ensuring the privacy of the business data of the client, so that the business model deployed on the server can better provide services externally, but also by the server sending the business model to the client, the model structures of the business models used for federated learning on the server and the client can be kept consistent, so that the server and the client have the same understanding of the business model, thereby enabling the business model on the server side to have better effects.

[0075] In some implementable embodiments, in step S210, the client can receive the business model sent by the server and the training method for training the business model. Correspondingly, in step S220, the client can train the business model according to the business data of the client in combination with the training method to obtain the first trained business model.

[0076] In some implementable embodiments, the training method includes data description information for screening the business data used for model training and / or a loss function for training the business model.

[0077] Among them, the data description information is used to instruct the client to select target business data matching the data description information from the business data of the client according to the data description information, and train the business model according to the target business data; the loss function is used to instruct the client to train the business model according to the loss function.

[0078] In some implementable embodiments, in step S230, the client can extract the vector representation corresponding to the business data from the intermediate layer of the first business model and send the vector representation to the server.

[0079] In some implementable embodiments, in step S210, the client can receive a preset application program issued by the server. Correspondingly, in step S220, the client can run the preset application program, obtain a business model from the server through the preset application program, obtain business data from the client, and train the business model according to the business data to obtain a first trained business model.

[0080] Of course, the client can also obtain the vector representation corresponding to the business data from the first business model through the preset application program and send the vector representation to the server.

[0081] It should be noted that for the detailed description on the client side, reference can be made to the relevant description on the server side, which will not be elaborated here.

[0082] Figure 3 It is a schematic diagram of a model training method shown according to an exemplary embodiment. As Figure 3 shown, a business model can be deployed on the business platform of the server. The server provides the business model and the training method to the client. The client uses the business data provided by the client business and the training method provided by the server through the federated learning module to perform model training on the business model provided by the server, obtain the vector representation of the business data, and return the vector representation of the business data to the server. The server uses the vector representation to perform model training on the business model on the server.

[0083] Figure 4 It is a schematic structural diagram of a model training device shown according to an exemplary embodiment. As Figure 4 shown, the present disclosure embodiment provides a model training device 400 configured on the server. The model training device 400 includes: A sending module 401, configured to send the business model deployed on the server to the client; the client is used to train the business model according to the business data of the client to obtain a first trained business model, obtain the vector representation corresponding to the business data from the first business model, and return the vector representation to the server; the business data is data related to the business model; A first training module 402, configured to train the business model deployed on the server based on the vector representation to obtain a second trained business model for providing services externally through the second business model.

[0084] Optionally, the sending module 401 is specifically configured to: Send the service model and the training method for training the service model to the client; the client is used to train the service model according to the service data of the client and in combination with the training method to obtain a first service model that has been trained.

[0085] Optionally, the training method includes data description information for screening service data for model training and / or a loss function for training the service model; The data description information is used to instruct the client to select target service data that matches the data description information from the service data of the client and to train the service model according to the target service data; the loss function is used to instruct the client to train the service model according to the loss function.

[0086] Optionally, the vector representation corresponding to the service data is the vector representation of the service data in the intermediate layer of the first service model that has been trained.

[0087] Optionally, the sending module 401 is specifically configured to: Send a preset application program to the client; the client is used to run the preset application program, obtain the service model from the server through the preset application program, obtain the service data from the client, and train the service model according to the service data to obtain a first service model that has been trained.

[0088] Regarding the model training device 400 in the above embodiments, the method logics executed by each functional module have been described in detail in the part regarding the method, and will not be elaborated here.

[0089] Figure 5 It is a schematic structural diagram of a model training device shown according to another exemplary embodiment. As Figure 5 shown, the present disclosure embodiment provides a model training device 500 configured in a client. The model training device 500 includes: A receiving module 501, configured to receive a service model deployed on the server sent by the server; A second training module 502, configured to train the service model according to the service data of the client to obtain a first service model that has been trained; the service data is data related to the service model; An obtaining module 503, configured to obtain the vector representation corresponding to the service data from the first service model and send the vector representation to the server; Among them, the server is used to train the business model deployed on the server based on the vector representation to obtain a second trained business model, so as to provide services externally through the second business model.

[0090] Optionally, the receiving module 501 is specifically configured to: Receive the business model sent by the server and the training method for training the business model. The second training module 502 is specifically configured to: Train the business model according to the business data of the client in combination with the training method to obtain a first trained business model.

[0091] Optionally, the training method includes data description information for screening business data for model training and / or a loss function for training the business model; the data description information is used to instruct the client to select target business data matching the data description information from the business data of the client and train the business model according to the target business data; the loss function is used to instruct the client to train the business model according to the loss function.

[0092] Optionally, the obtaining module 503 is specifically configured to: Extract the vector representation corresponding to the business data from the intermediate layer of the first business model and send the vector representation to the server.

[0093] Optionally, the receiving module 501 is specifically configured to: Receive the preset application program issued by the server. The second training module 502 is specifically configured to: Run the preset application program, obtain the business model from the server through the preset application program, obtain the business data from the client, and train the business model according to the business data to obtain a first trained business model.

[0094] Regarding the model training device 500 in the above embodiments, the method logics executed by each functional module have been described in detail in the part about the method, and will not be repeated here.

[0095] Next, refer to Figure 6, which shows a schematic structural diagram of an electronic device (such as a server or a client) 600 suitable for implementing the embodiments of the present disclosure. The client in the embodiments of the present disclosure may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), PMPs (Portable Multimedia Players), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The illustrated electronic device is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present disclosure.

[0096] As Figure 6 shown, the electronic device 600 may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 601, which may perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 602 or the program loaded from the storage device 608 into the random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604. The input / output (I / O) interface 605 is also connected to the bus 604.

[0097] Generally, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or wiredly to exchange data. Although Figure 6 the electronic device 600 with various devices is shown, it should be understood that it is not required to implement or have all the shown devices. More or fewer devices may be implemented or had alternatively.

[0098] Particularly, according to the embodiments of the present disclosure, the process described above with reference to the flowchart may be implemented as a computer software program. For example, the embodiments of the present disclosure include a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes program codes for performing the methods shown in the flowchart. In such an embodiment, the computer program may be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above functions defined in the methods of the embodiments of the present disclosure are executed.

[0099] It should be noted that the computer-readable medium described above in this disclosure can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of a computer-readable storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this disclosure, a computer-readable storage medium can be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium can be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0100] In some embodiments, the client and the server can communicate using any currently known or future-developed network protocol such as HTTP (HyperText Transfer Protocol), and can be interconnected with digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include local area networks ("LAN"), wide area networks ("WAN"), the Internet (e.g., the Internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed network.

[0101] The above computer-readable medium can be included in the above electronic device; or it can exist separately without being assembled into the electronic device.

[0102] The above computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: send a service model deployed on the server to the client; the client is used to train the service model according to the service data of the client to obtain a first trained service model, and obtain a vector representation corresponding to the service data from the first service model and return the vector representation to the server; the service data is data related to the service model; based on the vector representation, train the service model deployed on the server to obtain a second trained service model for providing services externally through the second service model.

[0103] Alternatively, the above computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: receive a service model deployed on the server sent by the server; train the service model according to the service data of the client to obtain a first trained service model; the service data is data related to the service model; obtain a vector representation corresponding to the service data from the first service model and send the vector representation to the server; wherein, the server is used to train the service model deployed on the server based on the vector representation to obtain a second trained service model for providing services externally through the second service model.

[0104] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages or combinations thereof. The above programming languages include, but are not limited to, object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., by using an Internet service provider to connect through the Internet).

[0105] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a portion of code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks may occur in a different order than noted in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or by a combination of dedicated hardware and computer instructions.

[0106] The modules described in the embodiments of the present disclosure can be implemented in software or in hardware. In some cases, the name of a module does not constitute a limitation on the module itself.

[0107] The functions described above herein can be performed, at least in part, by one or more hardware logic components. By way of example, and without limitation, exemplary types of hardware logic components that may be used include: Field Programmable Gate Arrays (FPGA), Application Specific Integrated Circuits (ASIC), Application Specific Standard Products (ASSP), System on a Chip (SOC), Complex Programmable Logic Devices (CPLD), and the like.

[0108] In the context of the present disclosure, a machine-readable medium may be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a Read-Only Memory (ROM), an Erasable Programmable Read-Only Memory (EPROM or Flash memory), an optical fiber, a portable Compact Disc Read-Only Memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0109] The above description is only a preferred embodiment of the present disclosure and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosure concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the present disclosure.

[0110] In addition, although the operations are depicted in a particular order, this should not be understood as requiring that the operations be performed in the particular order shown or in sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the present disclosure. Certain features described in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, the various features described in the context of a single embodiment may also be implemented separately or in any suitable sub-combination in multiple embodiments.

[0111] Although the subject matter has been described in language specific to structural features and / or methodological logical acts, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. On the contrary, the specific features and acts described above are merely example forms for implementing the claims. Regarding the apparatus in the above embodiments, the specific manner in which each module performs operations has been described in detail in the embodiments related to the method, and will not be elaborated here.

Claims

1. A model training method, characterized in that: Executed by the server, the method includes: Sending a business model deployed on the server to the client; the client is used to train the business model according to the business data of the client to obtain a trained first business model, and obtain a vector representation corresponding to the business data from the first business model and return the vector representation to the server; the business data is data related to the business model; Based on the vector representation, the business model deployed on the server is trained to obtain a trained second business model, so as to provide services externally through the second business model.

2. The method according to claim 1, characterized in that The sending the business model deployed on the server to the client includes: The business model and a training method for training the business model are sent to the client; the client is used to train the business model according to the business data of the client in combination with the training method to obtain a first business model that has been trained.

3. The method according to claim 2, characterized in that The training method includes data description information for screening business data for model training and / or a loss function for training the business model; The data description information is used to instruct the client to select target business data matching the data description information from the business data of the client according to the data description information, and to train the business model according to the target business data; the loss function is used to instruct the client to train the business model according to the loss function.

4. The method according to any one of claims 1 to 3, characterized in that The vector representation corresponding to the business data is the vector representation of the business data in the middle layer of the first business model after the training.

5. The method according to any one of claims 1 to 3, characterized in that The sending the business model deployed on the server to the client includes: A preset application is sent to the client; the client is used to run the preset application, obtain the business model from the server through the preset application, obtain the business data from the client, and train the business model according to the business data to obtain a first business model that has been trained.

6. A model training method, characterized in that: Executed by a client, the method includes: Receiving a service model sent by a server and deployed on the server; The business model is trained according to the business data of the client to obtain a trained first business model; the business data is data related to the business model; Acquire a vector representation corresponding to the business data from the first business model, and send the vector representation to the server; The server is used to train the business model deployed on the server based on the vector representation to obtain a trained second business model, so as to provide services externally through the second business model.

7. The method according to claim 6, characterized in that The receiving service model deployed on the server and sent by the service end includes: Receiving the business model and a training method for training the business model sent by the server; The training of the business model according to the business data of the client to obtain a trained first business model includes: The business model is trained according to the business data of the client in combination with the training method to obtain a trained first business model.

8. The method according to claim 7, characterized in that The training method includes data description information for screening business data for model training and / or a loss function for training the business model; the data description information is used to instruct the client to select target business data matching the data description information from the business data of the client according to the data description information, and train the business model according to the target business data; the loss function is used to instruct the client to train the business model according to the loss function.

9. The method according to any one of claims 6 to 8, characterized in that: The acquiring the vector representation corresponding to the business data from the first business model, and sending the vector representation to the server, includes: Extracting a vector representation corresponding to the business data from the middle layer of the first business model, and sending the vector representation to the server.

10. The method according to any one of claims 6 to 8, characterized in that: The receiving service model deployed on the server and sent by the service end includes: Receiving a preset application program sent by the server; The training of the business model according to the business data of the client to obtain a trained first business model includes: Run the preset application, obtain the business model from the server through the preset application, obtain the business data from the client, and train the business model according to the business data to obtain a trained first business model.

11. A computer readable medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processing device, the computer program implements the steps of the method according to any one of claims 1 to 5, or implements the steps of the method according to any one of claims 6 to 10.

12. An electronic device, characterized in that: include: a storage device having a computer program stored thereon; A processing device, used to execute the computer program in the storage device to implement the steps of the method according to any one of claims 1 to 5, or to implement the steps of the method according to any one of claims 6 to 10.

13. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 5, or implements the steps of the method according to any one of claims 6 to 10.

Images