CN112784998A

CN112784998A - Data processing method and device and computing equipment

Info

Publication number: CN112784998A
Application number: CN202110104038.3A
Authority: CN
Inventors: 张滋润; 章祯梁; 谭剑; 印卧涛
Original assignee: Alibaba Group Holding Ltd
Current assignee: Alibaba Group Holding Ltd
Priority date: 2021-01-26
Filing date: 2021-01-26
Publication date: 2021-05-11

Abstract

The embodiment of the application provides a data processing method, a data processing device and computing equipment, wherein the method comprises the following steps: detecting a parameter optimization request initiated by a target user; in response to a parameter optimization request, determining a first parameter space composed of a plurality of sub-parameters; selecting N sub-parameters from the plurality of sub-parameters to form a second parameter space; determining a target parameter meeting a parameter use condition based on a plurality of sampling parameters obtained by sampling in a second parameter space; and outputting the target parameters for the target user. The embodiment of the application improves the parameter sampling efficiency.

Description

Data processing method and device and computing equipment

Technical Field

The present application relates to the field of computer technologies, and in particular, to a data processing method and apparatus, and a computing device.

Background

Many parameters are involved in the use process of models such as machine learning models and neural network models, and the selection of the parameters often has important influence on the calculation structure of the models. Thus, a parameter optimization algorithm is employed to select parameters for the model that can be used. The bayesian optimization algorithm is a relatively common parameter adjusting algorithm.

In the prior art, when a Bayesian optimization algorithm is adopted to adjust parameters, firstly, an objective function of a proxy model is established. The agent model has small calculation amount and is used for approximately simulating more complex mathematical models such as a machine learning model, a neural network model and the like. The objective function is a calculation model of the black box, a parameter is input, and a target value is output, wherein the parameter can be a multi-dimensional parameter composed of a plurality of sub-parameters, and a complete parameter is formed when each sub-parameter takes a value. Thereafter, a data set may be initially acquired, the data set including a plurality of parameters and a target value for each parameter. And then carrying out Gaussian modeling by using the data set to obtain a Gaussian model. And then under the constraint of iteration times, continuously acquiring new parameters by using a Gaussian model under the influence of a sampling function. And after the new parameters are collected, inputting the new parameters into the objective function, and calculating to obtain the target values of the new parameters. And updating the data set by using the new parameters and the target values of the new parameters, returning to the step of carrying out Gaussian modeling by using the data set, and continuously executing the step of obtaining the Gaussian model until the iteration times are reached to obtain the target parameters with the highest target values.

However, under the constraint of the number of parameter selection times, when a gaussian model is used to collect new parameters under the influence of a sampling function, a target function is required to be used to calculate a target value every time a new parameter is generated, and for high-dimensional parameters, the calculation complexity is high and the calculation efficiency is low.

Disclosure of Invention

In view of this, embodiments of the present application provide a data processing method and apparatus, and a computing device, so as to solve the technical problem in the prior art that a parameter sampling process is low in computation efficiency, which results in low parameter sampling efficiency.

In a first aspect, an embodiment of the present application provides a data processing method, where the method includes:

detecting a parameter optimization request initiated by a target user;

in response to a parameter optimization request, determining a first parameter space composed of a plurality of sub-parameters;

selecting N sub-parameters from the plurality of sub-parameters to form a second parameter space;

determining a target parameter meeting a parameter use condition based on a plurality of sampling parameters obtained by sampling in the second parameter space;

and outputting the target parameters for the target user.

In a second aspect, an embodiment of the present application provides a data processing method, where the method includes:

responding to a request for calling a data processing interface, and determining a processing resource corresponding to the data processing interface;

executing the following steps by using the processing resource corresponding to the data processing interface:

detecting a parameter optimization request initiated by a target user;

and outputting the target parameters for the target user.

In a third aspect, an embodiment of the present application provides a data processing apparatus, including:

the request detection module is used for detecting a parameter optimization request initiated by a target user;

the request response module is used for responding to the parameter optimization request and determining a first parameter space formed by a plurality of sub-parameters;

the space selection module is used for selecting N sub-parameters from the plurality of sub-parameters to form a second parameter space;

the parameter sampling module is used for determining a target parameter meeting a parameter use condition based on a plurality of sampling parameters obtained by sampling in the second parameter space;

and the parameter output module is used for outputting the target parameters for the target user.

In a fourth aspect, an embodiment of the present application provides a computing device, including: a storage component and a processing component; the storage component is used for storing one or more computer instructions; the one or more computer instructions are invoked by the processing component to perform any of the data processing methods.

According to the embodiment of the application, a parameter optimization request initiated by a target user is detected, and a first parameter space formed by a plurality of sub-parameters can be determined in response to the parameter optimization request. N sub-parameters are selected from the plurality of sub-parameters to form a second parameter space. Determining a target parameter meeting a parameter use condition based on a plurality of sampling parameters obtained by sampling in the second parameter space; and outputting the target parameters for the target user. By reducing the spatial dimension during parameter sampling, the efficiency of parameter sampling can be improved, the time cost of parameter sampling is effectively reduced, the target parameters are quickly obtained and output for a user, and the optimization efficiency of the parameters and the interaction efficiency with the user are improved.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative efforts.

Fig. 1 is a flowchart of an embodiment of a data processing method according to an embodiment of the present application;

fig. 2 is a flowchart of another embodiment of a data processing method according to an embodiment of the present application;

fig. 3 is a flowchart of another embodiment of a data processing method according to an embodiment of the present application;

fig. 4 is a flowchart of another embodiment of a data processing method according to an embodiment of the present application;

fig. 5 is a diagram illustrating an application example of a data processing method according to an embodiment of the present application;

fig. 6 is a schematic structural diagram of an embodiment of a data processing apparatus according to an embodiment of the present application;

fig. 7 is a schematic structural diagram of an embodiment of a computing device according to an embodiment of the present application.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

The terminology used in the embodiments of the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the examples of this application and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise, and "a" and "an" typically include at least two, but do not exclude the presence of at least one.

It should be understood that the term "and/or" as used herein is merely one type of association that describes an associated object, meaning that three relationships may exist, e.g., a and/or B may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship.

The words "if," "if," as used herein may be interpreted as "at … …" or "at … …" or "in response to a determination" or "in response to a recognition," depending on the context. Similarly, the phrases "if determined" or "if identified (a stated condition or event)" may be interpreted as "when determined" or "in response to a determination" or "when identified (a stated condition or event)" or "in response to an identification (a stated condition or event)", depending on the context.

It is also noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a good or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such good or system. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a commodity or system that includes the element.

The technical scheme of the embodiment of the application can be applied to a parameter optimization scene, and the parameters are efficiently acquired by reducing the spatial dimension of parameter acquisition and increasing the parameter acquisition batch, so that the parameter acquisition efficiency is improved.

In a relatively common parameter optimization scenario, it is relatively common to employ a bayesian optimization algorithm for parameter adjustment. The Bayesian parameter adjusting process mainly comprises the steps of firstly establishing an objective function of a proxy model, wherein the objective function is usually a calculation model of a black box, and after a parameter is input to the objective function, the objective function can output a target value. In a bayesian optimization algorithm, the objective function typically outputs a probability value. After that, a data set may be initialized, where the data set may be denoted by D { (x1, y1),.. a.... a.. and (xn, yn) }, where xi is the ith parameter and yi is the target value corresponding to the ith parameter. And then carrying out Gaussian modeling by using the data set to obtain a Gaussian model. And then, under the constraint of the number of parameter selection, acquiring new parameters under the influence of a Gaussian model on a sampling function. And after the new parameters are collected, inputting the new parameters into the objective function, and calculating to obtain the target values of the new parameters. And updates the data set with the new parameter and the target value for the new parameter. And returning to the step of carrying out Gaussian modeling by using the data set to obtain the Gaussian model, continuously acquiring new parameters, and calculating the target value until the iteration times reach the parameter selection times.

However, each time a new parameter is acquired, the parameter needs to be calculated by using the objective function, for the parameter with high dimensionality, the calculation complexity of the new parameter by using the objective function is high, and if a poor parameter is acquired, invalid calculation occurs. Meanwhile, only one new parameter is generated each time, and then the target function is used for parameter calculation, so that the efficiency of parameter acquisition is low, the time cost of parameter selection is increased, and the parameter selection is not facilitated.

In the embodiment of the application, a parameter optimization request initiated by a target user is detected, and a first parameter space formed by a plurality of sub-parameters can be determined in response to the parameter optimization request. N sub-parameters are selected from the plurality of sub-parameters to form a second parameter space. Determining a target parameter meeting a parameter use condition based on a plurality of sampling parameters obtained by sampling in a second parameter space; and outputting the target parameters for the target user. By reducing the spatial dimension during parameter sampling, the efficiency of parameter sampling can be improved, the time cost of parameter sampling is effectively reduced, the target parameters are quickly obtained and output for a user, and the optimization efficiency of the parameters and the interaction efficiency with the user are improved.

The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

As shown in fig. 1, which is a flowchart of an embodiment of a data processing method provided in this embodiment of the present application, the method may include the following steps:

101: and detecting a parameter optimization request initiated by a target user.

102: in response to a parameter optimization request, a first parameter space of a plurality of sub-parameters is determined.

The embodiment of the application can be applied to a computing device, and the computing device can include: the embodiments of the present application do not limit the specific types of the computing devices. In practical applications, if the technical solution of the embodiment of the present application is applied to a computing device having a GPU (Graphics Processing Unit), the GPU may execute the technical solution of the embodiment of the present application.

The target user may interact with the computing device for data or information using a user side, which may include, for example: the embodiments of the present application do not limit the specific types of the user side.

The parameter space may be a value space of a plurality of parameter population distributions. Any value that can be used to randomly characterize a random variable can be referred to as a parameter, and the combination of the possible values of all parameters forms a parameter space. The plurality of sub-parameters form a first parameter space, and the spatial dimension of the first parameter space is the parameter number of the plurality of sub-parameters.

The parameter optimization request is generated when it is determined that a parameter sampling requirement exists. In a common parameter optimization scenario, a parameter optimization request may be generated when it is determined that parameter sampling is required. Taking a bayesian optimization scenario as an example, after an objective function of the proxy model is established, a parameter optimization request can be generated, and parameter sampling is executed by using the technical scheme of the embodiment of the application.

The parameter optimization request can also be initiated when the user has a sampling requirement, and the parameter optimization request can interact with the user to realize effective acquisition of the parameters by initiating the parameter sampling request by the user, so that the acquisition efficiency is improved.

103: n sub-parameters are selected from the plurality of sub-parameters to form a second parameter space.

The second parameter space is a partial space in the first parameter space, and the spatial dimension of the first parameter space is higher than that of the second parameter space. The first parameter space is composed of a plurality of sub-parameters. The second parameter space is composed of N sub-parameters selected from the plurality of sub-parameters. Wherein, N is a positive integer which is larger than 1 and smaller than the parameter number of the plurality of sub-parameters.

Any point in the first parameter space may be a parameter value corresponding to the point in each of the plurality of sub-parameters, and any point in the first parameter space may be referred to as a parameter of the first parameter space. Similarly, any point in the second parameter space may be a parameter value corresponding to the point in each of the selected N sub-parameters, and any point in the second parameter space may be referred to as a parameter of the second parameter space.

Take a plurality of sub-parameters as 3, each of which is ω₁、ω₂And ω₃，ω₁、ω₂And ω₃I.e. a parameter space can be formed, any kind of omega₁、ω₂And ω₃When the values are taken respectively, a parameter can be formed. From ω₁、ω₂And ω₃Two sub-parameters ω are selected₂And ω₃A second parameter space may be constructed. Omega₁、ω₂And ω₃Has a spatial dimension of 3, omega₂And ω₃Has a spatial dimension of 2.

104: and determining a target parameter meeting the parameter use condition based on a plurality of sampling parameters obtained by sampling in the second parameter space.

105: and outputting the target parameters for the target user.

Optionally, the outputting the target parameter for the target user may specifically include directly outputting the target parameter in a form of a web page, a short message, or the like, and may further include generating corresponding output information by using the target parameter and directly outputting the output information to the target user based on a processing target corresponding to the target user, so as to implement indirect output of the target parameter.

As shown in fig. 2, a flowchart of an embodiment of a data processing method provided in the embodiment of the present application may include the following steps:

201: and detecting a parameter optimization request initiated by a target user.

Some steps in the embodiment of the present application are the same as those in the embodiment shown in fig. 1, and are not described herein again.

202: in response to a parameter optimization request, a first parameter space of a plurality of sub-parameters is determined.

203: a second parameter space composed of N sub-parameters is selected from the plurality of sub-parameters.

204: and respectively performing parameter sampling by using a plurality of sampling functions in the second parameter space to obtain sampling parameters respectively corresponding to the plurality of sampling functions so as to obtain a plurality of sampling parameters.

A plurality of sampling functions may be provided to simultaneously sample the parameters in the second parameter space. More common sampling functions may include: the above sampling functions are merely exemplary, and the specific types of the sampling functions are not limited too much in the embodiments of the present application.

The sampling principle for different sampling functions is different, e.g. gain expectation is the new sampling parameter such that the gain expectation is maximal, and the sampling principle for gain probability is the gain probability that maximizes the sampling parameter. The confidence upper/lower limit is such that the sampling function searches for a sample point by maximizing the confidence upper/lower limit.

And simultaneously, a plurality of sampling functions are adopted to respectively perform parameter sampling, namely, sampling is performed by adopting different sampling principles, batch solution of a plurality of batches of models is realized by one-time sampling, a multi-directional sampling analysis result is obtained, and the sampling efficiency is improved, so that the parameter sampling is realized by using a batch processing mode, and the sampling efficiency is improved. In addition, compared with the first parameter space, sampling is performed in the second parameter space, so that the spatial dimension of parameter sampling is reduced, the computational complexity in the sampling process is reduced, and the parameter sampling efficiency is improved.

205: and mapping the plurality of sampling parameters from the second parameter space to the first parameter space respectively to obtain a plurality of target sampling parameters.

The plurality of sampling functions perform parameter sampling in a second parameter space. Mapping the sampling parameters respectively corresponding to the plurality of sampling functions to the first parameter space may include: determining at least one sub-parameter which is not located in the second parameter space in the first parameter space, determining values corresponding to the at least one sub-parameter respectively, and obtaining a plurality of target sampling functions according to the values corresponding to the at least one sub-parameter respectively and by combining the sampling parameters corresponding to the plurality of sampling functions respectively. When the value corresponding to at least one sub-parameter which is not located in the second parameter space in the first parameter space is determined, the values corresponding to the N sub-parameters in the second parameter space of the plurality of sampling parameters are combined to obtain a plurality of target sampling parameters. For example, suppose ω is selected₂And ω₃As the second subspace, when ω is determined in the second subspace₂Is 10, omega₃At 10, assume ω₁Is 5, a parameter point (5,10,10) in the first parameter space can be determined.

And determining a target sampling parameter corresponding to the sampling parameter in the first parameter space based on values corresponding to the sampling parameter in the N sub-parameters respectively and combining values corresponding to at least one sub-parameter which is not located in the second parameter space in the first parameter space, so as to obtain the target sampling parameter of the plurality of sampling parameters in the first parameter space respectively.

206: and determining target parameters meeting the parameter use conditions based on the plurality of target sampling parameters.

207: and outputting the target parameters for the target user.

The target user is the user who initiated the parameter optimization request. Optionally, the user terminal may detect a parameter optimization request initiated by the target user, and send the parameter optimization request to the computing device.

In the embodiment of the application, in response to a parameter optimization request, a first parameter space composed of a plurality of sub-parameters may be determined. N sub-parameters are selected from the plurality of sub-parameters to form a second parameter space. And respectively performing parameter sampling by using a plurality of sampling functions in the second parameter space to obtain sampling parameters respectively corresponding to the plurality of sampling functions, and mapping the sampling parameters respectively corresponding to the plurality of sampling functions to the first parameter space to obtain a plurality of target sampling parameters. Through adopting the space dimension of reducing the parameter, can improve the efficiency of parameter sampling, the sampling effect of different sampling functions is different, through adopting a plurality of sampling functions to carry out the sampling of parameter simultaneously, has realized the batch sampling of multi-angle, increases the parameter quantity of once sampling. Meanwhile, the efficiency and the number of parameter sampling are improved, the time cost of parameter sampling can be effectively reduced, and the optimization efficiency of parameters is improved.

Optionally, after selecting N sub-parameters from the plurality of sub-parameters to form the second parameter space, the method may further include: determining at least one sub-parameter of the plurality of sub-parameters of the first parameter space that is not located in the second parameter space; and determining the value of each of the at least one sub-parameter. Using the plurality of sampling functions to perform parameter sampling in the second parameter space, respectively, and obtaining sampling parameters corresponding to the plurality of sampling parameters may include: and taking the value corresponding to at least one sub-parameter as a sampling reference, and respectively sampling the parameters in a second parameter space by using a plurality of sampling functions. Also omega₁、ω₂And ω₃For example, assume that ω is selected₂And ω₃As a second parameter space, ω can be set₁Is a fixed value, e.g. 5, at ω₁When the value is 5, at ω₂And ω₃To obtain a plurality of target sampling parameters.

As an embodiment, performing parameter sampling in the second parameter space by using a plurality of sampling functions respectively, and obtaining sampling parameters corresponding to the plurality of sampling functions respectively, may include:

a parameter sampling strategy for a second parameter space is determined.

Based on the parameter sampling strategy, the parameter sampling is respectively carried out by using a plurality of sampling functions in the second parameter space, and the sampling parameters respectively corresponding to the plurality of sampling functions are obtained so as to obtain a plurality of sampling parameters.

By setting a parameter sampling strategy in the second parameter space, efficient sampling of parameters can be achieved.

Alternatively, the parameter sampling policy may be preset, and the parameter sampling policy may include, for example, trend information and interval information of parameter sampling.

In order to obtain parameters matching the effect of using the parameters in the second parameter space, in one possible design, the parameter sampling strategy for determining the second parameter space may include:

a plurality of base parameters in a second parameter space are obtained.

And estimating the use results of the plurality of basic parameters to obtain parameter estimation results of the plurality of basic parameters.

And generating a parameter sampling strategy of a second parameter space according to the parameter estimation result.

Wherein the base parameter may be a parameter that already exists before the second parameter spatial sampling. The plurality of basis parameters may be a basis for sampling in the second parameter space. As a possible implementation manner, the multiple basic parameters may be randomly generated in the second parameter space, and then the multiple basic parameters may be mapped to the first parameter space respectively to obtain multiple mapping parameters, and the multiple mapping parameters are input to the objective function respectively to obtain multiple target values through calculation, and the target values corresponding to the multiple mapping parameters respectively constitute the parameter estimation results of the multiple basic parameters. And obtaining a parameter sampling strategy of a second parameter space by sampling and analyzing the target values respectively corresponding to the plurality of mapping parameters.

Optionally, when any basic parameter is mapped to the first parameter space, at least one sub-parameter, which does not belong to the second parameter space, of the multiple sub-parameters in the first parameter space may be determined, and after values of the at least one sub-parameter are respectively determined, in combination with respective corresponding parameter values of the basic parameter in the N sub-parameters, mapping parameters formed by the respective corresponding parameter values of the multiple sub-parameters of the basic parameter in the first parameter space may be determined, so as to obtain the mapping parameter corresponding to the basic parameter.

In order to reduce the computational complexity of determining the target values corresponding to the basic parameters by using the objective function, an interpolation sampling method may be adopted to determine the target values corresponding to the multiple basic parameters. At this time, in some embodiments, obtaining the plurality of base parameters in the second parameter space may include:

a plurality of initial parameters in a first parameter space are randomly generated.

And mapping the plurality of initial parameters from the first parameter space to the second parameter space respectively to obtain a plurality of basic parameters.

Estimating the use results of the plurality of basic parameters, wherein the obtaining of the parameter estimation results of the plurality of basic parameters comprises:

and inputting the plurality of initial parameters into a preset objective function, and calculating to obtain target values corresponding to the plurality of initial parameters respectively.

Determining target values corresponding to the plurality of basic parameters respectively by adopting an estimation algorithm based on the target values corresponding to the plurality of initial parameters respectively;

and generating a parameter estimation result according to the target values respectively corresponding to the plurality of basic parameters.

Since the spatial dimension of the first parameter space is more complex than the spatial dimension of the second parameter space, when the plurality of initial parameters are mapped from the first parameter space to the second parameter space, the parameter dimensions of the plurality of initial parameters are reduced, a phenomenon that basic parameters are overlapped may exist in the second parameter space, and the number of the plurality of initial parameters is greater than the number of the plurality of basic parameters. The plurality of basic parameters may be parameters obtained by mapping the plurality of initial parameters to the second parameter space and performing parameter deduplication processing.

Alternatively, the plurality of initial parameters may be randomly generated in the first parameter space. The plurality of initial parameters are input to the objective function, and target values corresponding to the plurality of initial parameters can be calculated and obtained. And then mapping the plurality of initial parameters to a second parameter space respectively to obtain a plurality of basic parameters.

When the plurality of initial parameters are respectively mapped to the second parameter space, respective values of at least one sub-parameter that is not located in the second parameter space in the first parameter space may be determined, and when the respective values of the at least one sub-parameter are determined, the plurality of initial parameters are respectively mapped to the second parameter space, so as to obtain a plurality of basic parameters in the second parameter space under the condition that the at least one sub-parameter is locked. After the plurality of initial parameters are input to the objective function, target values corresponding to the plurality of initial parameters, respectively, may be obtained. Under the condition that the value of each of at least one sub-parameter which is not located in the second parameter space in the first parameter space is determined, an estimation algorithm can be adopted to estimate and obtain target values corresponding to a plurality of basic parameters respectively.

Alternatively, the estimation algorithm may include an interpolation estimation algorithm by which a target value with higher accuracy can be obtained quickly.

The parameter estimation result may be a mean, a variance, an extreme point, and/or the like between target values respectively corresponding to the plurality of basic parameters. The parameter estimation result can identify the use results possibly generated by a plurality of basic parameters in the actual use process, so that the quantization processing of the use results is realized, and the effective estimation of the plurality of basic parameters is realized.

In the Bayesian parameter optimization process, a Gaussian model of the data set can be calculated when the effect estimation is performed on the parameters. In the embodiment of the application, parameter sampling is performed by reducing the dimension of the parameter space, and when the effect estimation is performed on the parameters, target values corresponding to a plurality of basic parameters and a plurality of basic parameters respectively can be used as data sets in the second parameter space to calculate the gaussian model in the second parameter space. Therefore, as a possible implementation manner, generating the parameter estimation result according to the target values corresponding to the plurality of basic parameters respectively may include:

in a second parameter space, fitting a distribution model of target values corresponding to a plurality of basic parameters respectively to obtain a probability distribution model;

and determining the probability distribution model as a parameter estimation result.

According to the parameter estimation result, the parameter sampling strategy for determining the second parameter space comprises the following steps:

and determining a parameter sampling strategy of the second parameter space according to the probability distribution model.

Alternatively, the probability distribution model may be a gaussian model, a polynomial model, or a bernoulli model, and the specific type of the probability distribution model is not limited too much in the embodiments of the present application.

In the embodiment of the application, the probability distribution model in the second parameter space is calculated and participates in the generation process of the parameter sampling strategy in the second parameter space as the parameter estimation result, so that the quantitative analysis of the parameter use result is realized under a lower space dimensionality, the parameter use result is used as a reference to generate an accurate parameter sampling strategy, and efficient and accurate parameter sampling is realized.

When the generated parameter estimation result is the gaussian model, due to the model characteristics of the gaussian model, the information such as the mean value, the variance and/or the extreme point of the model can be directly obtained, and the information can be directly used as the generation basis of the sampling strategy. In some embodiments, determining the parameter sampling strategy of the second parameter space according to the parameter estimation result may include:

and determining sampling trend information and sampling interval information of the second parameter space according to the mean value, the variance and/or the extreme point corresponding to the probability distribution model.

Optionally, when the probability distribution model is a gaussian model, determining the sampling trend information and the sampling interval information of the second parameter space according to a mean, a variance, and/or an extreme point corresponding to the gaussian model specifically may include: and according to the mean value, the variance and/or the extreme point corresponding to the Gaussian model, taking a plurality of basic parameters as sampling trend information and sampling interval information corresponding to the sampling basic point.

Wherein the sampling trend information may include: when a plurality of basic parameters are taken as sampling basic points for parameter sampling, the acquisition direction and/or the acquisition angle of the acquired target sampling parameters relative to one or more basic parameters are/is determined. For example, assuming that a gaussian model is an increasing function in a certain segment, it can be determined that the value of the model increases when the coordinate point moves towards a first direction, and the value of the model decreases when the coordinate point moves towards a second direction opposite to the first direction. The determination of the sampling trend information can be determined according to specific acquisition requirements or according to acquisition targets of acquisition functions. The collection angle can accurately identify the angle of the target sampling parameter relative to one or more basic parameters, and accurate sampling can be realized. The sampling interval information may specifically be a coordinate distance of the acquired target sampling parameter relative to a certain basic parameter, and when the spatial dimension of the parameter space is complex, the sampling interval information may be a coordinate distance of the acquired target sampling parameter relative to each of the plurality of basic parameters.

When the parameter sampling policy includes sampling trend information and acquisition interval information, the parameter sampling may be directly performed according to the sampling trend information and the acquisition interval information, and therefore, in some embodiments, based on the parameter sampling policy, the parameter sampling is performed in the second parameter space using a plurality of sampling functions respectively, and the obtaining of the sampling parameters corresponding to the plurality of sampling functions respectively may include:

in the second parameter space, according to the sampling trend information and the acquisition interval information, parameter sampling is performed by using a plurality of sampling functions respectively, and sampling parameters corresponding to the plurality of sampling functions respectively are obtained so as to obtain a plurality of sampling parameters.

In practical applications, to implement parameter optimization, based on a plurality of target sampling parameters, determining target parameters satisfying the usage condition may include:

and respectively inputting the target functions to the target sampling parameters, and calculating to obtain target values corresponding to the target sampling parameters.

Selecting a candidate parameter with the highest target value from the plurality of basic parameters and the plurality of target sampling parameters according to the target values corresponding to the plurality of basic parameters and the target values corresponding to the plurality of target sampling parameters respectively;

judging whether the target value corresponding to the candidate parameter meets the preset parameter use condition;

if yes, determining the candidate parameter as a target parameter;

if not, returning to the step of selecting N sub-parameters from the plurality of sub-parameters to form the second parameter space and continuing to execute.

Optionally, candidate parameters are determined from a plurality of base parameters and a plurality of target sampling parameters. The candidate parameter may be a parameter with the highest target value obtained by preliminary selection, and when the candidate parameter is determined, the candidate parameter may be subjected to parameter setting to improve accuracy of parameter optimization.

Optionally, selecting, according to the target values corresponding to the plurality of basic parameters and the target values corresponding to the plurality of target sampling parameters, a candidate parameter with a highest target value from the plurality of basic parameters and the plurality of target sampling parameters may include: and sequencing the target values corresponding to the plurality of basic parameters and the target values corresponding to the plurality of target sampling parameters respectively according to a descending order or a descending order to obtain a plurality of sequenced basic parameters and a plurality of sequenced target sampling parameters, and determining the parameter with the largest target value as a candidate parameter.

The parameter with the largest target value may be used as the candidate parameter, specifically, the estimation criterion of the parameter may be used as the criterion that the target value is the largest, and the calculation method for adjusting the target value may be used to use the target value as the selection basis. For example, assuming that the goal of parameter optimization is to maximize the advertisement click rate, the parameter with the largest target value may be used as the target parameter. Assuming that the objective of the parameter optimization is to minimize the recommended error rate, a calculation method using 1 and the recommended error rate as target values may be used, and a parameter with the largest difference between 1 and the recommended error rate may be used as a candidate parameter.

In one possible design, whether the target value corresponding to the candidate parameter satisfies the preset parameter usage condition may be determined by:

judging whether the target value of the target parameter meets a preset target threshold value or not;

if so, determining that the target parameter meets the parameter use condition;

if not, determining that the target parameter does not meet the parameter use condition.

The target threshold may be preset. If the target value of the target parameter is greater than the target threshold, it may be determined that the target parameter reaches the use condition, and if the target value of the target parameter is not greater than the target value, it may be determined that the target parameter does not reach the use condition. The accuracy of the target parameters can be improved through the setting of the target parameters.

In the parameter optimization process, after the parameter calculation effect is optimized for a certain number of times, the optimization effect with a better result may not be generated. To reduce the invalid parameter optimization calculations, the number of iterations may be used to constrain the parameter optimization process. Therefore, in yet another possible design, after determining the first parameter space composed of the plurality of sub-parameters in response to the parameter optimization request, the method may further include:

recording the iteration times;

whether the target value corresponding to the candidate parameter meets the preset parameter use condition is determined by the following method:

judging whether the iteration times meet a preset iteration threshold value or not;

if so, determining that the target parameter meets the parameter use condition;

For one embodiment, after determining the target parameter, the target parameter may be output for the target user. In practical applications, in order to improve the user-specific use, after outputting the target parameters for the target user, the method may further include:

receiving a parameter adjustment request input by a target user aiming at a target parameter;

responding to the parameter adjustment request, and acquiring parameter adjustment information provided by a target user;

and adjusting the target parameters according to the parameter adjustment information to obtain the adjusted target parameters.

In the embodiment of the application, after the target parameter is output for the target user, if the target value of the target parameter is judged not to meet the expected use condition, the target user can adjust the target parameter so that the target parameter is closer to the expected use condition.

In many application fields, the parameter optimization problem may be directly involved, and in order to improve the parameter optimization efficiency, the technical solution of the embodiment of the present application may be used for parameter optimization.

In the process of allocating the electric power resources and the water resources, resource data of the electric power resources or the water resources and the like may be used as a parameter to be processed to initiate a parameter optimization request, where the parameter to be processed may specifically be a resource amount corresponding to each area, for example, a load capacity of the area in an electric power scene.

As an embodiment, detecting a target user initiated parameter optimization request may include:

and detecting a parameter optimization request initiated by a target user aiming at the to-be-processed parameters of the target resource.

The parameter to be processed comprises a plurality of sub-parameters.

After determining the target parameter satisfying the parameter usage condition based on the plurality of sampling parameters obtained by sampling in the second parameter space, the method may further include:

and generating resource processing information of the target resource according to the values of the target parameter corresponding to the sub-parameters respectively, so as to process the target resource according to the resource processing information.

Optionally, the resource element specifically represented by the parameter to be processed may be determined according to a processing target of the target resource. For example, when the processing target of the target resource is the electrical load capacity set for different regions, the total energy consumption of the grid is minimized. In this case, the power load capacities of different regions may be used as sub-parameters, the power load capacities corresponding to the plurality of regions may form a plurality of sub-parameters in the parameter to be processed, and the processing target may be the power load capacity. Processing information of the target resource can be generated according to the value of each sub-parameter in the to-be-processed parameter in the target parameter, that is, prompt information or setting instructions of power load capacities corresponding to the plurality of regions respectively can be generated according to the value of the to-be-processed parameter in the target parameter. The setting command allows the capacity setting to be performed according to the power load capacities corresponding to the plurality of regions. The prompt information can be output for the user, so that the user can set the capacity of the corresponding region according to the power load capacity of each region prompted in the prompt information.

Parameter optimization problems are also involved in the field of electronic commerce. Taking a relatively common product recommendation as an example, different contents or products can be recommended for a user due to different browsing characteristics of the user, such as consumption habits, attention fields, historical browsing behaviors, and the like. In practical applications, the click through rate may be used as a measure of the accuracy of the recommended content. In order to improve the click rate of the user, the browsing characteristics of the user such as consumption habits, attention fields and the like can be parameterized to generate different sub-parameters, product recommendation is performed according to the proportion formed when the sub-parameters are respectively taken, the click rate of the user on recommended contents or products is tracked, the browsing characteristics of the user are accurately analyzed, the analysis result is fed back to the parameter optimization process, and the products or the contents with higher attention of the user are searched. In the parameter optimization process, the parameter space is reduced, the complexity of the parameter analysis process can be reduced, and the parameter optimization efficiency is improved.

Thus, as an embodiment, detecting a target user-initiated parameter optimization request may include:

and detecting browsing operation initiated by the target user, and generating a parameter optimization request aiming at the browsing parameters of the target user.

The browsing parameter comprises a plurality of sub-parameters.

generating access recommendation information of the target parameter according to values of the target parameter corresponding to the plurality of sub-parameters respectively;

and searching the target product matched with the access recommendation information from the product database to output the target product for the target user.

And setting a plurality of sub-parameters for the browsing parameters of the target user. And generating and testing parameter examples by continuously carrying out browsing parameters to select and obtain target parameters. Of course, in practical application, the browsing parameter may correspond to a plurality of sub-parameters, and parameter values may be sampled respectively, so as to obtain sampling parameters formed by the parameter values corresponding to the plurality of sub-parameters. And selecting a target parameter meeting the parameter use condition from the plurality of sampling parameters to output the target parameter for the target user, and simultaneously generating access recommendation information of the target user according to the target parameter, searching a target product matched with the access recommendation information for the target user, and outputting the target product for the target user.

In some embodiments, the sub-parameters in the browsing parameters may represent product characteristics, the product characteristics may be feature vectors or feature matrices corresponding to different types of products, and the proportions of different product characteristics in the product searching process may be determined according to the values of the sub-parameters of the browsing parameters in the target parameters, so that the recommended characteristics may be obtained by performing weighted summation on the product characteristics according to the values of the sub-parameters in the target parameters, and the recommended characteristics may be used as access recommendation information.

In still other embodiments, the browsing parameter may be a ratio of different types of products, that is, products may be recommended to the user from a plurality of types of products, but the ratio of different types of products is different. Taking makeup and clothing products as main recommended types as examples, parameter optimization is carried out on recommended proportions of the makeup products and the service exhibits, target parameters are finally obtained to be 3:7, the makeup products are 3, the clothing products are 7, and at the moment, according to values of the target browsing parameters in the target parameters, generated access recommendation information can be the makeup products and the clothing products found according to the proportion of 3: 7. Therefore, 3 cosmetic products and 7 clothing products which match the access recommendation information can be searched from the product database and output to the user.

It should be noted that the application schemes shown in the present application are only exemplary, and should not constitute specific limitations on the technical solutions of the embodiments of the present application, and the embodiments of the present application may also be applied to other scenarios involving parameter optimization.

As shown in fig. 3, a flowchart of another embodiment of a data processing method provided in the embodiment of the present application may include the following steps:

301: and detecting a parameter optimization request initiated by a target user.

302: in response to a parameter optimization request, a first parameter space of a plurality of sub-parameters is determined.

303: n sub-parameters are selected from the plurality of sub-parameters to form a second parameter space.

304: randomly generating a plurality of initial parameters in a first parameter space;

305: mapping a plurality of initial parameters from a first parameter space to a second parameter space to obtain a plurality of basic parameters;

306: and inputting the plurality of initial parameters into a preset objective function, and calculating to obtain target values corresponding to the plurality of initial parameters respectively.

307: and determining the target values corresponding to the plurality of basic parameters respectively by adopting an interpolation estimation algorithm based on the target values corresponding to the plurality of initial parameters respectively.

308: and in the second parameter space, fitting the distribution models of the target values respectively corresponding to the plurality of basic parameters to obtain a Gaussian model.

309: and determining a parameter sampling strategy of the second parameter space according to the Gaussian model.

310: and based on a parameter sampling strategy, respectively performing parameter sampling in a second parameter space by using a plurality of sampling functions to obtain sampling parameters respectively corresponding to the plurality of sampling functions.

311: and mapping the sampling parameters respectively corresponding to the sampling functions to a first parameter space to obtain a plurality of target sampling parameters.

In the embodiment of the application, in response to a parameter optimization request, a first parameter space formed by a plurality of sub-parameters can be determined, and then N sub-parameters are selected from a plurality of word parameters to form a second parameter space. The second parameter space has a lower spatial complexity than the first parameter space. Then, a plurality of initial parameters in the first parameter space can be randomly generated, then the plurality of initial parameters are respectively input into a preset target function, target values corresponding to the plurality of initial functions are obtained through calculation, after the plurality of initial parameters are mapped to the second parameter space from the first parameter space, a plurality of basic parameters in the second parameter space can be obtained, and the space complexity of the plurality of basic parameters is lower than that of the parameters in the first parameter space. After the target values corresponding to the plurality of basic parameters are determined based on the target values corresponding to the plurality of initial parameters, respectively, by using an interpolation estimation algorithm, a gaussian model may be generated in the second parameter space according to the target values corresponding to the plurality of basic parameters, respectively. And then determining a parameter sampling strategy of a second parameter space according to the Gaussian model. And parameter sampling can be respectively carried out in the second parameter space by using a plurality of sampling functions through a parameter sampling strategy, so that sampling parameters respectively corresponding to the plurality of sampling functions are obtained. The parameter batch sampling is realized by using a plurality of sampling functions, and the parameter sampling efficiency is obviously improved. The sampling parameters respectively corresponding to the sampling functions are mapped to the first parameter space, a plurality of target sampling parameters can be obtained, the target sampling parameters are simultaneously used for parameter selection, and the parameter sampling efficiency is improved so as to realize efficient and accurate parameter acquisition.

In addition, after the sampling parameters respectively corresponding to the plurality of sampling functions are mapped to the first parameter space and the plurality of target sampling parameters are obtained, the method may further include: respectively inputting the target sampling parameters into a target function, and calculating to obtain target values corresponding to the target sampling parameters; and selecting the target parameter with the highest target value from the plurality of basic parameters and the plurality of target sampling parameters according to the target values corresponding to the plurality of basic parameters and the target values corresponding to the plurality of target sampling parameters respectively.

In a possible design, the technical solution of the embodiment of the present application may be configured in a server to form a service capable of providing parameter optimization to the outside. As shown in fig. 4, a flowchart of another embodiment of a data processing method provided in the embodiment of the present application may include the following steps:

401: responding to a request for calling a data processing interface, and determining a processing resource corresponding to the data processing interface;

402: and detecting a parameter optimization request initiated by a target user.

Alternatively, the parameter optimization request may be obtained by the data processing interface. The user side can provide the acquired parameter optimization request to the processing resource through the data processing interface.

403: in response to a parameter optimization request, a first parameter space of a plurality of sub-parameters is determined.

404: n sub-parameters are selected from the plurality of sub-parameters to form a second parameter space.

405: and determining a target parameter meeting the parameter use condition based on a plurality of sampling parameters obtained by sampling in the second parameter space.

406: and outputting the target parameters for the target user.

Optionally, when the target parameter is output for the target user, the target parameter may be sent to the user side of the target user through the data processing interface. And when the user side acquires the target parameters sent by the data processing interface, the target parameters can be output.

The specific steps executed by the processing resources corresponding to the data processing interface in the embodiment of the present application are the same as the processing steps executed by the data processing method shown in fig. 1, and the specific implementation manner and the technical effect of each technical feature are described in detail in the embodiment shown in fig. 1, and are not described again here.

For ease of understanding, referring to fig. 5, in practical applications, a user device, for example, a terminal device, such as a mobile phone terminal, an Internet of Things (IoT) terminal, or the like, may interact with a user, and the user device may communicate with a server capable of optimizing parameters. Take the example that the user equipment is the computer M1 and the server is the cloud server M2. The computer M1 may detect 501 a user-initiated parameter optimization request and send 502 the parameter optimization request to the cloud server M2.

The cloud server M2 may receive the parameter optimization request, and determine 503 a first parameter space composed of a plurality of sub-parameters in response to the parameter optimization request; selecting 504N sub-parameters from the plurality of sub-parameters constitutes a second parameter space. Thereby determining 505 a target parameter satisfying the parameter usage condition based on a plurality of sampling parameters sampled in the second parameter space; thereafter, a plurality of target parameters are output 506 for the target user. The calculation complexity in the parameter sampling process can be reduced by reducing the parameter sampling dimension, and the parameter sampling efficiency is improved. After obtaining the target parameters, the target parameters may be sent to the handset terminal M1.

After receiving the target parameters, the mobile phone terminal M1 may display 507 the target parameters for the user, and the user obtains and uses the target parameters, thereby implementing automatic sampling of the parameters of the user and improving sampling efficiency. The output mode of the target parameter may include various forms, for example, the output mode may be data, page, information, or message, and the embodiment of the present application does not make much limitation on the specific output mode of the target parameter.

In practical application, the target parameters obtained by the data processing method provided by the embodiment of the application can be directly applied to a model training scene of a machine learning model. For example, when the user-optimized parameter is a hyper-parameter, the obtained target parameter is a target hyper-parameter. The machine learning model can be constructed by utilizing the target hyper-parameters, the machine learning model is trained by utilizing the training data, the model parameters of the machine learning model constructed by utilizing the target hyper-parameters are obtained, the use result of the machine learning model is better, and for example, in the field of face recognition, the recognition accuracy of the face recognition model constructed by utilizing the target hyper-parameters is higher.

The technical scheme of the embodiment of the application can be applied to various fields such as artificial intelligence interaction, data retrieval, content recommendation, click rate prediction, intelligent factories and industrial control, and particularly has stronger applicability in the field of content recommendation, such as content recommendation in the fields of e-commerce, live video broadcast, social contact and online education, and in the field of resource allocation, such as financial product configuration, electric power resource, water resource and supply chain allocation.

For the convenience of understanding, the embodiments of the present application will be described in detail by taking the following problem cases in several practical field scenarios as examples.

(1) The e-commerce field. Application scenarios such as feature search, product recommendation in a live broadcast scenario, content recommendation, and advertisement click rate calculation are the most common in the e-commerce field, and in this embodiment, the content recommendation scenario is taken as an example, and example deployment is performed. The general recommendation process in the recommended scene may be to perform parameterization setting on elements of the selected scene, obtain a plurality of sub-parameters affecting the scene, and identify different features of the scene using the plurality of parameters. The plurality of sub-parameters may form a first parameter space, i.e. the first parameter space is a complete parameter space of the entire scene. The data processing scheme in the embodiment of the application is provided for the target user in a service form, the specific service form comprises a webpage service, a software program service or a program module existing in an SDK or API form, and the target user obtains the parameter optimization service in a form of browsing a parameter optimization interface, service software or a service module and the like. Therefore, the server for providing the parameter optimization service can detect a parameter optimization request initiated by a target user, and determine a first parameter space formed by a plurality of sub-parameters in response to the parameter optimization request. And selecting N sub-parameters from the plurality of sub-parameters to form a second parameter space, and determining a target parameter meeting the parameter use condition based on a plurality of sampling parameters obtained by sampling in the second parameter space, thereby outputting the target parameter for the target user.

Taking a word-clicking recommendation scenario as an example, when a user clicks a search box in an APP (Application), the system will recommend a partial search word (Query word) for the user. The purpose of recommending search terms for the user is to mine the potential purchase demand of the user, increase the use stickiness of the user and improve the total commodity transaction amount. The search system utilizes the following architecture and combines a deep learning Encode-Decode network, namely a target network, to predict the recommendation of search words. Assume that the number of search terms is parameter sampled. In the prior art, target parameters corresponding to the number of search terms are manually set according to manual experience. By using the data processing method, the number of the search words can be automatically optimized according to the parameter optimization process, and the target parameters are selected and obtained. In the parameter optimization process, after the first parameter space formed by the plurality of sub-parameters is determined, N sub-parameters can be selected from the plurality of sub-parameters to form the second parameter space. Based on a plurality of sampling parameters obtained by sampling in the second parameter space, a target parameter satisfying the use condition may be determined. Thereby outputting the target parameters for the target user. After the target user confirms the target parameters, the number of the search terms can be determined according to the values corresponding to the target parameters. By sampling the parameters of the parameter subspace of the number of the search terms, the complexity of parameter calculation can be reduced, and the optimization efficiency of the parameters can be improved.

(2) In the social field, it is also common to recommend content to social users and recommend materials to students. The recommendation in the social field is that a social user browses a social application program, and the display interface of the application program outputs social content which is interesting to the user. Generally, the recommendation of the social field is to constitute sub-parameters by options such as historical browsing behaviors of the user, the field of interest, and user information, and a plurality of sub-parameters may constitute browsing parameters. When the characteristic value of each sub-parameter is determined, characteristic information finally corresponding to the browsing parameter can be generated, and related content or products can be searched based on the characteristic information. In order to find content of interest to a social user, the number and kinds of parameters may be optimized to obtain accurate social user content.

The technical scheme of the embodiment of the application can be configured in the cloud server, the parameter optimization request can be initiated by an operation and maintenance person, and the operation and maintenance person, that is, a target user, can select N sub-parameters from a plurality of sub-parameters to form a second parameter space, wherein the first parameter space corresponds to the plurality of sub-parameters formed by the plurality of parameters and the related information of the user. And sampling parameters in the second parameter space, and selecting target parameters meeting the parameter use conditions according to a plurality of sampling parameters obtained by sampling. By continuously sampling the parameters in the second parameter space, and the space complexity of the second parameter space is lower than that of the first parameter space, the complexity of parameter optimization can be reduced, and the parameter optimization efficiency is improved.

(3) The financial field. Stock index simulation is a very important problem, and based on linear regression, models such as SVM (support vector machines) and LSTM (Long Short-Term Memory) are common to stock index simulation, and before the models are used, a proper model needs to be established. Many hyper-parameters can be involved in the model training process, such as time step in LSTM, feature _ dim, hidden feature, etc., and the contextual features of market, macroscopic factor, microscopic factor, emergency, etc., which can influence the selection of hyper-parameters. By utilizing the technology, the hyper-parameters influencing the construction of the mathematical model and the parameters such as the context characteristics influencing the selection of the hyper-parameters can be used as sub-parameters to form the parameters needing to be optimized. In order to obtain accurate target parameters, sub-parameters corresponding to the above features may be determined, and a first parameter space may be formed by using a plurality of sub-parameters.

By using the technical scheme of the embodiment of the application, the second parameter space formed by N sub-parameters can be selected from the plurality of sub-parameters. By sampling in the second parameter space to obtain a plurality of sampled parameters, a target parameter satisfying the parameter usage condition can be determined. The target parameters are actually screened according to the parameter use conditions, and the complexity of a parameter space is reduced, so that the parameter sampling difficulty can be reduced, the parameter sampling efficiency is improved, and the parameter selection efficiency is improved. After the target parameters are obtained, a machine learning model corresponding to the index simulation problem can be constructed by using the hyper-parameters in the obtained target parameters, model training is carried out, and model parameters are obtained. And then, performing simulation calculation on data such as RMSE (Root Mean square Error) difference values of actual stock indexes of the index simulation problem by using a machine learning model obtained by training, wherein the obtained data can be used for stock market investment guidance.

(4) The resource allocation field takes power resource allocation as an example. The allocation of power resources usually involves many areas, each area can be represented by a corresponding parameter, and these parameters can respectively allocate a certain proportion of resources, and the allocation of resources can affect the information of area economy, population, environment, etc. The technical scheme of the embodiment of the application can be applied to the problems of dynamic pricing of the power market and power economic load distribution. In the following, a detailed description is mainly given of a specific application field of the power system.

In the dynamic pricing problem of the electric power market, the influence of the type of users and the amount of power consumption on the electric power market is more critical. The user type, the electricity consumption and the like can be used as parameters to be optimized, the parameters to be optimized are sampled, and sampling parameters can be obtained. The value of the sub-parameter in the sampling parameter to the benefit/cost in the power system can be used as an optimization target, and the optimization target can also be used as a parameter using condition. In the parameter optimization process, the power resource manager can initiate a parameter optimization request, and the background server can respond to the parameter optimization request to obtain a first parameter space formed by a plurality of sub-parameters, wherein the first parameter space is a parameter space formed by all parameters such as user types and electricity consumption. Then, N sub-parameters are selected from the plurality of sub-parameters to form a second parameter space, parameter sampling can be performed in the second parameter space, a plurality of sampling parameters are obtained, and a target parameter meeting the parameter use condition is determined from the plurality of sampling parameters. After providing the target parameter for the power resource manager, the power resource manager may use the target parameter to perform dynamic pricing of the power resource. In the parameter selection process, since the parameter sampling is carried out in the second parameter space which is simpler than the first parameter space, the complexity of parameter sampling can be reduced, and the parameter sampling efficiency is improved, thereby further improving the parameter optimization efficiency.

In the power economic load distribution problem, a power supplier can simultaneously provide power resources to a plurality of regions, the power load capacity of each region can be used as a sub-parameter, and the lowest total power consumption of a power grid can be used as a parameter using condition. By using the technical scheme of the embodiment of the application, the respective power load capacities of a plurality of regions can be sampled to obtain a sampling parameter. In the sampling process, the spatial complexity of the first parameter space formed by the sub-parameters corresponding to the power load capacities of all regions is high. When N sub-parameters are selected from the plurality of sub-parameters to constitute the second parameter space, the complexity of the parameter space can be reduced. Therefore, when parameter sampling is carried out in the second parameter space, rapid sampling can be realized with lower sampling complexity, and the sampling efficiency is improved, so that the parameter optimization efficiency is improved. And then, continuously adjusting the second parameter space and continuously sampling the parameters to obtain the final target parameters. The distribution strategy of the power load can be obtained by utilizing the target parameters, the selection complexity of the parameters is reduced, and the selection efficiency is improved.

As shown in fig. 6, a schematic structural diagram of an embodiment of a data processing apparatus provided in the present application may include:

the request detection module 601: for detecting a parameter optimization request initiated by a target user.

The request response module 602: the method includes determining a first parameter space composed of a plurality of sub-parameters in response to a parameter optimization request.

The space selection module 603: for selecting N sub-parameters from the plurality of sub-parameters to form a second parameter space.

The parameter sampling module 604: the method comprises the steps of obtaining a plurality of sampling parameters by sampling in a second parameter space, and determining target parameters meeting parameter use conditions.

The parameter output module 605: for outputting target parameters for the target user.

As an embodiment, the parameter sampling module may include:

the parameter sampling unit is used for performing parameter sampling in a second parameter space by using a plurality of sampling functions respectively to obtain sampling parameters corresponding to the sampling functions respectively so as to obtain a plurality of sampling parameters;

a parameter mapping unit: the sampling parameter mapping module is used for mapping the plurality of sampling parameters from the second parameter space to the first parameter space respectively to obtain a plurality of target sampling parameters.

A parameter selection unit: and determining target parameters meeting the parameter use conditions based on the plurality of target sampling parameters.

As an embodiment, the parameter sampling unit may include:

the strategy determining subunit is used for determining a parameter sampling strategy of the second parameter space;

and the parameter sampling subunit is used for performing parameter sampling in the second parameter space by using the plurality of sampling functions respectively based on the parameter sampling strategy to obtain sampling parameters respectively corresponding to the plurality of sampling functions so as to obtain a plurality of sampling parameters.

In one possible design, the policy-determining subunit may include:

a parameter obtaining module, configured to obtain a plurality of basic parameters in a second parameter space;

the result estimation module is used for estimating the use results of the plurality of basic parameters to obtain the parameter estimation results of the plurality of basic parameters;

and the strategy generation module is used for determining a parameter sampling strategy of the second parameter space according to the parameter estimation result.

As a possible implementation manner, the parameter obtaining module may include:

a parameter generation unit for randomly generating a plurality of initial parameters in a first parameter space;

the parameter obtaining unit is used for mapping a plurality of initial parameters from a first parameter space to a second parameter space to obtain a plurality of basic parameters;

the result estimation module may include:

the target calculation unit is used for inputting the plurality of initial parameters into a preset target function and calculating to obtain target values corresponding to the plurality of initial parameters;

the target estimation unit is used for determining target values corresponding to the plurality of basic parameters by adopting an interpolation estimation algorithm based on the target values corresponding to the plurality of initial parameters;

and the parameter generating unit is used for generating a parameter estimation result according to the target values respectively corresponding to the plurality of basic parameters.

In some embodiments, the parameter generation unit may include:

the model fitting subunit is used for fitting the distribution models of the target values respectively corresponding to the plurality of basic parameters in the second parameter space to obtain a probability distribution model;

and the model estimation subunit is used for determining the probability distribution model as a parameter estimation result.

Optionally, the policy generation module includes:

and the strategy generating unit is used for determining a parameter sampling strategy of the second parameter space according to the probability distribution model.

In some embodiments, the parameter sampling strategy comprises: sampling trend information and sampling interval information; the policy generation unit may specifically be configured to: and determining sampling trend information and sampling interval information of the second parameter space according to the mean value, the variance and/or the extreme point corresponding to the probability distribution model.

As another embodiment, the parameter sampling subunit may specifically be configured to:

in the second parameter space, a plurality of sampling functions are respectively used for parameter sampling according to the sampling trend information and the sampling interval information, and sampling parameters respectively corresponding to the sampling functions are obtained so as to obtain a plurality of sampling parameters.

As still another embodiment, the parameter selection unit may include:

the parameter summarizing subunit is used for respectively inputting the target sampling parameters into the target function and calculating to obtain target values corresponding to the target sampling parameters;

the parameter selection subunit is used for selecting a target parameter with the highest target value from the plurality of basic parameters and the plurality of target sampling parameters according to the target values corresponding to the plurality of basic parameters and the target values corresponding to the plurality of target sampling parameters respectively;

and the first judgment unit is used for judging whether the target value corresponding to the candidate parameter meets the preset parameter use condition.

A first result unit, configured to determine the candidate parameter as the target parameter if yes.

And the second result unit is used for returning to the step of selecting N sub-parameters from the plurality of sub-parameters to form a second parameter space and continuing to execute if the second result unit does not exist.

In some embodiments, the first determining unit may specifically include:

the first judgment subunit is used for judging whether the target parameter meets a preset target threshold value; if so, determining that the target parameter meets the parameter use condition; if not, determining that the target parameter does not meet the parameter use condition.

In some embodiments, the apparatus may further comprise:

and the iteration recording module is used for recording the iteration times.

The first determining unit may be further configured to:

the second judgment subunit is used for judging whether the iteration times meet a preset iteration threshold value; if so, determining that the target parameter meets the parameter use condition; if not, determining that the target parameter does not meet the parameter use condition.

As an embodiment, the request detection module may include:

the first detection unit is used for detecting a parameter optimization request initiated by a target user aiming at the to-be-processed parameter of the target resource. The parameter to be processed comprises a plurality of sub-parameters.

The apparatus may further include:

and the resource processing module is used for generating resource processing information of the target resource according to the values of the target parameter corresponding to the plurality of sub-parameters respectively so as to process the target resource according to the resource processing information.

As another embodiment, the request detection module may include:

and the second detection unit is used for detecting the browsing operation initiated by the target user and generating a parameter optimization request aiming at the browsing parameter of the target user.

The browsing parameter comprises a plurality of sub-parameters.

The apparatus may further include:

and the recommendation processing module is used for generating the access recommendation information of the target user according to the values of the target parameter in the plurality of sub-parameters respectively.

And the product searching module is used for searching the target product matched with the access recommendation information from the product database so as to output the target product for the target user.

The data processing apparatus implementing fig. 6 may execute the data processing method in the embodiment shown in fig. 1, and details of the implementation principle and the technical effect are not repeated. The specific manner in which each step is performed by each module or unit in the above embodiments has been described in detail in relation to the embodiments of the method, and will not be elaborated upon here.

In practical applications, the data processing apparatus shown in fig. 7 may be configured as a computing device, and referring to fig. 7, for a schematic structural diagram of an embodiment of a computing device provided in the embodiment of the present application, the device may include: a storage component 701 and a processing component 702; storage component 701 is used to store one or more computer instructions; one or more computer instructions are invoked by the processing component 702 to perform the data processing method illustrated in the embodiments of fig. 1 and the like.

Among other things, the processing component 702 may include one or more processors to execute computer instructions to perform all or some of the steps of the methods described above. Of course, the processing elements may also be implemented as one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, micro-controllers, microprocessors or other electronic components configured to perform the above-described methods.

The storage component 701 is configured to store various types of data to support operations at the terminal. The memory components may be implemented by any type or combination of volatile or non-volatile memory devices such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disks.

Of course, a computing device may also necessarily include other components, such as input/output interfaces, communication components, and so forth. The input/output interface provides an interface between the processing components and peripheral interface modules, which may be output devices, input devices, etc. The communication component is configured to facilitate wired or wireless communication between the computing device and other devices, and the like.

In addition, embodiments of the present application also provide a computer-readable storage medium, where one or more computer instructions may be stored, and when executed, the one or more computer instructions are used to implement any data processing method in the embodiments of the present application.

The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by adding a necessary general hardware platform, and of course, can also be implemented by a combination of hardware and software. With this understanding in mind, the above-described technical solutions and/or portions thereof that contribute to the prior art may be embodied in the form of a computer program product, which may be embodied on one or more computer-usable storage media having computer-usable program code embodied therein (including but not limited to disk storage, CD-ROM, optical storage, etc.).

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.

Claims

1. A data processing method, comprising:

detecting a parameter optimization request initiated by a target user;

and outputting the target parameters for the target user.

2. The method of claim 1, wherein the determining the target parameter satisfying the parameter usage condition based on the plurality of sampled parameters sampled in the second parameter space comprises:

respectively performing parameter sampling by using a plurality of sampling functions in the second parameter space to obtain sampling parameters respectively corresponding to the plurality of sampling functions so as to obtain the plurality of sampling parameters;

mapping the plurality of sampling parameters from the second parameter space to the first parameter space respectively to obtain a plurality of target sampling parameters;

determining the target parameters meeting parameter use conditions based on the plurality of target sampling parameters.

3. The method of claim 2, wherein the parameter sampling in the second parameter space is performed by using a plurality of sampling functions, and obtaining sampling parameters corresponding to the plurality of sampling functions respectively to obtain the plurality of sampling parameters comprises:

determining a parameter sampling strategy of the second parameter space;

based on the parameter sampling strategy, the sampling functions are used for parameter sampling in the second parameter space respectively, and sampling parameters corresponding to the sampling functions respectively are obtained so as to obtain the sampling parameters.

4. The method of claim 3, wherein determining the parameter sampling strategy for the second parameter space comprises:

acquiring a plurality of basic parameters in the second parameter space;

estimating the use results of the plurality of basic parameters to obtain parameter estimation results of the plurality of basic parameters;

and determining a parameter sampling strategy of the second parameter space according to the parameter estimation result.

5. The method of claim 4, wherein the obtaining the plurality of base parameters in the second parameter space comprises:

randomly generating a plurality of initial parameters in the first parameter space;

mapping the plurality of initial parameters from the first parameter space to the second parameter space to obtain the plurality of base parameters;

the estimating the usage results of the plurality of basic parameters, and obtaining the parameter estimation results of the plurality of basic parameters includes:

inputting the initial parameters into a preset objective function, and calculating to obtain target values corresponding to the initial parameters respectively;

and generating the parameter estimation result according to the target values respectively corresponding to the plurality of basic parameters.

6. The method according to claim 5, wherein the generating the parameter estimation result according to the target values respectively corresponding to the plurality of basic parameters comprises:

in the second parameter space, fitting a distribution model of the target values corresponding to the plurality of basic parameters respectively to obtain a probability distribution model;

determining the probability distribution model as the parameter estimation result;

the determining the parameter sampling strategy of the second parameter space according to the parameter estimation result comprises:

7. The method of claim 6, wherein the parameter sampling strategy comprises: sampling trend information and sampling interval information; the determining a parameter sampling strategy of the second parameter space according to the probability distribution model includes:

8. The method of claim 7, wherein the parameter sampling in the second parameter space using the plurality of sampling functions respectively based on the parameter sampling strategy to obtain the sampling parameters corresponding to the plurality of sampling functions respectively comprises:

in the second parameter space, the plurality of sampling functions are respectively used for parameter sampling according to the sampling trend information and the sampling interval information, and sampling parameters respectively corresponding to the plurality of sampling functions are obtained so as to obtain the plurality of sampling parameters.

9. The method of claim 5, wherein the determining the target parameter that satisfies a parameter usage condition based on the plurality of target sampling parameters comprises:

respectively inputting the target sampling parameters into the target function, and calculating to obtain target values corresponding to the target sampling parameters;

judging whether the target value corresponding to the candidate parameter meets the preset parameter use condition or not;

if yes, determining the candidate parameter as the target parameter;

if not, returning to the step of selecting N sub-parameters from the plurality of sub-parameters to form a second parameter space and continuing to execute.

10. The method according to claim 9, wherein whether the target value corresponding to the candidate parameter satisfies a preset parameter use condition is determined by:

if yes, determining that the target parameter meets a parameter use condition;

11. The method of claim 9, wherein after determining the first parameter space of the plurality of sub-parameters in response to the parameter optimization request, further comprising:

recording the iteration times;

if yes, determining that the target parameter meets a parameter use condition;

12. The method of claim 1, wherein after outputting the target parameters for the target user, further comprising:

receiving a parameter adjustment request input by the target user for the target parameter;

responding to the parameter adjustment request, and acquiring parameter adjustment information provided by the target user;

13. The method of claim 1, wherein the detecting a target user initiated parameter optimization request comprises:

detecting a parameter optimization request initiated by the target user aiming at the to-be-processed parameters of the target resources; wherein the parameter to be processed comprises a plurality of sub-parameters;

after determining the target parameter satisfying the parameter usage condition based on the plurality of sampling parameters obtained by sampling in the second parameter space, the method further includes:

14. The method of claim 1, wherein the detecting a target user initiated parameter optimization request comprises:

detecting browsing operation initiated by the target user, and generating a parameter optimization request aiming at the browsing parameter of the target user; wherein the browsing parameter comprises a plurality of sub-parameters;

generating access recommendation information of the target user according to values of the target parameter corresponding to the plurality of sub-parameters respectively;

and searching a target product matched with the access recommendation information from a product database to output the target product for the target user.

15. A data processing method, comprising:

detecting a parameter optimization request initiated by a target user;

and outputting the target parameters for the target user.

16. A data processing apparatus, comprising:

17. A computing device, comprising: a storage component and a processing component; the storage component is used for storing one or more computer instructions; the one or more computer instructions being invoked by the processing component to perform the data processing method of any of claims 1 to 14.