Data processing method, device, computer system and readable storage medium

Abstract

The present disclosure provides a data processing method, comprising: iteratively performing at least one model training until a preset iteration stopping condition is met to obtain a target model; wherein the model training comprises: training an initial model by using at least one training sample set to obtain at least one training model corresponding to the at least one training sample set; determining, based on a test sample set, an accuracy rate corresponding to the initial model and accuracy rates corresponding to the at least one training model respectively; and determining a model with the highest accuracy rate from the accuracy rate corresponding to the initial model and the accuracy rates corresponding to the at least one training model, and taking the model with the highest accuracy rate as the initial model. The present disclosure also provides a data prediction method, device, computer system, readable storage medium and computer program product.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and more specifically, to a data processing method, a data prediction method, an apparatus, a computer system, a readable storage medium, and a computer program product. Background Technology

[0002] In recent years, in sales, finance, and many other industries, data isolation and the emphasis on data privacy are becoming the next challenge for artificial intelligence, given factors such as intellectual property, privacy protection, and data security. Federated learning offers new hope for data isolation and privacy protection. It can build a unified model for multiple collaborators (such as multiple different enterprises) while protecting local data, allowing them to jointly utilize the jointly built model while ensuring data security.

[0003] In realizing the concept disclosed herein, the inventors discovered at least the following problems in the related technology: once one or more collaborators withdraw, the data provided by the withdrawing collaborators will be completely blocked, and the recognition accuracy of the model built through federated learning will drop significantly. Summary of the Invention

[0004] In view of the above, this disclosure provides a data processing method, a data prediction method, an apparatus, a computer system, a readable storage medium, and a computer program product.

[0005] One aspect of this disclosure provides a data processing method, comprising:

[0006] Iteratively perform model training at least once until a preset iteration stopping condition is met to obtain the target model; wherein, model training includes:

[0007] The initial model is trained using at least one training sample set to obtain at least one training model corresponding to the at least one training sample set. The initial model is obtained by training using the initial training sample set. The feature dimension of the initial training samples in the initial training sample set is greater than the feature dimension of the training samples in each training sample set. Each training sample set includes a labeled training sample set and an unlabeled training sample set.

[0008] Based on the test sample set, determine the accuracy corresponding to the initial model and the accuracy corresponding to at least one trained model, wherein the feature dimension of the test samples in the test sample set is the same as the feature dimension of the training samples in each training sample set; and

[0009] The model with the highest accuracy is determined from the accuracy corresponding to the initial model and the accuracy corresponding to at least one trained model, and the model with the highest accuracy is used as the initial model.

[0010] According to embodiments of this disclosure, training an initial model using at least one set of training samples to obtain at least one trained model corresponding to the at least one set of training samples includes:

[0011] The initial model is trained using multiple sets of training samples to obtain multiple training models, where each training model corresponds to one of the training sample sets in the multiple sets of training samples.

[0012] According to embodiments of this disclosure, training an initial model using multiple training sample sets to obtain multiple training models includes:

[0013] The unlabeled training samples from each of the multiple training sample sets are input into the initial model to obtain multiple training sample sets with predicted labels. Each of these multiple training sample sets with predicted labels corresponds to one of the multiple training sample sets.

[0014] The initial model is trained using multiple sets of labeled training samples and multiple sets of training samples with predicted labels to obtain multiple training models.

[0015] According to embodiments of this disclosure, the preset iteration stop condition includes:

[0016] The accuracy corresponding to at least one trained model is less than or equal to the accuracy corresponding to the initial model.

[0017] According to embodiments of this disclosure, the preset iteration stop condition further includes:

[0018] If the accuracy corresponding to at least one trained model is greater than the accuracy corresponding to the initial model, the number of iterations is greater than or equal to a preset threshold number.

[0019] According to an embodiment of this disclosure, the ratio of the number of labeled training samples in the labeled training sample set to the number of unlabeled training samples in the unlabeled training sample set is 1:1.

[0020] According to embodiments of this disclosure, each test sample in the test sample set includes labeled data;

[0021] Based on the test sample set, determine the accuracy corresponding to the initial model and the accuracy corresponding to at least one trained model, including:

[0022] Each test sample in the test sample set is input into the initial model and at least one training model to obtain a first prediction result set corresponding to the initial model and at least one second prediction result set corresponding to at least one training model.

[0023] Based on the first set of prediction results and the labeled data set in the test sample set, determine the accuracy corresponding to the initial model; and

[0024] Based on at least one second set of prediction results and a set of labeled data in the test sample set, determine the accuracy corresponding to at least one trained model.

[0025] Another aspect of this disclosure provides a data prediction method using a target model trained by data processing methods, comprising:

[0026] Obtain the feature data to be identified, wherein the feature dimension of the feature data to be identified is less than or equal to the feature dimension of at least one training sample in the training sample set; and

[0027] The feature data to be identified is input into the target model to obtain the prediction result corresponding to the feature data to be identified.

[0028] Another aspect of this disclosure provides a data processing apparatus, comprising:

[0029] The model training module is used to train the initial model using at least one training sample set in each model training session, so as to obtain at least one training model corresponding to the at least one training sample set. The initial model is obtained by training using the initial training sample set. The feature dimension of the initial training samples in the initial training sample set is greater than the feature dimension of the training samples in each training sample set. Each training sample set includes a labeled training sample set and an unlabeled training sample set.

[0030] The first determining module is used to determine, based on the test sample set, the accuracy corresponding to the initial model and the accuracy corresponding to at least one training model, respectively, wherein the feature dimension of the test samples in the test sample set is the same as the feature dimension of the training samples in each training sample set; and

[0031] The second determining module is used to determine the model with the highest accuracy from the accuracy corresponding to the initial model and the accuracy corresponding to at least one trained model, and to use the model with the highest accuracy as the initial model.

[0032] The process involves iteratively performing model training at least once until a preset iteration stopping condition is met, thus obtaining the target model.

[0033] Another aspect of this disclosure provides a data prediction apparatus, comprising:

[0034] The acquisition module is used to acquire feature data to be identified; wherein the feature dimension of the feature data to be identified is less than or equal to the feature dimension of at least one training sample in the training sample set; and

[0035] The prediction module is used to input the feature data to be identified into the target model and obtain the prediction result corresponding to the feature data to be identified.

[0036] Another aspect of this disclosure provides a computer system comprising:

[0037] One or more processors;

[0038] Memory, used to store one or more programs.

[0039] When the above one or more programs are executed by the above one or more processors, the above one or more processors implement the above method.

[0040] Another aspect of this disclosure provides a computer-readable storage medium storing computer-executable instructions that, when executed, are used to implement the above-described method.

[0041] Another aspect of this disclosure provides a computer program product including computer-executable instructions that, when executed, implement the method described above.

[0042] According to embodiments of this disclosure, because model training is iteratively performed at least once until a preset iteration stopping condition is met, a target model is obtained; wherein, model training includes: training an initial model using at least one training sample set to obtain at least one training model corresponding to at least one training sample set, wherein the initial model is obtained by training using the initial training sample set, wherein the feature dimension of the initial training samples in the initial training sample set is greater than the feature dimension of the training samples in each training sample set, wherein each training sample set includes a labeled training sample set and an unlabeled training sample set; and based on the test sample set, determining the accuracy corresponding to the initial model and the accuracy corresponding to at least one training sample set. The accuracy corresponding to a training model, wherein the feature dimension of the test samples in the test sample set is the same as the feature dimension of the training samples in each training sample set; and the technical means of determining the model with the highest accuracy from the accuracy corresponding to the initial model and the accuracy corresponding to at least one training model, and using the model with the highest accuracy as the initial model, thereby mining the masked information in the initial training sample set and maximizing the use of the implicit data in the initial training sample set. Therefore, it at least partially overcomes the technical problem in the prior art where the original data is masked and the model trained with the masked data has low recognition accuracy, thereby achieving the technical effect of fully mining the implicit information in the masked data. Attached Figure Description

[0043] The above and other objects, features, and advantages of this disclosure will become clearer from the following description of embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0044] Figure 1 This illustration schematically shows an exemplary system architecture to which the data processing methods and apparatus of this disclosure can be applied according to embodiments of this disclosure;

[0045] Figure 2 This illustration schematically depicts an application scenario of the data processing method according to an embodiment of the present disclosure;

[0046] Figure 3 A flowchart illustrating a data processing method according to an embodiment of the present disclosure is shown schematically.

[0047] Figure 4 A flowchart illustrating a data prediction method according to an embodiment of the present disclosure is shown schematically;

[0048] Figure 5 A flowchart illustrating a data processing method and a data prediction method according to another embodiment of the present disclosure is shown schematically;

[0049] Figure 6 A block diagram of a data processing apparatus according to an embodiment of the present disclosure is shown schematically;

[0050] Figure 7 A block diagram of a data prediction apparatus according to an embodiment of the present disclosure is schematically shown; and

[0051] Figure 8 A block diagram of a computer system suitable for implementing data processing methods and data prediction methods according to embodiments of the present disclosure is shown schematically. Detailed Implementation

[0052] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0053] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0054] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0055] When using expressions such as "at least one of A, B, and C," the expression should generally be interpreted in accordance with the meaning commonly understood by a person skilled in the art (e.g., "a system having at least one of A, B, and C" should include, but is not limited to, systems having A alone, having B alone, having C alone, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.). Similarly, when using expressions such as "at least one of A, B, or C," the expression should generally be interpreted in accordance with the meaning commonly understood by a person skilled in the art (e.g., "a system having at least one of A, B, or C" should include, but is not limited to, systems having A alone, having B alone, having C alone, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.).

[0056] Embodiments of this disclosure provide a data processing method. The method includes iteratively performing at least one model training iteration until a preset iteration stopping condition is met to obtain a target model. The model training includes: training an initial model using at least one training sample set to obtain at least one training model corresponding to the at least one training sample set, wherein the initial model is obtained by training using the initial training sample set, and the feature dimension of the initial training samples in the initial training sample set is greater than the feature dimension of the training samples in each training sample set, wherein each training sample set includes a labeled training sample set and an unlabeled training sample set; determining, based on a test sample set, the accuracy corresponding to the initial model and the accuracy corresponding to the at least one training model, respectively, wherein the feature dimension of the test samples in the test sample set is the same as the feature dimension of the training samples in each training sample set; and determining the model with the highest accuracy from the accuracy corresponding to the initial model and the accuracy corresponding to the at least one training model, and using the model with the highest accuracy as the initial model.

[0057] The data processing method of this disclosure can fully mine the masked information in the initial training sample set and maximize the use of the hidden data in the initial training sample set, thus achieving the technical effect of fully mining the hidden information in the masked data to train the model.

[0058] Embodiments of this disclosure provide a data prediction method using a target model trained by the data processing method described above. The method includes acquiring feature data to be identified; wherein the feature dimension of the feature data to be identified is less than or equal to the feature dimension of at least one training sample in a training sample set; and inputting the feature data to be identified into the target model to obtain a prediction result corresponding to the feature data to be identified.

[0059] The data prediction method using the embodiments of this disclosure yields prediction results with high accuracy and precision.

[0060] Figure 1 An exemplary system architecture 100 to which data processing methods and apparatus can be applied according to embodiments of this disclosure is illustrated. It should be noted that... Figure 1 The examples shown are merely examples of system architectures that can be applied to the embodiments of this disclosure, in order to help those skilled in the art understand the technical content of this disclosure, but do not mean that the embodiments of this disclosure cannot be used in other devices, systems, environments or scenarios.

[0061] like Figure 1As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, and 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the terminal devices 101, 102, and 103 and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0062] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as transaction payment applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software, etc. (for example only).

[0063] Terminal devices 101, 102, and 103 can be various electronic devices with displays and web browsing capabilities, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0064] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using terminal devices 101, 102, and 103 (for example only). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0065] It should be noted that the data processing method provided in this embodiment can be executed by server 105. Correspondingly, the data processing apparatus provided in this embodiment can be located in server 105. The data processing method provided in this embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with terminal devices 101, 102, 103 and / or server 105. Correspondingly, the data processing apparatus provided in this embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with terminal devices 101, 102, 103 and / or server 105.

[0066] For example, at least one training sample set may originally be stored in any one of terminal devices 101, 102, or 103 (e.g., terminal device 101, but not limited thereto), or it may be stored on an external storage device and imported into terminal device 101. Then, terminal device 101 may send at least one training sample set to other terminal devices, servers, or server clusters, and the other servers or server clusters that receive the at least one training sample set may execute the data processing method provided in the embodiments of this disclosure.

[0067] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0068] Figure 2 The illustration depicts an application scenario of the data processing method according to an embodiment of the present disclosure.

[0069] In most industries, data often exists in silos due to issues such as industry competition, privacy and security concerns, and complex administrative procedures. Furthermore, in reality, integrating data scattered across different locations and institutions is nearly impossible, or the required costs are enormous. Even achieving centralized data integration between different departments within the same company faces significant obstacles. Federated machine learning has emerged as an important means to address these problems.

[0070] Federated Machine Learning is a machine learning framework that effectively helps multiple organizations (which can be referred to as collaborators) to use data from multiple collaborators for machine learning modeling and training a unified model while meeting the requirements of user privacy protection, data security and government regulations, without compromising the privacy and security of this data.

[0071] like Figure 2 As shown, for example, there are two collaborators, collaborator A and collaborator B, who jointly train and model the model using a vertical federated learning approach. Collaborator A provides m features fa1, fa2, ..., fam, while collaborator B provides n features fb1, fb2, ..., fbn. These features are combined to obtain m+n features, and the final model, ModelAB, is obtained by training the model using the combined features from collaborators A and B.

[0072] However, because federated learning imposes weak constraints on the collaborating parties, any party may withdraw at any time. In vertical federated learning, all parties are required to participate in both training and inference. Once a collaborating party withdraws, the only solution is to reduce the dimensionality of the training features, necessitating retraining of the model.

[0073] like Figure 2 As shown, due to some reason, collaborator B no longer participates in federated learning, so ModelAB can no longer be used because it cannot provide the fb1, fb2, ..., fbn features required for inference. This will result in the features provided by collaborator B being removed during prediction using ModelAB; that is, all feature information provided by collaborator B is discarded. Only the feature information provided by collaborator A is used for prediction. With only feature information provided by collaborator A, the prediction accuracy and precision of ModelAB will decrease significantly.

[0074] Existing technologies can also utilize the feature information provided by collaborator A to reconstruct and train the model, obtaining Model A. However, using this method will discard Model AB, resulting in wasted preliminary work, time consumption, and increased labor costs. More seriously, the recognition accuracy will decrease.

[0075] According to embodiments of this disclosure, a method for retraining a model is provided using the data processing method provided in the embodiments of this disclosure. In the case of only feature information provided by collaborator A, the method strives to mine the feature information provided by collaborator B hidden in ModelAB, thereby achieving better model accuracy.

[0076] Figure 3 A flowchart illustrating a data processing method according to an embodiment of the present disclosure is shown schematically.

[0077] like Figure 3 As shown, the data processing method includes operations S310 to S340.

[0078] In operation S310, the initial model is trained using at least one training sample set to obtain at least one training model corresponding to the at least one training sample set. The initial model is obtained by training using the initial training sample set. The feature dimension of the initial training samples in the initial training sample set is greater than the feature dimension of the training samples in each training sample set. Each training sample set includes a labeled training sample set and an unlabeled training sample set.

[0079] According to embodiments of this disclosure, the training sample set may include one or more sets, and the number of training samples in each of the multiple training sample sets may be the same or different.

[0080] According to embodiments of this disclosure, the feature dimension can be understood as the number of feature vectors or feature information.

[0081] According to embodiments of this disclosure, the initial training sample set can be the training sample set provided by each collaborator during federated learning, which is a training sample set with relatively complete feature dimensions. However, the training sample set in this disclosure can be the remaining training sample set after one or more collaborators have withdrawn during the federated learning process, and their provided training sample sets have been masked. Because this federated learning is vertical federated learning, after masking the data provided by one or more collaborators, the feature dimensions of the training samples in the remaining training sample set are reduced compared to the initial training sample set.

[0082] According to embodiments of this disclosure, the training sample set in these embodiments includes a labeled training sample set and an unlabeled training sample set. It should be noted that manually labeling samples is a time-consuming and labor-intensive task; therefore, the training sample set in these embodiments includes both labeled and unlabeled training sample sets.

[0083] According to an optional embodiment of this disclosure, an initial model is trained using a training sample set. This initial model can be an initial model trained from the initial training samples through federated learning, without any collaborating parties withdrawing.

[0084] It should be noted that this initial model is a machine learning model, which can be a regression model, but is not limited to this; it can also be a classification model, depending on the actual situation. Furthermore, the loss function used during training can also be selected according to the specific circumstances, which will not be elaborated upon here.

[0085] Training using the initial model from embodiments of this disclosure maximizes the use of information obtained from the federated learning process and saves resources.

[0086] In operation S320, based on the test sample set, the accuracy corresponding to the initial model and the accuracy corresponding to at least one training model are determined respectively, wherein the feature dimension of the test samples in the test sample set is the same as the feature dimension of the training samples in each training sample set.

[0087] According to embodiments of this disclosure, at least one trained model can be obtained by training an initial model using at least one set of training samples. In embodiments of this disclosure, the quality of the trained model can be verified using a test set of samples.

[0088] According to optional embodiments of this disclosure, accuracy can be used for evaluation, but it is not limited to this. Other evaluation criteria can also be used, which will not be elaborated here.

[0089] In operation S330, the model with the highest accuracy is determined from the accuracy corresponding to the initial model and the accuracy corresponding to at least one trained model, and the model with the highest accuracy is used as the initial model.

[0090] According to embodiments of this disclosure, an optimal model can be determined based on accuracy and used as an update to the initial model. Iterative training is then performed in this manner.

[0091] In operation S340, based on the accuracy corresponding to the initial model and the accuracy corresponding to at least one trained model, it is determined whether the preset iteration stopping condition is met, and the target model is obtained.

[0092] According to embodiments of this disclosure, a preset iteration stop condition can be used as a criterion for determining whether the iteration has stopped.

[0093] According to an optional embodiment of this disclosure, the model with the highest accuracy can be used as the target model to replace the initial model.

[0094] The data processing method provided by the embodiments of this disclosure not only does not discard the initial model obtained by federated learning, but also fully mines the masked feature information under the condition of feature dimensionality reduction, maximizes the use of the data value of federated learning, and makes the target model trained subsequently have high recognition accuracy and reduced cost.

[0095] The following describes specific embodiments. Figure 3 The method shown will be further explained.

[0096] According to embodiments of this disclosure, an initial model can be trained using multiple sets of training samples to obtain multiple training models, wherein each of the multiple training models corresponds to one of the training sample sets in the multiple sets of training samples.

[0097] According to embodiments of this disclosure, multiple training sample sets are used to train the initial model to obtain multiple training models. Compared with training the initial model using a single training sample set to obtain only one training model, the efficiency and accuracy of the final target model are relatively higher.

[0098] According to embodiments of this disclosure, unlabeled training samples in the training sample set can be selected from those with rich features as unlabeled training samples.

[0099] According to embodiments of this disclosure, feature saturation can be understood as the feature vectors of each dimension in a multi-dimensional feature vector being able to obtain corresponding data or the feature vector dimensions containing data accounting for the majority of the total number of dimensions, for example, 70%, 80% or 90%.

[0100] According to embodiments of this disclosure, training with training samples rich in features is beneficial for improving model accuracy.

[0101] According to embodiments of this disclosure, unlabeled training samples from each of multiple training sample sets can be input into an initial model to obtain multiple training sample sets with predicted labels, wherein each of the multiple training sample sets with predicted labels corresponds to one training sample set in the multiple training sample sets; and

[0102] The initial model is trained using multiple sets of labeled training samples and multiple sets of training samples with predicted labels to obtain multiple training models.

[0103] According to embodiments of this disclosure, an initial model is used to automatically label unlabeled training samples in an unlabeled training sample set, saving manpower, time, and costs.

[0104] According to embodiments of this disclosure, the ratio of the number of labeled training samples in the labeled training sample set to the number of unlabeled training samples in the unlabeled training sample set is 1:1.

[0105] According to embodiments of this disclosure, the ratio of the number of labeled training samples in the labeled training sample set to the number of unlabeled training samples in the unlabeled training sample set is not limited to 1:1, and can be set according to actual conditions. For example, the number of labeled training samples in the labeled training sample set may be more or less than the number of unlabeled training samples in the unlabeled training sample set.

[0106] However, according to the optional embodiments of this disclosure, the model training effect is best when a ratio of 1:1 is used, which not only results in fast model convergence speed, but also high model recognition accuracy and precision.

[0107] According to embodiments of this disclosure, after evaluating the accuracy of the training model and the initial model using a test sample set, it is determined whether a preset iteration stopping condition is met.

[0108] According to an optional embodiment of this disclosure, the preset iteration stopping condition can be that the accuracy corresponding to at least one trained model is less than or equal to the accuracy corresponding to the initial model. That is, after at least one iteration of model training, the accuracy determined by the trained model is less than the accuracy determined by the initial model. This indicates that continuing the iteration operation will not effectively improve the model's accuracy, and the iterated initial model can be used as the target model.

[0109] It should be noted that once the model has been trained to a certain extent, its convergence speed will decrease and the training time will increase significantly. Using the model's accuracy as the final preset stopping condition for iteration will seriously affect work efficiency.

[0110] According to other embodiments of this disclosure, the number of iterations can also be used as the iteration stopping condition to solve the problem of reduced work efficiency caused by using the model's accuracy as the evaluation criterion. In embodiments of this disclosure, the preset iteration stopping condition is that the number of iterations is greater than or equal to a preset threshold number when the accuracy corresponding to at least one trained model is greater than the accuracy corresponding to the initial model.

[0111] According to embodiments of this disclosure, it should also be noted that the methods for determining the accuracy corresponding to the initial model and the accuracy corresponding to at least one training model based on the test sample set can be performed through the following operations.

[0112] For example, each test sample in the test sample set is input into an initial model and at least one training model to obtain a first prediction result set corresponding to the initial model and at least one second prediction result set corresponding to at least one training model; based on the first prediction result set and the labeled data set in the test sample set, the accuracy corresponding to the initial model is determined; and based on the at least one second prediction result set and the labeled data set in the test sample set, the accuracy corresponding to at least one training model is determined.

[0113] According to embodiments of this disclosure, each test sample in the test sample set includes labeled data. First prediction result data in the first prediction result set can be compared with the labeled data to determine whether it is correct; the accuracy corresponding to the initial model is determined based on the ratio of the number of correct results to the total number of test samples in the total test sample set. Similarly, second prediction result data in the second prediction result set can be compared with the labeled data to determine whether it is correct; the accuracy corresponding to the trained model is determined based on the ratio of the number of correct results to the total number of test samples in the total test sample set.

[0114] According to embodiments of this disclosure, the method for determining whether the first and second predicted result data are correct can be specifically limited according to the model category. For example, for a classification model, the first or second predicted result data can be compared with the corresponding labeled data to see if they are the same. If they are the same, it is correct; if they are different, it is incorrect. However, it is not limited to this. For a regression model, a preset threshold condition can be used to compare the first or second predicted result data with the preset threshold condition. If the preset threshold condition is met, it is correct; otherwise, it is incorrect.

[0115] The method for determining the accuracy of the embodiments disclosed herein is simple, effective, practical, and easy to operate.

[0116] Figure 4 A flowchart illustrating a data prediction method according to an embodiment of the present disclosure is shown schematically.

[0117] like Figure 4 As shown, the data prediction method includes operations S410 to S420.

[0118] In operation S410, feature data to be identified is obtained, wherein the feature dimension of the feature data to be identified is less than or equal to the feature dimension of at least one training sample in the training sample set.

[0119] In operation S420, the feature data to be identified is input into the target model to obtain the prediction result corresponding to the feature data to be identified.

[0120] The target model obtained using the data processing method provided in this disclosure minimizes the loss of recognition accuracy caused by the withdrawal of collaborators. The target model obtained using the data processing method of this disclosure can identify or predict feature data with small feature dimensions, and the identification and prediction accuracy is high.

[0121] The technical solutions of this disclosure will be further described below with reference to specific embodiments. However, it should be noted that the following embodiments are only used to illustrate the technical solutions of this disclosure, but this disclosure is not limited thereto.

[0122] The following is a detailed description of an embodiment of this disclosure, using residential community pedestrian flow prediction as an example.

[0123] In the field of advertising placement decisions, pedestrian traffic is a crucial indicator, and residential communities are a significant scenario. Machine learning algorithms can be used to predict pedestrian traffic in residential communities by collecting multiple relevant features for each community. These features include, for example, community area, number of building units, number of parking spaces, building year, and reported business locations. Through feature collection, feature data for 400,000 residential communities nationwide can be obtained.

[0124] For example, Company A selects 1,000 residential communities and obtains the pedestrian traffic of the corresponding residential communities through data collection surveys or purchases, which serves as the labeled data for the corresponding samples.

[0125] Using the above 1,000 data points, combined with machine learning regression algorithms, a model can be trained to predict and infer the pedestrian flow in a 400,000-square-meter residential community.

[0126] However, the above prediction method may have problems. Among the collected features, there may be insufficient feature dimensions or insufficient feature fullness. For example, the reported points of Company A are relatively sparse.

[0127] To address the above issues, more effective features can be incorporated through federated learning.

[0128] For example, through federated learning, feature data provided by one or more collaborators can be accessed, such as the reporting points of Company B, the male-to-female ratio, and other features.

[0129] By combining the characteristics of Company A and Company B, a regression model ModelAB is trained, thereby enabling the prediction of pedestrian flow in residential communities.

[0130] However, for some reason, Company B withdrew from the federated learning process. Because of Company B's withdrawal, features provided by Company B, such as its reporting locations and gender ratio, cannot be used, reducing the feature dimensionality. Therefore, when predicting pedestrian traffic in a residential area, using only Company A's features and ModelAB will significantly reduce prediction accuracy, rendering the predictions unusable as accurate results.

[0131] In this context, the data processing and prediction methods provided in this disclosure can be employed to fully extract high-value information that has been masked, thereby obtaining a new model with high prediction accuracy, while reducing the feature dimensions. This minimizes the accuracy loss caused by the withdrawal of collaborating parties.

[0132] Figure 5 A flowchart illustrating a data processing method and a data prediction method according to embodiments of the present disclosure is shown.

[0133] like Figure 5 As shown, operations S510 to S560 can be performed.

[0134] When operating S510, prepare data and models.

[0135] The prepared data may include labeled samples and unlabeled samples. The labeled samples may be, for example, feature samples (labeled data) of 1,300 residential communities with prepared labeled data (i.e., known pedestrian traffic). The unlabeled samples may be 20,000 residential communities with rich features and unknown pedestrian traffic selected from 400,000 residential communities.

[0136] Alternatively, the labeled samples can be further divided according to a certain ratio, for example, selecting 1000 samples as training samples and the remaining 300 samples as test samples.

[0137] The prepared model can be the initial model (i.e., bestModel) using the pre-trained federated model ModelAB.

[0138] When operating S520, the bestModel is used to predict the 20,000 unlabeled samples selected to obtain the prediction results, i.e., pseudo labels.

[0139] When operating the S530, the samples containing pseudo-labels are split and the training data is combined separately to train the initial model.

[0140] The unlabeled samples that have already been pseudo-labeled are split into, for example, 20 parts. Each part contains 1000 entries, which are then combined with 1000 entries of training data with true labels (i.e., labeled data) in a 1:1 ratio. Using only the features of Company A, such as the area of ​​the residential area, the number of units in the building, the number of parking spaces, the year of construction, and the location reported by Company A, the initial model is trained to obtain 20 training models.

[0141] When operating the S540, test data is used to select the model with the highest accuracy from the initial model and the trained model.

[0142] The 20 trained models and the initial model were used to predict the test data, and the accuracy of the 20 trained models (newacc1, newacc2, ..., newacc20) and the accuracy of the initial model (acc) were obtained. The model with the highest accuracy was selected as the new model.

[0143] Among them, the model with the highest accuracy can be determined by comparing newacc and acc. For example, update acc = max(acc, newacc1, newacc2, ..., newacc20) and update the corresponding initial model bestmodel.

[0144] In operation S550, it is determined whether the preset iteration stop condition is met. If yes, the target model is determined and the process exits. Otherwise, the iteration begins, i.e., operations S520 to S550 are executed iteratively.

[0145] The stopping condition for iterative training can be that max(newacc1, newacc2, ..., newacc20) is less than or equal to acc. Alternatively, if max(newacc1, newacc2, ..., newacc20) is greater than acc, the number of iterations is greater than or equal to the preset number of iterations.

[0146] In operation S560, the features of Company A for a residential community to be predicted, provided by Company A, are input into the target model to obtain the pedestrian flow information of the residential community to be predicted.

[0147] It should be noted that the collaborators in federated learning can be Company A and Company B, but are not limited to these; multiple companies can also collaborate, such as Company A, Company B, Company C, and Company D. Furthermore, companies that withdraw midway through the process are not limited to a single company; multiple companies can also participate.

[0148] Furthermore, the data processing method and data prediction method provided in this disclosure are not limited to the prediction of pedestrian traffic in residential communities. They can also be applied to other fields such as finance and sales, depending on the specific circumstances, which will not be elaborated here.

[0149] In summary, during federated learning, when a collaborating party withdraws, the data processing and prediction methods provided in this disclosure can be employed. With a reduced feature dimension, the model obtained through federated learning can fully extract previously hidden high-value information, thereby yielding a new model with high prediction accuracy. This minimizes the accuracy loss caused by the withdrawal of collaborating parties.

[0150] Figure 6 A block diagram of a data processing apparatus according to an embodiment of the present disclosure is shown schematically.

[0151] like Figure 6 As shown, the data processing device 600 includes a model training module 610, a first determination module 620, and a second determination module 630.

[0152] The model training module 610 is used to train the initial model using at least one training sample set in each model training session to obtain at least one training model corresponding to the at least one training sample set. The initial model is obtained by training using the initial training sample set. The feature dimension of the initial training samples in the initial training sample set is greater than the feature dimension of the training samples in each training sample set. Each training sample set includes a labeled training sample set and an unlabeled training sample set.

[0153] The first determining module 620 is used to determine, based on the test sample set, the accuracy corresponding to the initial model and the accuracy corresponding to at least one training model, respectively, wherein the feature dimension of the test samples in the test sample set is the same as the feature dimension of the training samples in each training sample set; and

[0154] The second determining module 630 is used to determine the model with the highest accuracy from the accuracy corresponding to the initial model and the accuracy corresponding to at least one trained model, and to use the model with the highest accuracy as the initial model.

[0155] The process involves iteratively performing model training at least once until a preset iteration stopping condition is met, thus obtaining the target model.

[0156] According to the embodiments of this disclosure, by making reasonable use of the initial model obtained by federated learning, and fully mining the masked feature information under the condition of feature dimensionality reduction, the data value of federated learning is maximized, so that the target model trained subsequently has high recognition accuracy and reduced cost.

[0157] According to embodiments of this disclosure, the model training module 610 includes a model training submodule.

[0158] The model training submodule is used to train the initial model using multiple sets of training samples to obtain multiple training models, wherein each training model corresponds to one of the training sample sets in the multiple sets of training samples.

[0159] According to embodiments of this disclosure, the model training submodule includes a first training unit and a second training unit.

[0160] The first training unit is used to input unlabeled training samples from each of the multiple training sample sets into the initial model, resulting in multiple training sample sets with predicted labels. Each of these training sample sets with predicted labels corresponds to one of the multiple training sample sets.

[0161] The second training unit is used to train the initial model based on multiple sets of labeled training samples and multiple sets of training samples with predicted labels, so as to obtain multiple training models.

[0162] According to embodiments of this disclosure, the preset iteration stopping condition includes an accuracy corresponding to at least one trained model that is less than or equal to the accuracy corresponding to the initial model.

[0163] According to embodiments of this disclosure, the preset iteration stopping condition further includes that the number of iterations is greater than or equal to a preset threshold number when the accuracy corresponding to at least one trained model is greater than the accuracy corresponding to the initial model.

[0164] According to an embodiment of this disclosure, the ratio of the number of labeled training samples in the labeled training sample set to the number of unlabeled training samples in the unlabeled training sample set is 1:1.

[0165] According to embodiments of this disclosure, each test sample in the test sample set includes labeled data.

[0166] According to embodiments of this disclosure, the first determining module includes an input unit, a first determining unit, and a second determining unit.

[0167] The input unit is used to input each test sample in the test sample set into the initial model and at least one training model respectively, to obtain a first prediction result set corresponding to the initial model and at least one second prediction result set corresponding to at least one training model;

[0168] The first determining unit is configured to determine the accuracy corresponding to the initial model based on the first prediction result set and the labeled data set in the test sample set; and

[0169] The second determining unit is used to determine the accuracy corresponding to at least one training model based on at least one second prediction result set and a set of labeled data in a test sample set.

[0170] Figure 7 A block diagram of a data prediction apparatus according to an embodiment of the present disclosure is shown schematically.

[0171] like Figure 7 As shown, the data processing device 700 includes an acquisition module 710 and a prediction module 720.

[0172] The acquisition module 710 is used to acquire feature data to be identified; wherein the feature dimension of the feature data to be identified is less than or equal to the feature dimension of at least one training sample in the training sample set; and

[0173] The prediction module 720 is used to input the feature data to be identified into the target model and obtain the prediction result corresponding to the feature data to be identified.

[0174] According to embodiments of this disclosure, the loss of recognition accuracy due to the withdrawal of collaborating parties is minimized. The target model obtained using embodiments of this disclosure can identify or predict feature data with small feature dimensions, and its recognition and prediction accuracy is higher than that of models constructed through federated learning, which require large feature dimensions.

[0175] Any one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure, or at least part of the functions of any one or more of them, can be implemented in one module. Any one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure can be implemented by dividing them into multiple modules. Any one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure can be at least partially implemented as hardware circuitry, such as Field Programmable Gate Arrays (FPGAs), Programmable Logic Arrays (PLAs), Systems-on-Chip, Systems-on-Substrate, Systems-on-Package, Application-Specific Integrated Circuits (ASICs), or implemented in hardware or firmware by any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure can be at least partially implemented as computer program modules, which, when run, can perform corresponding functions.

[0176] For example, any multiple of the model training module 610, the first determination module 620, and the second determination module 630, or the acquisition module 710 and the prediction module 720, can be combined into one module / unit / subunit, or any one of these modules / units / subunits can be split into multiple modules / units / subunits. Alternatively, at least part of the functionality of one or more of these modules / units / subunits can be combined with at least part of the functionality of other modules / units / subunits and implemented in one module / unit / subunit. According to embodiments of this disclosure, at least one of the model training module 610, the first determination module 620, and the second determination module 630, or the acquisition module 710 and the prediction module 720, can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the model training module 610, the first determining module 620, and the second determining module 630, or the acquisition module 710 and the prediction module 720, can be at least partially implemented as a computer program module that can perform corresponding functions when the computer program module is run.

[0177] It should be noted that the data processing device part in the embodiments of this disclosure corresponds to the data processing method part in the embodiments of this disclosure. The specific description of the data processing device part is referred to in the data processing method part, and will not be repeated here.

[0178] Figure 8 A block diagram of a computer system suitable for implementing the methods described above, according to embodiments of the present disclosure, is illustrated schematically. Figure 8 The computer system shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0179] like Figure 8 As shown, a computer system 800 according to an embodiment of this disclosure includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage portion 808 into a random access memory (RAM) 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this disclosure.

[0180] RAM 803 stores various programs and data required for the operation of system 800. Processor 801, ROM 802, and RAM 803 are interconnected via bus 804. Processor 801 performs various operations of the method flow according to embodiments of the present disclosure by executing programs in ROM 802 and / or RAM 803. It should be noted that the programs may also be stored in one or more memories other than ROM 802 and RAM 803. Processor 801 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in said one or more memories.

[0181] According to embodiments of this disclosure, system 800 may further include an input / output (I / O) interface 805, which is also connected to bus 804. System 800 may also include one or more of the following components connected to I / O interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, modem, etc. Communication section 809 performs communication processing via a network such as the Internet. Drive 810 is also connected to I / O interface 805 as needed. Removable media 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 810 as needed so that computer programs read from there are installed into storage section 808 as needed.

[0182] According to embodiments of this disclosure, the method flow according to embodiments of this disclosure can be implemented as a computer software program. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for performing the methods shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network via communication section 809, and / or installed from removable medium 811. When the computer program is executed by processor 801, it performs the functions defined in the system of embodiments of this disclosure. According to embodiments of this disclosure, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0183] This disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs that, when executed, implement the method according to the embodiments of this disclosure.

[0184] According to embodiments of this disclosure, the computer-readable storage medium can be a non-volatile computer-readable storage medium. Examples include, but are not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this disclosure, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0185] For example, according to embodiments of this disclosure, a computer-readable storage medium may include the ROM 802 and / or RAM 803 described above and / or one or more memories other than ROM 802 and RAM 803.

[0186] Embodiments of this disclosure also include a computer program product comprising a computer program containing program code for performing the methods provided in the embodiments of this disclosure. When the computer program product is run on an electronic device, the program code is used to enable the electronic device to implement the data processing methods provided in the embodiments of this disclosure.

[0187] When the computer program is executed by the processor 801, it performs the functions defined in the system / apparatus of this disclosure embodiments. According to embodiments of this disclosure, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0188] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 809, and / or installed from a removable medium 811. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0189] According to embodiments of this disclosure, program code for executing the computer programs provided in embodiments of this disclosure can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can execute entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0190] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions. Those skilled in the art will understand that the features recited in the various embodiments and / or claims of this disclosure can be combined and / or combined in various ways, even if such combinations or combinations are not expressly described in this disclosure. In particular, the features described in the various embodiments and / or claims of this disclosure may be combined and / or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0191] The embodiments of this disclosure have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of this disclosure. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. The scope of this disclosure is defined by the appended claims and their equivalents. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of this disclosure, and all such substitutions and modifications should fall within the scope of this disclosure.

Claims

1. A data processing method, comprising: The model training is performed iteratively at least once until a preset iteration stopping condition is met to obtain the target model; wherein, the model training includes: An initial model is trained using at least one set of training samples to obtain at least one training model corresponding to the at least one set of training samples. The initial model is obtained by training using the initial set of training samples. The feature dimension of the initial training samples in the initial set of training samples is greater than the feature dimension of the training samples in each set of training samples. Each set of training samples includes a set of labeled training samples and a set of unlabeled training samples. Based on the test sample set, the accuracy corresponding to the initial model and the accuracy corresponding to the at least one training model are determined respectively, wherein the feature dimension of the test samples in the test sample set is the same as the feature dimension of each training sample in the training sample set; and The model with the highest accuracy is determined from the accuracy corresponding to the initial model and the accuracy corresponding to the at least one trained model, and the model with the highest accuracy is used as the initial model.

2. The method according to claim 1, wherein, The step of training the initial model using at least one set of training samples to obtain at least one trained model corresponding to the at least one set of training samples includes: The initial model is trained using multiple sets of training samples to obtain multiple training models, wherein each of the multiple training models corresponds to one of the training sample sets in the multiple sets of training samples.

3. The method according to claim 2, wherein, The process of training the initial model using multiple training sample sets to obtain multiple trained models includes: The unlabeled training samples from each of the multiple training sample sets are input into the initial model to obtain multiple training sample sets with predicted labels, wherein each of the multiple training sample sets with predicted labels corresponds to one of the training sample sets in the multiple training sample sets; and The initial model is trained based on multiple sets of labeled training samples and multiple sets of training samples with predicted labels to obtain multiple training models.

4. The method according to claim 1, wherein, The preset iteration stop condition includes: The accuracy corresponding to the at least one trained model is less than or equal to the accuracy corresponding to the initial model.

5. The method according to claim 4, wherein, The preset iteration stopping condition also includes: If the accuracy corresponding to the at least one trained model is greater than the accuracy corresponding to the initial model, the number of iterations is greater than or equal to a preset threshold number.

6. The method according to claim 1, wherein, The ratio of the number of labeled training samples in the labeled training sample set to the number of unlabeled training samples in the unlabeled training sample set is 1:

1.

7. The method according to claim 1, wherein, Each test sample in the test sample set includes labeled data; The determination of the accuracy corresponding to the initial model and the accuracy corresponding to the at least one trained model based on the test sample set includes: Each test sample in the test sample set is input into the initial model and the at least one training model to obtain a first prediction result set corresponding to the initial model and at least one second prediction result set corresponding to the at least one training model. Based on the first set of prediction results and the set of labeled data in the test sample set, the accuracy corresponding to the initial model is determined. as well as Based on the at least one set of second prediction results and the set of labeled data in the test sample set, the accuracy corresponding to the at least one training model is determined.

8. A data prediction method using a target model trained by the data processing method as described in any one of claims 1 to 7, comprising: Obtain feature data to be identified, wherein the feature dimension of the feature data to be identified is less than or equal to the feature dimension of the training samples in the at least one training sample set; and The feature data to be identified is input into the target model to obtain the prediction result corresponding to the feature data to be identified.

9. A data processing apparatus, comprising: The model training module is used to train an initial model using at least one set of training samples in each model training session to obtain at least one training model corresponding to the at least one set of training samples. The initial model is obtained by training using the initial set of training samples. The feature dimension of the initial training samples in the initial set of training samples is greater than the feature dimension of the training samples in each set of training samples. Each set of training samples includes a set of labeled training samples and a set of unlabeled training samples. The first determining module is configured to determine, based on a test sample set, the accuracy corresponding to the initial model and the accuracy corresponding to the at least one training model, respectively, wherein the feature dimension of the test samples in the test sample set is the same as the feature dimension of the training samples in each of the training sample sets; and The second determining module is used to determine the model with the highest accuracy from the accuracy corresponding to the initial model and the accuracy corresponding to the at least one trained model, and to use the model with the highest accuracy as the initial model; The process involves iteratively performing model training at least once until a preset iteration stopping condition is met, thus obtaining the target model.

10. A data prediction apparatus for performing data prediction using a target model trained by the data processing apparatus as described in claim 9, comprising: An acquisition module is used to acquire feature data to be identified; wherein the feature dimension of the feature data to be identified is less than or equal to the feature dimension of the training samples in the at least one training sample set; and The prediction module is used to input the feature data to be identified into the target model to obtain the prediction result corresponding to the feature data to be identified.

11. A computer system, comprising: One or more processors; Memory, used to store one or more programs. Wherein, when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method of any one of claims 1 to 7 or 8.

12. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to perform the method of any one of claims 1 to 7 or 8.

13. A computer program product comprising: Computer-executable instructions, when executed, are used to implement the method of any one of claims 1 to 7 or 8.

Images