A data processing method, device and equipment

Abstract

The present invention provides a data processing method, device and equipment, which relate to the field of communication technology. The method comprises: when a target device is enabled, obtaining first target data of the target device corresponding to the current user and / or scene; training a first decision model according to the first target data to obtain a second decision model; wherein the first decision model is a common decision model determined by training the second target data, and the second target data is data of users and / or scenes under multiple tasks; calling the second decision model to make a decision on the third target data of the current user and / or scene. The solution of the present invention solves the problem that the factory-cured decision model cannot be applied.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a data processing method, device and equipment. Background Art

[0002] With the intelligent development of the Internet of Things, all devices are connected and intelligent. Currently, there are many Internet of Things application services that facilitate people's production and life.

[0003] However, in current IoT devices, the training of decision models is usually completed before leaving the factory, and a large amount of user data or scenario data is indiscriminately collected to build the model. As a result, after the device is delivered and put into use, for a specific user or usage environment, the actual performance of the model may be far lower than the reference value in the model test due to changes in the spatial distribution of the generated data, making the decision of the IoT device unable to meet the personalized needs of users and scenarios. Summary of the invention

[0004] The purpose of the present invention is to provide a data processing method, device and equipment to solve the problem that the decision of the equipment cannot be applied.

[0005] To achieve the above object, an embodiment of the present invention provides a data processing method comprising:

[0006] When the target device is enabled, obtaining first target data of the target device corresponding to the current user and / or scene;

[0007] The first decision model is trained according to the first target data to obtain a second decision model; wherein the first decision model is a common decision model determined by training according to the second target data, and the second target data is data of users and / or scenarios under multiple tasks;

[0008] The second decision model is called to make a decision on the third target data of the current user and / or scene.

[0009] Optionally, before acquiring the first target data of the target device corresponding to the current user and / or scene, the method further includes:

[0010] acquiring the second target data;

[0011] Training a preset decision model according to the second target data to obtain a third decision model corresponding to each task; wherein the preset decision model is a neural network model;

[0012] The first decision model is determined according to the third decision model.

[0013] Optionally, determining the first decision model according to the third decision model includes:

[0014] selecting the mode of the model structure parameter of the third decision model as the value of the model structure parameter of the first decision model;

[0015] According to the training results of each iteration in the training process of the third decision model, the initial values ​​of the model calculation parameters of the first decision model are determined.

[0016] Optionally, the training result includes values ​​of model calculation parameters obtained after the current iteration;

[0017] The step of determining the initial value of the model calculation parameter of the first decision model according to the training result of each iteration in the training process of the third decision model comprises:

[0018] Through the formula φ * =argmin φ L(φ), calculate the initial value φ of the model calculation parameter * ;in, θ n Indicates the value of the model calculation parameter after the nth iteration, l n (θ n ) represents the loss function of the model after the nth iteration, and k represents the total number of iterations corresponding to the third decision model.

[0019] To achieve the above object, an embodiment of the present invention provides a data processing device, including:

[0020] A first acquisition module, used for acquiring first target data of the target device corresponding to a current user and / or scene when the target device is enabled;

[0021] A first processing module, configured to train a first decision model according to the first target data to obtain a second decision model; wherein the first decision model is a common decision model determined by training the second target data, and the second target data is data of users and / or scenarios under multiple tasks;

[0022] The second processing module is used to call the second decision model to make a decision on the third target data of the current user and / or scene.

[0023] Optionally, the device further comprises:

[0024] A second acquisition module, used for acquiring the second target data;

[0025] a third processing module, configured to train a preset decision model according to the second target data to obtain a third decision model corresponding to each task; wherein the preset decision model is a neural network model;

[0026] The fourth processing module is used to determine the first decision model according to the third decision model.

[0027] Optionally, the fourth processing module includes:

[0028] a first processing submodule, configured to select the mode of the model structure parameter of the third decision model as the value of the model structure parameter of the first decision model;

[0029] The second processing submodule is used to determine the initial value of the model calculation parameter of the first decision model according to the training result of each iteration in the training process of the third decision model.

[0030] Optionally, the training result includes values ​​of model calculation parameters obtained after the current iteration;

[0031] The second processing submodule is further used for:

[0032] Through the formula φ * =argmin φ L(φ), calculate the initial value φ of the model calculation parameter * ;in, θ n Indicates the value of the model calculation parameter after the nth iteration, l n (θ n ) represents the loss function of the model after the nth iteration, and k represents the total number of iterations corresponding to the third decision model.

[0033] To achieve the above object, an embodiment of the present invention provides a communication device, including a transceiver and a processor; wherein:

[0034] The transceiver is used for acquiring first target data corresponding to a current user and / or scene of the target device when the target device is enabled;

[0035] The processor is used to train a first decision model according to the first target data to obtain a second decision model; wherein the first decision model is a common decision model determined by training the second target data, and the second target data is data of users and / or scenarios under multiple tasks;

[0036] The processor is further configured to call the second decision model to make a decision on third target data of the current user and / or scene.

[0037] Optionally, the transceiver is further used to obtain the second target data;

[0038] The processor is further configured to: train a preset decision model according to the second target data to obtain a third decision model corresponding to each task; wherein the preset decision model is a neural network model;

[0039] The processor is further configured to: determine a first decision model according to the third decision model.

[0040] Optionally, the processor is further configured to:

[0041] selecting the mode of the model structure parameter of the third decision model as the value of the model structure parameter of the first decision model;

[0042] According to the training results of each iteration in the training process of the third decision model, the initial values ​​of the model calculation parameters of the first decision model are determined.

[0043] Optionally, the training result includes values ​​of model calculation parameters obtained after the current iteration;

[0044] The processor is further configured to:

[0045] Through the formula φ * =argmin φ L(φ), calculate the initial value φ of the model calculation parameter * ;in, θ n Indicates the value of the model calculation parameter after the nth iteration, l n (θ n ) represents the loss function of the model after the nth iteration, and k represents the total number of iterations corresponding to the third decision model.

[0046] To achieve the above-mentioned purpose, an embodiment of the present invention provides a communication device, including a transceiver, a processor, a memory, and a computer program stored in the memory and executable on the processor; when the processor executes the computer program, the data processing method as described above is implemented.

[0047] To achieve the above objective, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps in the data processing method described above are implemented.

[0048] The beneficial effects of the above technical solution of the present invention are as follows:

[0049] The method of the embodiment of the present invention, when the target device is enabled, obtains the data of the current user and / or scene corresponding to the target device as the first target data, thereby training the first decision model based on the first target data to obtain the second decision model. Since the first decision model is a common decision model determined by training the data of users and / or scenes under multiple tasks, and then undergoes personalized training with the first target data, the obtained second decision model is a personalized decision model established by learning the characteristics of the current user and / or scene, so that the target device can call the second decision model to make a decision on the third target data of the current user and / or scene, so that the decision of the target device is more suitable for the current user or scene, thereby improving the applicability of the target device. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 One of the flow charts of the data processing method according to an embodiment of the present invention;

[0051] Figure 2 This is a second flowchart of the data processing method according to an embodiment of the present invention;

[0052] Figure 3 It is a schematic diagram of the application of the method of the embodiment of the present invention;

[0053] Figure 4 is a structural diagram of a data processing device according to an embodiment of the present invention;

[0054] Figure 5 is a structural diagram of a communication device according to an embodiment of the present invention;

[0055] Figure 6 This is a structural diagram of a communication device according to another embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0057] It should be understood that the references to "one embodiment" or "an embodiment" throughout the specification mean that the specific features, structures, or characteristics associated with the embodiment are included in at least one embodiment of the present invention. Therefore, the references to "in one embodiment" or "in an embodiment" appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0058] In various embodiments of the present invention, it should be understood that the size of the serial numbers of the following processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0059] Additionally, the terms "system" and "network" are often used interchangeably herein.

[0060] In the embodiments provided in the present application, it should be understood that "B corresponding to A" means that B is associated with A, and B can be determined according to A. However, it should also be understood that determining B according to A does not mean determining B only according to A, and B can also be determined according to A and / or other information.

[0061] like Figure 1 As shown, a data processing method according to an embodiment of the present invention includes:

[0062] Step 101, when a target device is enabled, obtaining first target data of the target device corresponding to a current user and / or scene;

[0063] Step 102: training a first decision model according to the first target data to obtain a second decision model; wherein the first decision model is a common decision model determined by training according to the second target data, and the second target data is data of users and / or scenarios under multiple tasks;

[0064] Step 103: Call the second decision model to make a decision on the third target data of the current user and / or scene.

[0065] Here, the first decision model is a common decision model determined by training data of users and / or scenes under multiple tasks. The method of the embodiment of the present invention, through steps 101 and 102, first obtains the data of the current user and / or scene corresponding to the target device as the first target data when the target device is enabled, so as to train the first decision model based on the first target data to obtain the second decision model. Since the first decision model is a common decision model determined by training data of users and / or scenes under multiple tasks, and then undergoes personalized training of the first target data, the obtained second decision model is a personalized decision model established by learning the characteristics of the current user and / or scene. In this way, the target device can call the second decision model to make a decision on the third target data of the current user and / or scene through step 103, so that the decision of the target device is more suitable for the current user or scene, thereby improving the applicability of the target device.

[0066] It should be known that the decision model is mostly used for communication devices of the Internet of Things network, such as vehicle terminals in the Internet of Vehicles, smart home devices (such as cameras, speakers, infrared sensors, etc.), wearable smart devices (such as smart watches, smart glasses), etc. In this way, the method of the embodiment of the present invention can be applied to the use of communication devices of the Internet of Things network, and by learning the current user and / or scenario, a personalized decision model is established to improve the applicability of the communication device.

[0067] Taking the camera as an example, before activation, a common decision model, namely the first decision model, has been obtained through data training of users and / or scenes under multiple tasks. When the camera is activated, it can obtain the data of the current user and / or scene as the first target data, and retrain the first decision model based on the first target data to obtain a second decision model that is more suitable for the current user and / or scene. In this way, for different application scenarios, such as day or night, the decision model can be personalized after the camera is activated, and then the decision model suitable for the current scene is called to make a decision. Assuming that the decision model is used to decide the image brightness adjustment strategy, and the current scene is night, the adjusted decision model will be more suitable for night. The adjusted decision model is called to make a decision on the current scene data, and the resulting image brightness adjustment strategy must be an image brightness adjustment strategy suitable for night, which improves the applicability of the camera.

[0068] The multiple tasks corresponding to the second target data are set to obtain training data for more diverse users or scenarios. For example, for smart air conditioners, multiple tasks are respectively for different seasons or for users of different age groups. The third target data is the data obtained again corresponding to the current user and / or scenario after the personalized training of the decision model, after the first target data.

[0069] In this embodiment, in order to improve the timeliness of personalized training, optionally, before step 101, Figure 2 As shown, it also includes:

[0070] Step 201, obtaining the second target data;

[0071] Step 202, training a preset decision model according to the second target data to obtain a third decision model corresponding to each task; wherein the preset decision model is a neural network model;

[0072] Step 203: Determine the first decision model according to the third decision model.

[0073] Here, the preset decision model is the model originally set. Optionally, the preset decision model is a neural network model. After obtaining the data of users and / or scenarios under multiple tasks as the second target data through the above steps 201-203, the preset decision model can be trained according to the second target data to obtain a third decision model corresponding to each task, and finally the first decision model is determined by the obtained third decision model. By pre-establishing an initial decision model preferred by the public, the personalized training time can be reduced.

[0074] Of course, based on the preset decision model, the first decision model, the second decision model and the third decision model will also be neural network models. Among them, the neural network model can also adopt a deep neural network model.

[0075] In this embodiment, the parameters of the decision model mainly include: model structure parameters and model calculation parameters. Among them, the model structure parameters include the number of neurons, the number of channels, the activation function, etc. After determining the model structure parameters, the main structure of the model is obtained, and the model training is mainly aimed at the model calculation parameters. The specific implementation of the model calculation parameters is based on the functional settings of the decision model, which will not be listed here.

[0076] Thus, optionally, step 203 includes:

[0077] selecting the mode of the model structure parameter of the third decision model as the value of the model structure parameter of the first decision model;

[0078] According to the training results of each iteration in the training process of the third decision model, the initial values ​​of the model calculation parameters of the first decision model are determined.

[0079] Considering that the same parameter may obtain multiple results, here, firstly, for the model structure parameters of the first decision model, the mode of all model structure parameters can be selected as the value of the model structure parameters of the first decision model based on the third decision model obtained by multiple tasks. For the model calculation parameters of the first decision model, the initial values ​​of the model calculation parameters of the first decision model can be determined based on the training results of each iteration during the training process of the third decision model.

[0080] In addition, considering that the parameters are initialized with random numbers, more training samples are needed, and multiple iterations are required to obtain the optimal parameter values. Therefore, optionally, in this embodiment, the training results include the values ​​of the model calculation parameters obtained after the current iteration;

[0081] The step of determining the initial value of the model calculation parameter of the first decision model according to the training result of each iteration in the training process of the third decision model comprises:

[0082] Through the formula φ * =argmin φ L(φ), calculate the initial value φ of the model calculation parameter * ;in, θ n Indicates the value of the model calculation parameter after the nth iteration, l n (θ n ) represents the loss function of the model after the nth iteration, and k represents the total number of iterations corresponding to the third decision model.

[0083] In this way, combined with the training results of each iteration in the training process of the third decision model, all intermediate values ​​in the model calculation parameter calculation process are added to the loss function, the control of parameter calculation is strengthened, and the initial value of the model calculation parameter of the first decision model is obtained as the optimized value, and a small amount of the first target data can be used as a sample to complete the training, and the number of iterations in the training will also be reduced. In this way, the communication device using the method of the embodiment of the present invention does not need to have strong processing power and storage performance.

[0084] Among them, the initial value of the model calculation parameter φ * The solution process is as follows:

[0085]

[0086] In practical problems, the second-order derivatives and above can be ignored in the calculation process. Assuming that the training data of the third decision model has N tasks, according to the above solution process, the training data in task i is used for k iterations in the calculation process to obtain the value of the model calculation parameter Calculate the gradient value Use this as the gradient to update φ, and after m updates, the optimal value of φ is obtained. * , which is the initial value of the model calculation parameter of the first decision model. Among them, β is the learning rate, Computes the values ​​of the parameters for the model from the previous iteration.

[0087] Take the method of applying the embodiment of the present invention to a camera as an example. Figure 3 As shown, firstly, a task set (T1, T2, ..., T N ) to obtain the second target data. Then, the preset model is trained according to the second target data corresponding to the task set until the training of N tasks is completed. After that, the mode of all model structure parameters obtained by training can be selected as the value of the model structure parameters of the first decision model, that is, the optimal model structure is determined. In addition, the random initialization of the model calculation parameters can refer to the training results of each iteration in the training process of the third decision model to determine the initial values ​​of the model calculation parameters of the first decision model. Finally, new task data, that is, the data of the current user and / or scene as the first target data, is obtained, and the first decision model with the optimal model structure and model calculation parameters as the above initial values ​​is trained to obtain a personalized second decision model. In this way, the camera can train the initial decision model to make facial recognition decisions, such as unlocking, alarming, etc., based on a small number of facial image samples provided by the user. Moreover, the training process only needs to use limited computing and storage performance to quickly train a decision model that meets user needs. In this way, the subsequent camera can call the personalized trained decision model to make facial recognition decisions, which improves the convenience and reliability of device use.

[0088] In addition, for smart home devices, an easy-to-train initialization model is obtained by training with common data from many users. After training a personalized model based on a small amount of data from a specific household, smart home devices can call on the personalized model to make decisions that are in line with the user's actual usage habits and needs, helping to increase the types and quality of services that can be provided.

[0089] In summary, the method of the embodiment of the present invention first obtains the data of the current user and / or scene corresponding to the target device as the first target data when the target device is enabled, and then trains the first decision model based on the first target data to obtain the second decision model. Since the first decision model is a common decision model determined by training the data of users and / or scenes under multiple tasks, and then undergoes personalized training with the first target data, the obtained second decision model is a personalized decision model established by learning the characteristics of the current user and / or scene. In this way, the target device can call the second decision model to make a decision on the third target data of the current user and / or scene, so that the decision of the target device is more suitable for the current user or scene, thereby improving the applicability of the target device.

[0090] like Figure 4 As shown, a data processing device according to an embodiment of the present invention includes:

[0091] A first acquisition module 410 is used to acquire first target data of the target device corresponding to a current user and / or scene when the target device is enabled;

[0092] A first processing module 420 is used to train a first decision model according to the first target data to obtain a second decision model; wherein the first decision model is a common decision model determined by training the second target data, and the second target data is data of users and / or scenarios under multiple tasks;

[0093] The second processing module 430 is used to call the second decision model to make a decision on the third target data of the current user and / or scene.

[0094] Optionally, the device further comprises:

[0095] A second acquisition module, used for acquiring the second target data;

[0096] a third processing module, configured to train a preset decision model according to the second target data to obtain a third decision model corresponding to each task; wherein the preset decision model is a neural network model;

[0097] The fourth processing module is used to determine the first decision model according to the third decision model.

[0098] Optionally, the fourth processing module includes:

[0099] a first processing submodule, configured to select the mode of the model structure parameter of the third decision model as the value of the model structure parameter of the first decision model;

[0100] The second processing submodule is used to determine the initial value of the model calculation parameter of the first decision model according to the training result of each iteration in the training process of the third decision model.

[0101] Optionally, the training result includes values ​​of model calculation parameters obtained after the current iteration;

[0102] The second processing submodule is further used for:

[0103] Through the formula φ * =argmin φ L(φ), calculate the initial value φ of the model calculation parameter * ;in, θ n Indicates the value of the model calculation parameter after the nth iteration, l n (θ n ) represents the loss function of the model after the nth iteration, and k represents the total number of iterations corresponding to the third decision model.

[0104] The device of this embodiment will obtain the data of the current user and / or scene corresponding to the target device as the first target data when the target device is enabled, so as to train the first decision model based on the first target data to obtain the second decision model. Since the first decision model is a common decision model determined by training the data of users and / or scenes under multiple tasks, and then the second decision model obtained through personalized training of the first target data is a personalized decision model established by learning the characteristics of the current user and / or scene, the target device can call the second decision model to make a decision on the third target data of the current user and / or scene, so that the decision of the target device is more suitable for the current user or scene, thereby improving the applicability of the target device.

[0105] It should be noted that the device is a device to which the above method is applied, and the implementation method of the above method embodiment is applicable to the device and can also achieve the same technical effect.

[0106] like Figure 5 As shown, a communication device according to an embodiment of the present invention includes: a transceiver 510 and a processor 520; wherein,

[0107] The transceiver 510 is used to obtain first target data corresponding to the current user and / or scene of the target device when the target device is enabled;

[0108] The processor 520 is used to train the first decision model according to the first target data to obtain a second decision model; wherein the first decision model is a common decision model determined by training the second target data, and the second target data is data of users and / or scenarios under multiple tasks;

[0109] The processor 520 is further configured to call the second decision model to make a decision on the third target data of the current user and / or scene.

[0110] Optionally, the transceiver is further used to obtain the second target data;

[0111] The processor is further configured to: train a preset decision model according to the second target data to obtain a third decision model corresponding to each task; wherein the preset decision model is a neural network model;

[0112] The processor is further configured to: determine a first decision model according to the third decision model.

[0113] Optionally, the processor is further configured to:

[0114] selecting the mode of the model structure parameter of the third decision model as the value of the model structure parameter of the first decision model;

[0115] According to the training results of each iteration in the training process of the third decision model, the initial values ​​of the model calculation parameters of the first decision model are determined.

[0116] Optionally, the training result includes values ​​of model calculation parameters obtained after the current iteration;

[0117] The processor is further configured to:

[0118] Through the formula φ * =argmin φ L(φ), calculate the initial value φ of the model calculation parameter * ;in, θ n Indicates the value of the model calculation parameter after the nth iteration, l n (θ n ) represents the loss function of the model after the nth iteration, and k represents the total number of iterations corresponding to the third decision model.

[0119] The communication device of this embodiment will obtain the data of the current user and / or scene corresponding to the target device as the first target data when the target device is enabled, so as to train the first decision model based on the first target data to obtain the second decision model. Since the first decision model is a common decision model determined by training the data of users and / or scenes under multiple tasks, and then the second decision model obtained through personalized training of the first target data is a personalized decision model established by learning the characteristics of the current user and / or scene, the target device can call the second decision model to make a decision on the third target data of the current user and / or scene, so that the decision of the target device is more suitable for the current user or scene, thereby improving the applicability of the target device.

[0120] A communication terminal according to another embodiment of the present invention, Figure 6 As shown, it includes a transceiver 610, a processor 600, a memory 620, and a computer program stored in the memory 620 and executable on the processor 600; when the processor 600 executes the computer program, the above-mentioned data processing method applied to the communication terminal is implemented.

[0121] The transceiver 610 is used to receive and send data under the control of the processor 600 .

[0122] Among them, Figure 6 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 600 and various circuits of memory represented by memory 620 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 610 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium. For different user devices, the user interface 630 may also be an interface capable of externally and internally connecting required devices, and the connected devices include but are not limited to a keypad, a display, a speaker, a microphone, a joystick, and the like.

[0123] The processor 600 is responsible for managing the bus architecture and general processing, and the memory 620 can store data used by the processor 600 when performing operations.

[0124] A computer-readable storage medium according to an embodiment of the present invention stores a computer program, and when the computer program is executed by a processor, the steps in the data processing method described above are implemented, and the same technical effect can be achieved. To avoid repetition, it is not described here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0125] It should be further explained that the communication devices described in this specification include but are not limited to smart phones, tablet computers, etc., and many of the functional components described are called modules in order to more particularly emphasize the independence of their implementation methods.

[0126] In the embodiment of the present invention, the module can be implemented with software so that it can be executed by various types of processors. For example, an executable code module of an identification can include one or more physical or logical blocks of computer instructions, for example, it can be constructed as an object, process or function. Nevertheless, the executable code of the identified module does not need to be physically located together, but can include different instructions stored in different positions, and when these instructions are logically combined together, it constitutes a module and realizes the specified purpose of the module.

[0127] In fact, executable code module can be a single instruction or many instructions, and can even be distributed on a plurality of different code segments, distributed among different programs, and distributed across a plurality of memory devices. Similarly, operating data can be identified in the module, and can be implemented and organized in the data structure of any appropriate type according to any appropriate form. The operating data can be collected as a single data set, or can be distributed in different locations (including on different storage devices), and can only be present on a system or network as an electronic signal at least in part.

[0128] When a module can be implemented by software, considering the level of existing hardware technology, a person skilled in the art can build a corresponding hardware circuit to implement the corresponding function of the module that can be implemented by software without considering the cost. The hardware circuit includes a conventional very large scale integration (VLSI) circuit or gate array and existing semiconductors such as logic chips, transistors, or other discrete components. The module can also be implemented by a programmable hardware device, such as a field programmable gate array, a programmable array logic, a programmable logic device, etc.

[0129] The above exemplary embodiments are described with reference to the accompanying drawings, and many different forms and embodiments are feasible without departing from the spirit and teachings of the present invention. Therefore, the present invention should not be constructed as a limitation of the exemplary embodiments proposed herein. More specifically, these exemplary embodiments are provided so that the present invention will be perfect and complete, and the scope of the present invention will be conveyed to those who are familiar with the technology. In these figures, the component sizes and relative sizes may be exaggerated for clarity. The terms used here are only based on the purpose of describing specific exemplary embodiments and are not intended to be limiting. As used herein, unless the text clearly indicates otherwise, the singular forms "one", "an" and "the" are intended to include these multiple forms. It will be further understood that the terms "including" and / or "comprising" when used in this specification indicate the presence of the features, integers, steps, operations, components and / or components, but do not exclude the presence or increase of one or more other features, integers, steps, operations, components, components and / or their groups. Unless otherwise indicated, when stated, a range of values ​​includes the upper and lower limits of that range and any subranges therebetween.

[0130] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A data processing method, characterized in that: include: When the target device is enabled, obtaining first target data of the target device corresponding to the current user and / or scene; The first decision model is trained according to the first target data to obtain a second decision model; wherein the first decision model is a common decision model determined by training according to the second target data, and the second target data is data of users and / or scenarios under multiple tasks; Calling the second decision model to make a decision on third target data of the current user and / or scene, where the third target data is data acquired again corresponding to the current user and / or scene after the first target data is trained individually for the decision model; Before obtaining the first target data of the target device corresponding to the current user and / or scene, the method further includes: acquiring the second target data; Training a preset decision model according to the second target data to obtain a third decision model corresponding to each task; wherein the preset decision model is a neural network model; Determine the first decision model according to the third decision model; Wherein, determining the first decision model according to the third decision model includes: selecting the mode of the model structure parameter of the third decision model as the value of the model structure parameter of the first decision model; According to the training results of each iteration in the training process of the third decision model, the initial values ​​of the model calculation parameters of the first decision model are determined.

2. The method according to claim 1, characterized in that The training results include the values ​​of the model calculation parameters obtained after the current iteration; The step of determining the initial value of the model calculation parameter of the first decision model according to the training result of each iteration in the training process of the third decision model comprises: Through the formula φ * =argmin φ L(φ), calculate the initial value φ of the model calculation parameter * ;in, θ n Indicates the value of the model calculation parameter after the nth iteration, l n (θ n ) represents the loss function of the model after the nth iteration, and k represents the total number of iterations corresponding to the third decision model.

3. A data processing device, characterized in that: include: A first acquisition module, used for acquiring first target data of the target device corresponding to a current user and / or scene when the target device is enabled; A first processing module, configured to train a first decision model according to the first target data to obtain a second decision model; wherein the first decision model is a common decision model determined by training the second target data, and the second target data is data of users and / or scenarios under multiple tasks; A second processing module is used to call the second decision model to make a decision on third target data of the current user and / or scene, where the third target data is data acquired again corresponding to the current user and / or scene after the first target data is obtained after personalized training of the decision model; Wherein, the device further comprises: A second acquisition module, used for acquiring the second target data; a third processing module, configured to train a preset decision model according to the second target data to obtain a third decision model corresponding to each task; wherein the preset decision model is a neural network model; A fourth processing module, used to determine the first decision model according to the third decision model; Wherein, the fourth processing module includes: a first processing submodule, configured to select the mode of the model structure parameter of the third decision model as the value of the model structure parameter of the first decision model; The second processing submodule is used to determine the initial value of the model calculation parameter of the first decision model according to the training result of each iteration in the training process of the third decision model.

4. The device according to claim 3, characterized in that The training results include the values ​​of the model calculation parameters obtained after the current iteration; The second processing submodule is further used for: Through the formula φ * =argmin φ L(φ), calculate the initial value φ of the model calculation parameter * ;in, θ n Indicates the value of the model calculation parameter after the nth iteration, l n (θ n ) represents the loss function of the model after the nth iteration, and k represents the total number of iterations corresponding to the third decision model.

5. A communication device, characterized in that: include: transceiver and processor; wherein, The transceiver is used for acquiring first target data corresponding to a current user and / or scene of the target device when the target device is enabled; The processor is used to train a first decision model according to the first target data to obtain a second decision model; wherein the first decision model is a common decision model determined by training the second target data, and the second target data is data of users and / or scenarios under multiple tasks; The processor is further configured to call the second decision model to make a decision on third target data of the current user and / or scene, where the third target data is data acquired again corresponding to the current user and / or scene after the first target data is obtained after personalized training of the decision model; Wherein, the transceiver is further used to obtain the second target data; The processor is further configured to: train a preset decision model according to the second target data to obtain a third decision model corresponding to each task; wherein the preset decision model is a neural network model; The processor is further configured to: determine a first decision model according to the third decision model; Wherein, the processor is further used for: selecting the mode of the model structure parameter of the third decision model as the value of the model structure parameter of the first decision model; According to the training results of each iteration in the training process of the third decision model, the initial values ​​of the model calculation parameters of the first decision model are determined.

6. A communication device, comprising: A transceiver, a processor, a memory, and a computer program stored in the memory and executable on the processor; wherein the processor implements the data processing method according to any one of claims 1 to 2 when executing the computer program.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps in the data processing method according to any one of claims 1 to 2 are implemented.

Images