CN111797302A - Model processing method, device, storage medium and electronic device - Google Patents

Abstract

The embodiment of the present application provides a model processing method, device, storage medium and electronic device, the model processing method includes: obtaining first information, and inputting the first information as a training sample into a prediction model for training, obtaining model parameters of the trained prediction model; sending the model parameters to a server; receiving a common model parameter returned by the server, the common model parameter being calculated from the model parameter and the model parameters corresponding to other users; and obtaining a second prediction model according to the common model parameter. The prediction accuracy and generalization ability of the prediction model are significantly improved, and the user's data privacy can be well protected.

Description

Model processing method, device, storage medium and electronic device

technical field

The present application relates to the field of electronic technology, and in particular, to a model processing method, device, storage medium and electronic device.

Background technique

With the development of artificial intelligence, electronic devices such as smartphones are becoming more and more intelligent. Electronic devices can provide users with various intelligent functions according to the collected data.

In the related art, the prediction model in the electronic device will make predictions according to the collected information, and then provide corresponding services to users. Such as recommending corresponding applications, etc. However, due to the limited information collected, the prediction model in the related art results in insufficient accuracy of the learned algorithm.

SUMMARY OF THE INVENTION

Embodiments of the present application provide a model processing method, apparatus, storage medium, and electronic device, which can improve the prediction accuracy of the prediction model.

The embodiment of the present application provides a model processing method, which includes:

obtaining first information, and inputting the first information as a training sample into a prediction model for training, to obtain target model parameters of the trained first prediction model;

sending the target model parameters to the server;

receiving the shared model parameters returned by the server, where the shared model parameters are calculated from the target model parameters and model parameters corresponding to other users;

A second prediction model is obtained according to the shared model parameters.

The embodiment of the present application also provides a model processing method, which includes:

According to a set of model parameters corresponding to the prediction model of each user, multiple sets of model parameters corresponding to multiple users are obtained;

Adjusting the multiple sets of model parameters to multiple sets of standard model parameters of the same standard;

The multiple groups of standard model parameters are calculated to obtain a group of shared model parameters;

The shared model parameters are sent to the prediction model corresponding to each user, so as to use the shared model parameters as second model parameters of the prediction model corresponding to each user.

The embodiment of the present application also provides a model processing device, which includes:

a first acquisition module for model parameters, configured to acquire first information, and input the first information as a training sample into a prediction model for training, to obtain target model parameters of the trained first prediction model;

a first sending module, configured to send the target model parameters to a server;

a receiving module, configured to receive the shared model parameters returned by the server, where the shared model parameters are calculated from the target model parameters and model parameters corresponding to other users;

A processing module, configured to obtain a second prediction model according to the shared model parameters.

The embodiment of the present application also provides a model processing device, which includes:

a second acquisition module for model parameters, configured to obtain multiple sets of model parameters corresponding to multiple users according to a set of model parameters corresponding to the prediction model of each user;

an adjustment module for adjusting the multiple sets of model parameters to multiple sets of standard model parameters of the same standard;

a shared model parameter acquisition module for calculating the multiple groups of standard model parameters to obtain a group of shared model parameters;

The second sending module is configured to send the shared model parameter to the prediction model corresponding to each user, so as to use the shared model parameter as the second model parameter of the prediction model corresponding to each user.

An embodiment of the present application further provides a storage medium, on which a computer program is stored, and when the computer program runs on a computer, the computer is made to execute the steps of the above model processing method.

An embodiment of the present application further provides an electronic device, the electronic device includes a processor and a memory, the memory stores a computer program, and the processor invokes the computer program stored in the memory to execute The steps of the above model processing method.

In the model processing method, device, storage medium, and electronic device provided by the embodiments of the present application, first information is obtained, and the first information is used as a training sample to input into a prediction model for training, and the first prediction model after training is obtained. target model parameters; then send the model parameters to the server; then receive the shared model parameters returned by the server, where the shared model parameters are calculated from the target model parameters and the model parameters corresponding to other users; The shared model parameters are used to obtain a second prediction model. By adopting the idea of federated learning, the model parameters of other users can be collaboratively calculated without uploading user data, which helps the local terminal to make better predictions, significantly improves the prediction accuracy and generalization ability of the model, and at the same time, it can be well protect user data privacy.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present application more clearly, the following briefly introduces the accompanying drawings that are used in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative effort.

FIG. 1 is a schematic diagram of an application scenario of a model processing method provided by an embodiment of the present application.

FIG. 2 is a first schematic flowchart of a model processing method provided by an embodiment of the present application.

FIG. 3 is a schematic flowchart of a second model processing method provided by an embodiment of the present application.

FIG. 4 is a third schematic flowchart of a model processing method provided by an embodiment of the present application.

FIG. 5 is a schematic diagram of another application scenario of the model processing method provided by the embodiment of the present application.

FIG. 6 is a schematic structural diagram of an information processing apparatus provided by an embodiment of the present application.

FIG. 7 is another schematic structural diagram of an information processing apparatus provided by an embodiment of the present application.

FIG. 8 is a schematic diagram of a first structure of an electronic device provided by an embodiment of the present application.

FIG. 9 is a schematic diagram of a second structure of an electronic device provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of this application.

Please refer to FIG. 1. FIG. 1 is a schematic diagram of an application scenario of the model processing method provided by the embodiment of the present application. The model processing method is applied to electronic devices. A panoramic perception architecture is provided in the electronic device. Panoramic perception architecture is the integration of hardware and software in electronic devices for implementing model processing methods.

Among them, the panoramic perception architecture includes an information perception layer, a data processing layer, a feature extraction layer, a scenario modeling layer, and an intelligent service layer.

The information perception layer is used to acquire the information of the electronic device itself and/or the information in the external environment. The information perception layer may include multiple sensors. For example, the information perception layer includes a distance sensor, a magnetic field sensor, a light sensor, an acceleration sensor, a fingerprint sensor, a Hall sensor, a position sensor, a gyroscope, an inertial sensor, an attitude sensor, a barometer, a heart rate sensor, and other sensors.

Among them, the distance sensor can be used to detect the distance between the electronic device and the external object. The magnetic field sensor can be used to detect the magnetic field information of the environment in which the electronic device is located. The light sensor can be used to detect the light information of the environment where the electronic device is located. Acceleration sensors can be used to detect acceleration data of electronic devices. The fingerprint sensor can be used to collect the user's fingerprint information. Hall sensor is a magnetic field sensor made according to the Hall effect, which can be used to realize automatic control of electronic equipment. The location sensor can be used to detect the current geographic location of the electronic device. Gyroscopes can be used to detect the angular velocity of electronic devices in various directions. Inertial sensors can be used to detect motion data of electronic devices. The attitude sensor can be used to sense the attitude information of the electronic device. A barometer can be used to detect the air pressure in the environment in which the electronic device is located. The heart rate sensor may be used to detect the user's heart rate information.

The data processing layer is used to process the data obtained by the information perception layer. For example, the data processing layer can perform data cleaning, data integration, data transformation, data reduction and other processing on the data obtained by the information perception layer.

Among them, data cleaning refers to cleaning a large amount of data obtained by the information perception layer to eliminate invalid data and duplicate data. Data integration refers to integrating multiple single-dimensional data obtained by the information perception layer into a higher or more abstract dimension to comprehensively process multiple single-dimensional data. Data transformation refers to converting the data type or format of the data obtained by the information perception layer, so that the transformed data can meet the processing requirements. Data reduction refers to reducing the amount of data to the greatest extent possible on the premise of keeping the original data as much as possible.

The feature extraction layer is used to perform feature extraction on the data processed by the data processing layer to extract features included in the data. The extracted features may reflect the state of the electronic device itself, the state of the user, or the environmental state of the environment in which the electronic device is located.

Among them, the feature extraction layer can extract features or process the extracted features by filtering method, packaging method, integration method and other methods.

The filtering method refers to filtering the extracted features to remove redundant feature data. The packing method is used to filter the extracted features. The integration method refers to the integration of multiple feature extraction methods to construct a more efficient and accurate feature extraction method for feature extraction.

The scenario modeling layer is used to construct a model according to the features extracted by the feature extraction layer, and the obtained model can be used to represent the state of the electronic device, the state of the user, or the environment state, etc. For example, the scenario modeling layer can construct a key value model, a pattern identification model, a graph model, an entity relationship model, an object-oriented model, etc. according to the features extracted by the feature extraction layer.

The intelligent service layer is used to provide users with intelligent services according to the model constructed by the scenario modeling layer. For example, the intelligent service layer can provide users with basic application services, can perform system intelligent optimization for electronic devices, and can also provide users with personalized intelligent services.

In addition, the panoramic perception architecture can also include multiple algorithms, each of which can be used to analyze and process data, and multiple algorithms can form an algorithm library. For example, the algorithm library may include Markov algorithm, latent Dirichlet distribution algorithm, Bayesian classification algorithm, support vector machine, K-means clustering algorithm, K-nearest neighbor algorithm, conditional random field, residual network, long Algorithms such as short-term memory networks, convolutional neural networks, and recurrent neural networks.

The embodiment of the present application provides a model processing method, and the model processing method can be applied to an electronic device. Electronic devices may be smartphones, tablets, gaming devices, Augmented Reality (AR) devices, automobiles, data storage devices, audio playback devices, video playback devices, notebooks, desktop computing devices, wearable devices such as watches, glasses , helmets, electronic bracelets, electronic necklaces, electronic clothing and other equipment.

Please refer to FIG. 2 , FIG. 2 is a schematic flowchart of a first type of model processing method provided by an embodiment of the present application. The model processing method includes the following steps:

101. Acquire first information, and input the first information as a training sample into a prediction model for training, to obtain target model parameters of the trained first prediction model.

The first information may be all information about the user. For example, it may include three categories of information about the environment where the user is located, the operation information of the electronic device used by the user, and the user behavior information. The environmental information may include the temperature, humidity, location, brightness, etc. of the environment, and the environmental information may also include the user's physical information, such as blood pressure, pulse, heart rate, and the like. Specifically, the environmental information may be obtained by the sensor. For example, through at least one of a distance sensor, a magnetic field sensor, a light sensor, an acceleration sensor, a fingerprint sensor, a Hall sensor, a position sensor, a gyroscope, an inertial sensor, an attitude sensor, a barometer, a blood pressure sensor, a pulse sensor, a heart rate sensor, etc. A fetched environment information. The environmental information can also be the current audio information obtained by the microphone, and the current image information obtained by the camera module.

The operation information of the electronic device may include the startup time, shutdown time, standby time, memory usage rate at each time point, main chip usage rate at each time point, current running program information, background running program information, running time of each program, each Program downloads, etc.

The user behavior information may include the user's action track information, browsing information, payment information, travel information, and the like.

The first information may also include configuration information of the electronic device, user information stored in the electronic device, and the like. The user information includes the user's identity information, personal hobbies, browsing records, personal collections and other human-computer interaction information.

It should be noted that some of the first information may be in two or three categories of environmental information, electronic device operation information, and user behavior information at the same time.

After the first information is obtained, the first information may be input into the prediction model as a training sample for training, and target model parameters of the trained first prediction model are obtained.

For example, the first information includes travel information of the user, which may specifically include living location, home time period, going out time, travel vehicle, vehicle use time period, work location, stay location, stay time period, and the like. In some embodiments, user travel related modes can also be defined, and seven travel modes such as bus, subway, driving, cycling, walking, high-speed rail, and airplane can be defined. The prediction model can predict the travel mode that the user is most likely to use next, and set the corresponding function according to the travel mode. If the next most likely travel mode is the subway, the application content used on the subway can be preloaded, which may specifically include the content of the news application, video content, application content, and subway payment application.

Among them, each user's electronic device can locally use a deep learning framework to establish a prediction model, such as a travel pattern recognition model.

102. Send the target model parameters to the server.

After the target model parameters of the trained first prediction model are obtained, the target model parameters are sent to the server. The server can be a pre-built remote server or a cloud server. Among them, uploading the target model parameters to the server instead of directly uploading the first information avoids the leakage of the first information, so as to better protect user privacy and comply with relevant laws and regulations, such as EU data privacy protection regulations.

103. Receive the shared model parameters returned by the server, where the shared model parameters are obtained by calculating the target model parameters and model parameters corresponding to other users.

The shared model parameters returned by the server are received, and the shared model parameters are calculated from the target model parameters and the model parameters corresponding to other users. The server receives the model parameters corresponding to multiple users, and then calculates the shared model parameters according to the multiple model parameters of the multiple users.

Wherein, the server may use multiple algorithms to calculate multiple sets of standard model parameters. For example, the average method, the maximum value method or the median method is used. The averaging method assumes that the aligned parameter vector of each user is w_u=(w_u1,w_u2,...,w_un), where u represents the uth user and n represents the number of parameters, then the calculation method based on the averaging method takes all users Average over each parameter dimension. The maximum method is similar to the average method, except that the maximum value in each parameter dimension is selected. The median method is also similar to the mean method, except that the median in each parameter dimension is chosen.

104. Obtain a second prediction model according to the shared model parameters.

The prediction model obtains the second prediction model according to the shared model parameters returned by the server. The training process of the prediction model of a single user can well calculate the model parameters of other users, and the model parameters of other users are obtained according to the first information of other users' electronic devices, that is, the data knowledge and behavior habits of other users are calculated. Significantly improve the recognition accuracy and generalization ability of the prediction model.

In related technologies, it is difficult not only to protect the privacy of user data, but also to perform computational learning and training of data when there are differences in data between different electronic devices, resulting in greater limitations in the accuracy and adaptability of prediction algorithms. Based on the idea of federated transfer learning, this embodiment can achieve stronger generalization ability and higher precision, and can protect user data privacy. Specifically, by adopting the federated learning idea, the first information of other users can be collaboratively calculated without uploading the first information of the user, which helps the local terminal to make better predictions, and significantly improves the accuracy and generalization of the prediction model. It can protect the data privacy of users well; through transfer learning, the model parameters of the prediction model are aligned under the premise of the difference between the first information, which ensures that the prediction model has stronger robustness Not only can the user first information be the same, but also the scenarios where the user first information is different can be handled, which greatly improves the application scope of the constructed prediction mode.

Please refer to FIG. 3 , FIG. 3 is a schematic flowchart of a first type of model processing method provided by an embodiment of the present application. The model processing method includes the following steps:

201. Acquire first information, and input the first information as a training sample into a prediction model for training, to obtain target model parameters of the trained first prediction model.

The first information may be all information about the user. For example, it may include three categories of information about the environment where the user is located, the operation information of the electronic device used by the user, and the user behavior information. The environmental information may include the temperature, humidity, location, brightness, etc. of the environment, and the environmental information may also include the user's physical information, such as blood pressure, pulse, heart rate, and the like. Specifically, the environmental information may be obtained by the sensor. For example, through at least one of a distance sensor, a magnetic field sensor, a light sensor, an acceleration sensor, a fingerprint sensor, a Hall sensor, a position sensor, a gyroscope, an inertial sensor, an attitude sensor, a barometer, a blood pressure sensor, a pulse sensor, a heart rate sensor, etc. A fetched environment information. The environmental information can also be the current audio information obtained by the microphone, and the current image information obtained by the camera module.

The operation information of the electronic device may include the startup time, shutdown time, standby time, memory usage rate at each time point, main chip usage rate at each time point, current running program information, background running program information, running time of each program, each Program downloads, etc.

The user behavior information may include the user's action track information, browsing information, payment information, travel information, and the like.

The first information may also include configuration information of the electronic device, user information stored in the electronic device, and the like. The user information includes the user's identity information, personal hobbies, browsing records, personal collections and other human-computer interaction information.

It should be noted that some of the first information may be in two or three categories of environmental information, electronic device operation information, and user behavior information at the same time.

After the first information is obtained, the first information may be input into the prediction model as a training sample for training, and target model parameters of the trained first prediction model are obtained.

For example, the first information includes travel information of the user, which may specifically include living location, home time period, going out time, travel vehicle, vehicle use time period, work location, stay location, stay time period, and the like. In some embodiments, user travel related modes can also be defined, and seven travel modes such as bus, subway, driving, cycling, walking, high-speed rail, and airplane can be defined. The prediction model can predict the travel mode that the user is most likely to use next, and set the corresponding function according to the travel mode. If the next most likely travel mode is the subway, the application content used on the subway can be preloaded, which may specifically include the content of the news application, video content, application content, and subway payment application.

Among them, each user's electronic device can locally use a deep learning framework to establish a prediction model, such as a travel pattern recognition model.

202. Send the target model parameters to the server.

After the target model parameters of the trained first prediction model are obtained, the target model parameters are sent to the server. The server can be a pre-built remote server or a cloud server. Among them, uploading the target model parameters to the server instead of directly uploading the first information avoids the leakage of the first information, so as to better protect user privacy and comply with relevant laws and regulations, such as EU data privacy protection regulations.

203. Receive the shared model parameters returned by the server, where the shared model parameters are calculated from the target model parameters and the model parameters corresponding to other users.

The shared model parameters returned by the server are received, and the shared model parameters are calculated from the target model parameters and the model parameters corresponding to other users. The server receives the model parameters corresponding to multiple users, and then calculates the shared model parameters according to the multiple model parameters of the multiple users.

Wherein, the server may use multiple algorithms to calculate multiple sets of standard model parameters. For example, the average method, the maximum value method or the median method is used. The averaging method assumes that the aligned parameter vector of each user is w_u=(w_u1,w_u2,...,w_un), where u represents the uth user and n represents the number of parameters, then the calculation method based on the averaging method takes all users Average over each parameter dimension. The maximum method is similar to the average method, except that the maximum value in each parameter dimension is selected. The median method is also similar to the mean method, except that the median in each parameter dimension is chosen.

204. Obtain the matching degree between the target model parameter and the shared model parameter.

The matching degree of the target model parameters and the shared model parameters is calculated, and the matching degree of the two is calculated.

205. When the matching degree is less than the preset matching degree, retrain the prediction model by using the shared model parameters and the first information to obtain the second prediction model.

When the matching degree is less than the preset matching degree, it means that the target model parameters obtained by the training of the prediction model are far from the shared model parameters returned by the server, and directly using the shared model parameters as the second model parameters of the prediction model may have a large error , in order to predict the accuracy of the model, the prediction model can be retrained by using the shared model parameters and the first information. For example, retrain the prediction model using the shared model parameters, or calculate the shared model parameters and the target model parameters, and obtain the intermediate value of the two as the model parameters of the prediction mode for retraining.

206. When the matching degree is greater than the preset matching degree, obtain a second prediction model according to the shared model parameters.

According to the prediction model, the second prediction model is obtained according to the shared model parameters returned by the server. The training process of the prediction model of a single user can well calculate the model parameters of other users, and the model parameters of other users are obtained according to the first information of other users' electronic devices, that is, the data knowledge and behavior habits of other users are calculated. Significantly improve the recognition accuracy and generalization ability of the prediction model.

In some embodiments, obtaining the second prediction model according to the shared model parameters includes:

The prediction model using the shared model parameters will be used as the second prediction model.

Directly using the shared model parameters as the second model parameters of the prediction model can also be understood as the prediction model using the shared model parameters as the second prediction model.

In some embodiments, obtaining the second prediction model according to the shared model parameters includes:

Adjust the shared model parameters according to the target model parameters to obtain the second model parameters;

The prediction model using the second model parameters will be used as the second prediction model.

Directly using the shared model parameters as the model parameters of the prediction model may have deviations. In this case, the second model parameters can be obtained by adjusting the target model parameters obtained by the previous training of the prediction model.

In some embodiments, obtaining the second prediction model according to the shared model parameters includes:

The prediction model using the shared model parameters is trained according to the first information to adjust the shared model parameters to obtain the second model parameters;

The prediction model using the second model parameters will be used as the second prediction model.

Directly using the shared model parameters as the model parameters of the prediction model may have deviations. In this case, the prediction model with the shared model parameters can be used to retrain according to the first information to adjust the shared model parameters to obtain the second model parameters. The retraining can A smaller sample is used for training, or all samples can be used for training; finally, the prediction model using the parameters of the second model is used as the second prediction model.

In related technologies, it is difficult not only to protect the privacy of user data, but also to perform computational learning and training of data when there are differences in data between different electronic devices, resulting in greater limitations in the accuracy and adaptability of prediction algorithms. Based on the idea of federated transfer learning, this embodiment can achieve stronger generalization ability and higher precision, and can protect user data privacy. Specifically, by adopting the federated learning idea, the first information of other users can be collaboratively calculated without uploading the first information of the user, which helps the local terminal to make better predictions, and significantly improves the accuracy and generalization of the prediction model. It can protect the data privacy of users well; through transfer learning, the model parameters of the prediction model are aligned under the premise of the difference between the first information, which ensures that the prediction model has stronger robustness Not only can the user first information be the same, but also the scenarios where the user first information is different can be handled, which greatly improves the application scope of the constructed prediction mode.

Please refer to FIG. 4 , which is a schematic flowchart of a third type of model processing method provided by an embodiment of the present application. The model processing method includes the following steps:

301. Obtain multiple sets of model parameters corresponding to multiple users according to a set of model parameters corresponding to the prediction model of each user.

The prediction model corresponding to each user is first trained according to the first information of the corresponding user to obtain a trained model and a set of model parameters, and corresponding to multiple users, there are multiple sets of model parameters.

302. Adjust the multiple sets of model parameters to multiple sets of standard model parameters of the same standard.

After acquiring multiple sets of model parameters corresponding to multiple users, adjust the multiple sets of model parameters to multiple sets of standard model parameters of the same standard. Specifically, the transfer learning method is used to align the model parameters uploaded by each user, so that the parameters of all users are in the same space. For example, the prediction model of each user is a travel prediction model, and each travel prediction model performs travel prediction based on the first information of the corresponding electronic device.

For example, the first information includes travel information of the user, which may specifically include living location, home time period, going out time, travel vehicle, vehicle use time period, work location, stay location, stay time period, and the like. In some embodiments, user travel related modes can also be defined, and seven travel modes such as bus, subway, driving, cycling, walking, high-speed rail, and airplane can be defined. The prediction model can predict the travel mode that the user is most likely to use next, and set the corresponding function according to the travel mode. If the next most likely travel mode is the subway, the application content used on the subway can be preloaded, which may specifically include the content of the news application, video content, application content, and subway payment application.

In this way, the model parameters corresponding to each user are in the same space, and are all identified for the same standard (if present).

303. Calculate multiple sets of standard model parameters to obtain a set of common model parameters.

Multiple sets of standard model parameters can be calculated using a variety of algorithms. For example, the average method, the maximum value method or the median method is used. The averaging method assumes that the aligned parameter vector of each user is w_u=(w_u1,w_u2,...,w_un), where u represents the uth user and n represents the number of parameters, then the calculation method based on the averaging method takes all users Average over each parameter dimension. The maximum method is similar to the average method, except that the maximum value in each parameter dimension is selected. The median method is also similar to the mean method, except that the median in each parameter dimension is chosen.

304. Send the shared model parameters to the prediction model corresponding to each user, so as to use the shared model parameters as the second model parameters of the prediction model corresponding to each user.

After the shared model parameters are obtained, they are sent to the prediction model corresponding to each user. After the prediction model obtains the shared model parameters, the second model parameters are obtained according to the shared model parameters.

In some embodiments, before using the common model parameter as the second model parameter of the prediction model corresponding to each user, the matching degree between the model parameter and the common model parameter may be obtained first.

When the matching degree is less than the preset matching degree, the prediction model is retrained by using the shared model parameters and the first information.

When the matching degree is greater than the preset matching degree, a second prediction model is obtained according to the shared model parameters.

When the matching degree is less than the preset matching degree, it means that the model parameters obtained by the training of the prediction model are far from the shared model parameters returned by the server. There may be a large error in directly using the shared model parameters as the second model parameters of the prediction model. In order to predict the accuracy of the model, the prediction model may be retrained using the shared model parameters and the first information. For example, retrain the prediction model using the shared model parameters, or calculate the shared model parameters and the model parameters, and obtain the intermediate value of the two as the model parameters of the prediction mode for retraining.

When the matching degree is greater than the preset matching degree, the second prediction model is obtained according to the prediction model according to the shared model parameters returned by the server. The training process of the prediction model of a single user can well calculate the model parameters of other users, and the model parameters of other users are obtained according to the first information of other users' electronic devices, that is, the data knowledge and behavior habits of other users are calculated. Significantly improve the recognition accuracy and generalization ability of the prediction model.

In some embodiments, obtaining the second prediction model according to the shared model parameters may include:

The prediction model using the shared model parameters will be used as the second prediction model.

Directly using the shared model parameters as the second model parameters of the prediction model can also be understood as the prediction model using the shared model parameters as the second prediction model.

In some embodiments, obtaining the second prediction model according to the shared model parameters may include:

Adjust the shared model parameters according to the model parameters to obtain the second model parameters;

The prediction model using the second model parameters will be used as the second prediction model.

Directly using the shared model parameters as the final parameters of the prediction model may have deviations. In this case, you can adjust the model parameters obtained by the previous training of the prediction model to obtain the final model parameters.

In some embodiments, obtaining the second prediction model according to the shared model parameters includes:

The prediction model using the shared model parameters is trained according to the first information to adjust the shared model parameters to obtain the second model parameters;

The prediction model using the second model parameters will be used as the second prediction model.

Directly using the shared model parameters as the final parameters of the prediction model may be biased. In this case, the prediction model using the shared model parameters can be retrained according to the first information to adjust the shared model parameters to obtain the second model parameters. The retraining can A smaller sample is used for training, or all samples can be used for training; finally, the prediction model using the parameters of the second model is used as the second prediction model.

In related technologies, it is difficult not only to protect the privacy of user data, but also to perform computational learning and training of data when there are differences in data between different electronic devices, resulting in greater limitations in the accuracy and adaptability of prediction algorithms. Based on the idea of federated transfer learning, this embodiment can achieve stronger generalization ability and higher precision, and can protect user data privacy. Specifically, by adopting the federated learning idea, the first information of other users can be collaboratively calculated without uploading the first information of the user, which helps the local terminal to make better predictions, and significantly improves the accuracy and generalization of the prediction model. It can protect the data privacy of users well; through transfer learning, the model parameters of the prediction model are aligned under the premise of the difference between the first information, which ensures that the prediction model has stronger robustness Not only can the user first information be the same, but also the scenarios where the user first information is different can be handled, which greatly improves the application scope of the constructed prediction mode.

Please refer to FIG. 5 , which is a schematic diagram of a scenario of a model processing method provided by an embodiment of the present application. First obtain the first information of the user's electronic device, and then input the obtained first information into the prediction model for training to obtain the trained first prediction model and target model parameters, and then upload the target model parameters to the server, and the server is based on migration learning Align the uploaded target model parameters with the model parameters uploaded by other users to avoid the problem of difficult direct integration of model parameters caused by data inconsistency. Next, the aligned model parameters are calculated to obtain the shared model parameters, and then the shared model parameters are sent to the prediction model to learn again to obtain the second prediction model. The electronic device can perform prediction using the second prediction model, and perform function control according to the prediction result. For example, if the prediction is made based on the user's travel information, and the prediction result is that there will be a high-speed rail trip, ticket purchasing software, map software, car rental software, taxi-hailing software, train number inquiry software, etc. can be recommended.

In some embodiments, the model processing method may specifically include: firstly acquiring information of the user's electronic device through the information perception layer (specifically including electronic device operation information, user behavior information, information acquired by various sensors, electronic device status information, electronic device Display content information, download information on electronic devices, etc.), and then process the information of electronic devices (such as invalid data deletion, etc.) through the data processing layer, and then extract the required information from the information processed by the data processing layer through the feature extraction layer. the first information (for details of the first information, please refer to the description of the above-mentioned embodiment), and then input the first information into the scenario modeling layer. The scenario modeling layer includes a pre-stored prediction model, and the prediction model of the scenario modeling layer is based on The first information is trained to obtain the trained first prediction model and target model parameters, and then the target model parameters are uploaded to the server through a transmission module (such as a radio frequency module), and the server uploads the uploaded target model parameters and other users based on transfer learning. The model parameters are aligned to avoid the problem that the model parameters are difficult to directly integrate due to data inconsistency. Next, the aligned model parameters are calculated to obtain the shared model parameters, and then the shared model parameters are sent back to the electronic device. After the electronic device receives the shared model parameters, the shared model parameters are input into the scenario modeling layer. The parameters of the prediction model in are replaced with the shared model parameters to obtain the second prediction model. The prediction module can directly replace the original model parameters with the shared model parameters, or can also use the shared model parameters to replace the original model parameters and then perform training and learning to obtain the second prediction model. Finally, the intelligent service layer can use the second prediction model to make predictions, and perform function control according to the prediction results. For example, if the prediction is made based on the user's travel information, and the prediction result is that there will be a high-speed rail trip, ticket purchasing software, map software, car rental software, taxi-hailing software, train number inquiry software, etc. can be recommended.

Please refer to FIG. 6 , which is a schematic structural diagram of a model processing apparatus provided by an embodiment of the present application. The model processing apparatus 400 includes a model parameter first obtaining module 401 , a first sending module 402 , a receiving module 403 and a processing module 404 .

The model parameter first obtaining module 401 is configured to obtain first information, and input the first information as a training sample into a prediction model for training, and obtain target model parameters of the trained first prediction model.

The first sending module 402 is configured to send the target model parameters to the server.

The receiving module 403 is configured to receive the shared model parameters returned by the server, where the shared model parameters are calculated from the target model parameters and the model parameters corresponding to other users.

The processing module 404 is configured to obtain the second prediction model according to the shared model parameters.

The model processing apparatus of this embodiment may be provided in an electronic device used by a user.

In some embodiments, the processing module 404 is further configured to obtain the matching degree between the target model parameter and the shared model parameter; when the matching degree is less than the preset matching degree, use the shared model parameter and the first information to retrain the prediction model; When the matching degree is greater than the preset, a second prediction model is obtained according to the shared model parameters.

In some embodiments, the processing module 404 is further configured to use the prediction model using the shared model parameters as the second prediction model.

In some embodiments, the processing module 404 is further configured to adjust the shared model parameters according to the target model parameters to obtain the second model parameters; the prediction model using the second model parameters is used as the second prediction model.

In some embodiments, the processing module 404 is further configured to perform training using the prediction model using the shared model parameters according to the first information, so as to adjust the shared model parameters to obtain the second model parameters; and use the prediction model using the second model parameters as the second prediction Model.

Please refer to FIG. 7 , which is another schematic structural diagram of a model processing apparatus provided by an embodiment of the present application. The model processing apparatus 500 includes a second model parameter obtaining module 501 , an adjustment module 502 , a shared model parameter obtaining module 503 and a second sending module 504 .

A second model parameter acquisition module 501, configured to obtain multiple sets of model parameters corresponding to multiple users according to a set of model parameters corresponding to the prediction model of each user;

an adjustment module 502, configured to adjust multiple groups of model parameters to multiple groups of standard model parameters of the same standard;

a shared model parameter acquisition module 503, configured to calculate multiple sets of standard model parameters to obtain a set of shared model parameters;

The second sending module 504 is configured to send the shared model parameters to the prediction model corresponding to each user, so as to use the shared model parameters as the second model parameters of the prediction model corresponding to each user.

The modules in this embodiment may be used in cooperation with the modules in the foregoing embodiments. For example, the second model parameter acquisition module 501 receives the model parameters sent by the first sending module 402 . The receiving module 403 receives the shared model parameters and the like sent by the second sending module 504 .

The model processing apparatus of this embodiment may be provided in the server.

Please refer to FIG. 8 , which is a schematic diagram of a first structure of an electronic device 600 according to an embodiment of the present application. The electronic device 600 includes a processor 601 and a memory 602 . The processor 601 is electrically connected to the memory 602 .

The processor 601 is the control center of the electronic device 600, uses various interfaces and lines to connect various parts of the entire electronic device, executes the electronic device by running or calling the computer program stored in the memory 602, and calling the data stored in the memory 602. Various functions of the device and processing data, so as to carry out the overall monitoring of the electronic device.

In this embodiment, the processor 601 in the electronic device 600 loads the instructions corresponding to the processes of one or more computer programs into the memory 602 according to the following steps, and is executed by the processor 601 and stored in the memory 602 A computer program in , which implements various functions:

acquiring first information, and inputting the first information as a training sample into a prediction model for training, to obtain target model parameters of the trained first prediction model;

Send the target model parameters to the server;

Receive the shared model parameters returned by the server, and the shared model parameters are calculated from the target model parameters and the model parameters corresponding to other users;

A second prediction model is obtained according to the shared model parameters.

In some embodiments, when obtaining the second prediction model according to the shared model parameters, the processor 601 performs the following steps:

Obtain the matching degree between the target model parameters and the shared model parameters;

When the matching degree is less than the preset matching degree, the prediction model is retrained by using the shared model parameters and the first information;

When the matching degree is greater than the preset matching degree, a second prediction model is obtained according to the shared model parameters.

In some embodiments, when obtaining the second prediction model according to the shared model parameters, the processor 601 performs the following steps:

The prediction model using the shared model parameters will be used as the second prediction model.

In some embodiments, when obtaining the second prediction model according to the shared model parameters, the processor 601 performs the following steps:

Adjust the shared model parameters according to the target model parameters to obtain the second model parameters;

The prediction model using the second model parameters will be used as the second prediction model.

In some embodiments, when obtaining the second prediction model according to the shared model parameters, the processor 601 performs the following steps:

The prediction model using the shared model parameters is trained according to the first information to adjust the shared model parameters to obtain the second model parameters;

The prediction model using the second model parameters will be used as the second prediction model.

In some embodiments, processor 601 performs the following steps:

According to a set of model parameters corresponding to the prediction model of each user, multiple sets of model parameters corresponding to multiple users are obtained;

Adjust multiple sets of model parameters to multiple sets of standard model parameters of the same standard;

Calculate multiple sets of standard model parameters to obtain a set of common model parameters;

The shared model parameters are sent to the prediction model corresponding to each user, so as to use the shared model parameters as the second model parameters of the prediction model corresponding to each user.

In some embodiments, please refer to FIG. 9 , which is a schematic diagram of a second structure of an electronic device 600 according to an embodiment of the present application.

The electronic device 600 further includes: a display screen 603 , a control circuit 604 , an input unit 605 , a sensor 606 and a power supply 607 . The processor 601 is electrically connected to the display screen 603 , the control circuit 604 , the input unit 605 , the sensor 606 and the power supply 607 , respectively.

The display screen 603 may be used to display information input by or provided to the user and various graphical user interfaces of the electronic device, which may be composed of images, text, icons, videos, and any combination thereof.

The control circuit 604 is electrically connected to the display screen 603 for controlling the display screen 603 to display information.

The input unit 605 can be used to receive input numbers, character information or user characteristic information (eg fingerprint), and generate keyboard, mouse, joystick, optical or trackball signal input related to user settings and function control. Wherein, the input unit 605 may include a fingerprint identification module.

The sensor 606 is used to collect the information of the electronic device itself or the user's information or the external environment information. For example, sensors 606 may include distance sensors, magnetic field sensors, light sensors, acceleration sensors, fingerprint sensors, Hall sensors, position sensors, gyroscopes, inertial sensors, attitude sensors, barometers, heart rate sensors, and the like.

Power supply 607 is used to power various components of electronic device 600 . In some embodiments, the power supply 607 may be logically connected to the processor 601 through a power management system, so as to implement functions such as managing charging, discharging, and power consumption through the power management system.

Although not shown in FIG. 9 , the electronic device 600 may further include a camera, a Bluetooth module, and the like, which will not be repeated here.

It can be seen from the above that the embodiment of the present application provides an electronic device, and the processor in the electronic device performs the following steps: acquiring first information, and inputting the first information as a training sample into a prediction model for training, and obtaining the prediction model of the trained prediction model. model parameters; sending the model parameters to the server; receiving the shared model parameters returned by the server, where the shared model parameters are calculated from model parameters corresponding to other users; and obtaining a second prediction model according to the shared model parameters.

An embodiment of the present application further provides a storage medium, where a computer program is stored in the storage medium, and when the computer program runs on the computer, the computer executes the model processing method described in any of the foregoing embodiments.

For example, in some embodiments, when the computer program is run on a computer, the computer performs the following steps:

acquiring first information, and inputting the first information as a training sample into a prediction model for training, to obtain target model parameters of the trained first prediction model;

Send the target model parameters to the server;

Receive the shared model parameters returned by the server, and the shared model parameters are calculated from the target model parameters and the model parameters corresponding to other users;

A second prediction model is obtained according to the shared model parameters.

It should be noted that those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium , the storage medium may include, but is not limited to, a read only memory (ROM, Read Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and the like.

The model processing method, apparatus, storage medium, and electronic device provided by the embodiments of the present application have been described in detail above. The principles and implementations of the present application are described herein using specific examples, and the descriptions of the above embodiments are only used to help understand the methods and core ideas of the present application; meanwhile, for those skilled in the art, according to the Thoughts, there will be changes in specific embodiments and application scopes. To sum up, the contents of this specification should not be construed as limitations on the present application.

Claims (10)

1. A method of model processing, comprising:

acquiring first information, inputting the first information into a prediction model as a training sample for training, and obtaining a target model parameter of the trained first prediction model;

sending the target model parameters to a server;

receiving a common model parameter returned by the server, wherein the common model parameter is obtained by calculating the target model parameter and model parameters corresponding to other users;

and obtaining a second prediction model according to the common model parameters.

2. The model processing method of claim 1, wherein deriving a second predictive model from the common model parameters comprises:

obtaining the matching degree of the target model parameters and the common model parameters;

when the matching degree is smaller than a preset matching degree, retraining the prediction model by using the common model parameter and the first information;

and when the matching degree is greater than the preset matching degree, obtaining a second prediction model according to the common model parameters.

3. The model processing method according to claim 1 or 2, wherein deriving a second predictive model from the common model parameters comprises:

using the prediction model using the common model parameters as a second prediction model.

4. The model processing method of claim 1, wherein deriving a second predictive model from the common model parameters comprises:

adjusting the common model parameter according to the target model parameter to obtain a second model parameter;

using the prediction model using the second model parameters as a second prediction model.

5. The model processing method of claim 1, wherein deriving a second predictive model from the common model parameters comprises:

training the prediction model of the common model parameter according to the first information to adjust the common model parameter to obtain a second model parameter;

using the prediction model using the second model parameters as a second prediction model.

6. A method of model processing, comprising:

obtaining a plurality of groups of model parameters corresponding to a plurality of users according to a group of model parameters corresponding to the prediction model of each user;

adjusting the multiple groups of model parameters to multiple groups of standard model parameters of the same standard;

calculating the multiple groups of standard model parameters to obtain a group of common model parameters;

and sending the common model parameters to the prediction model corresponding to each user so as to take the common model parameters as second model parameters of the prediction model corresponding to each user.

7. A model processing apparatus, comprising:

the model parameter first acquisition module is used for acquiring first information, inputting the first information into a prediction model as a training sample and training to obtain a target model parameter of the trained first prediction model;

the first sending module is used for sending the target model parameters to a server;

the receiving module is used for receiving the common model parameters returned by the server, wherein the common model parameters are obtained by calculating the target model parameters and the model parameters corresponding to other users;

and the processing module is used for obtaining a second prediction model according to the common model parameters.

8. A model processing apparatus, comprising:

the second model parameter acquisition module is used for acquiring a plurality of groups of model parameters corresponding to a plurality of users according to a group of model parameters of the prediction model corresponding to each user;

the adjusting module is used for adjusting the multiple groups of model parameters into multiple groups of standard model parameters of the same standard;

the shared model parameter acquisition module is used for calculating the multiple groups of standard model parameters to obtain a group of shared model parameters;

and the second sending module is used for sending the common model parameters to the prediction model corresponding to each user, and taking the common model parameters as second model parameters of the prediction model corresponding to each user.

9. A storage medium having stored thereon a computer program, characterized in that, when the computer program is run on a computer, it causes the computer to execute the model processing method according to any one of claims 1 to 6.

10. An electronic device, characterized in that the electronic device comprises a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to execute the model processing method according to any one of claims 1 to 6 by calling the computer program stored in the memory.

Images