CN114240506B - Multi-task model modeling method, promotion content processing method and related device - Google Patents
Multi-task model modeling method, promotion content processing method and related device Download PDFInfo
- Publication number
- CN114240506B CN114240506B CN202111573367.9A CN202111573367A CN114240506B CN 114240506 B CN114240506 B CN 114240506B CN 202111573367 A CN202111573367 A CN 202111573367A CN 114240506 B CN114240506 B CN 114240506B
- Authority
- CN
- China
- Prior art keywords
- task
- tasks
- model
- positive sample
- mutual information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0277—Online advertisement
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Game Theory and Decision Science (AREA)
- Medical Informatics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
本公开提供了一种多任务模型的建模方法、推广内容处理方法及相关装置，涉及人工智能技术领域。该多任务模型的建模方法包括：获取用于构建任务的特征，根据所述用于构建任务的特征构建初始任务集合；确定所述初始任务集合中不同任务之间的互信息；根据所述不同任务之间的互信息，获得相关任务集合，所述相关任务集合中包括的任务的互信息满足第一预设条件；根据所述相关任务集合中各任务对应的特征，生成多任务模型的样本，利用所述多任务模型的样本进行模型训练，获得所述多任务模型。如此，该方法能够使得多任务模型中多个任务之间的存在较强的相关性，提高多任务模型的学习效率。
The present disclosure provides a modeling method of a multi-task model, a promotion content processing method and related devices, and relates to the field of artificial intelligence technology. The modeling method of the multi-task model includes: obtaining features for constructing tasks, constructing an initial task set according to the features for constructing tasks; determining the mutual information between different tasks in the initial task set; obtaining a related task set according to the mutual information between the different tasks, and the mutual information of the tasks included in the related task set meets a first preset condition; generating samples of the multi-task model according to the features corresponding to each task in the related task set, and performing model training using the samples of the multi-task model to obtain the multi-task model. In this way, the method can make the multiple tasks in the multi-task model have a strong correlation, and improve the learning efficiency of the multi-task model.
Description
Technical Field
The present disclosure relates to the field of artificial intelligence (ARTIFICIAL INTELLIGENCE, AI) technology, and in particular to a method of modeling a multitasking model, a method of promotional content processing, an apparatus, a device, a computer-readable storage medium, and a computer program product.
Background
With the development of computer technology, especially artificial intelligence technology, the application of artificial intelligence technology is becoming wider and wider. For example, in a scenario where promotional content (e.g., advertisements) is pushed, the conversion rate of the promotional content may be predicted based on artificial intelligence techniques, and the promotional content may be pushed to the user based on the conversion rate.
In order to improve the generalization capability of the model and improve the accuracy of the predicted conversion rate, a multitasking model is generally constructed by adopting a multitasking learning (muti TASK LEARNING, MTL) mode. At present, a plurality of tasks are selected based on artificial subjective experience, and multi-task learning is performed, so that learned features are shared among different tasks, and further learning efficiency and generalization capability of a multi-task model are improved.
However, the multiple tasks selected by relying on subjective experience may have poor correlation, which may cause negative effects among the multiple tasks, and reduce learning efficiency of the multi-task model.
Disclosure of Invention
The invention aims to provide a modeling method, a popularization content processing method, a device, equipment, a computer readable storage medium and a computer program product of a multi-task model, which can improve the learning efficiency of the multi-task model.
In a first aspect, the present disclosure provides a method for modeling a multitasking model, including:
The method comprises the steps of obtaining characteristics for constructing a task, and constructing an initial task set according to the characteristics for constructing the task, wherein the initial task set comprises at least two of conversion rate of promotion content, play time of the promotion content, presentation type of the promotion content and information of promotion objects in the promotion content;
determining mutual information among different tasks in the initial task set;
Acquiring a related task set according to the mutual information of different tasks, wherein the mutual information of the tasks included in the related task set meets a first preset condition;
And generating a sample of a multi-task model according to the characteristics corresponding to each task in the related task set, and performing model training by using the sample of the multi-task model to obtain the multi-task model.
In a second aspect, the present disclosure provides a promotional content processing method, including:
acquiring the attribute of the behavior of the user on the promotion content;
Obtaining an inference result of the multitask model according to the attribute of the behavior of the user on the promotion content and the multitask model, wherein the multitask model is obtained based on a sample of the multitask model generated by the characteristics corresponding to each task in a related task set, the related task set is obtained based on mutual information among different tasks, the different tasks are tasks in an initial task set constructed by the characteristics used for constructing the tasks, and the inference result comprises a plurality of types of conversion rate of the promotion content, play time of the promotion content, presentation type of the promotion content or information of promotion objects in the promotion content;
and adjusting the popularization strategy for the popularization content according to the reasoning result.
In a third aspect, the present disclosure provides a modeling apparatus for a multitasking model, comprising:
the acquisition module is used for acquiring characteristics for constructing a task and constructing an initial task set according to the characteristics for constructing the task, wherein the initial task set comprises at least two of conversion rate of promotion content, play time of the promotion content, presentation type of the promotion content and information of promotion objects in the promotion content;
The mutual information determining module is used for determining mutual information among different tasks in the initial task set;
the related task determining module is used for obtaining a related task set according to the mutual information among different tasks, wherein the mutual information of the tasks included in the related task set meets a first preset condition;
And the training module is used for generating samples of the multi-task model according to the characteristics corresponding to each task in the related task set, and carrying out model training by using the samples of the multi-task model to obtain the multi-task model.
In a fourth aspect, the present disclosure provides a promotion content processing apparatus, including:
the acquisition module is used for acquiring the attribute of the behavior of the user on the promotion content;
The reasoning module is used for obtaining a reasoning result of the multitasking model according to the attribute of the behavior of the user on the popularization content and the multitasking model; the multi-task model is obtained based on a sample of the multi-task model generated by the characteristics corresponding to each task in a related task set, the related task set is obtained based on mutual information among different tasks, and the different tasks are tasks in an initial task set constructed by the characteristics used for constructing the tasks;
and the processing module is used for adjusting the popularization strategy for the popularization content according to the reasoning result.
In a fifth aspect, the present disclosure provides a computer readable medium having stored thereon a computer program which when executed by a processing device performs the steps of the method of any of the first or second aspects of the present disclosure.
In a sixth aspect, the present disclosure provides an electronic device, comprising:
a storage device having a computer program stored thereon;
Processing means for executing the computer program in the storage means to implement the steps of the method of any one of the first or second aspects of the present disclosure.
In a seventh aspect, the present disclosure provides a computer program product comprising instructions which, when run on a device, cause the device to perform the method of any one of the implementations of the first or second aspects described above.
From the above technical solution, the present disclosure has the following advantages:
The method comprises the steps of firstly obtaining characteristics for constructing tasks, constructing an initial task set based on the characteristics for constructing the tasks, then determining mutual information among different tasks in the initial task set, screening the initial task set based on the mutual information to obtain a relevant task set with stronger correlation, and then generating a sample of the multi-task model based on the characteristics corresponding to the tasks in the relevant task set so as to train the multi-task model. Compared with a plurality of tasks which are selected by relying on subjective experiences, the method and the device have the advantage that the correlation between the tasks in the related task set obtained by screening the initial task set is stronger. In this way, the learning efficiency of the multi-task model can be improved in the multi-task learning process.
Additional features and advantages of the present disclosure will be set forth in the detailed description which follows.
Drawings
In order to more clearly illustrate the technical method of the embodiments of the present disclosure, the drawings that are required to be used in the embodiments will be briefly described below.
FIG. 1 is a flow chart of a method for modeling a multitasking model provided by an embodiment of the present disclosure;
FIG. 2 is a schematic diagram of acquiring a set of related tasks according to an embodiment of the present disclosure;
FIG. 3 is a schematic diagram of a multi-tasking model provided by embodiments of the present disclosure;
FIG. 4 is a flow chart of a method of modeling a further multitasking model provided by an embodiment of the present disclosure;
fig. 5 is a flowchart of a method for processing promotion content according to an embodiment of the present disclosure;
FIG. 6 is a schematic diagram of a modeling apparatus for a multitasking model provided in an embodiment of the disclosure;
fig. 7 is a schematic diagram of a promotion content processing apparatus according to an embodiment of the present disclosure;
Fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the disclosure.
Detailed Description
The terms "first," "second," and the like in the presently disclosed embodiments are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature.
Some technical terms related to the embodiments of the present disclosure will be first described.
A multitasking model refers to a model built based on the manner of multitasking learning. The core of multi-task learning is that multiple tasks train in parallel and share learned features with each other. In a promotional content push scenario, it is also often desirable to predict multiple items of content, each of which may be abstracted into a task such as predicting conversion rate, predicting potential users, predicting whether users click on promotional content, etc. Based on this, a multitasking model may be constructed based on the manner of multitasking learning to predict multiple items of content.
In the multi-task learning process, when the correlation among a plurality of tasks is strong, the plurality of tasks share learned characteristics, so that the learning efficiency of a multi-task model can be improved, when the correlation among the plurality of tasks is weak, the plurality of tasks share learned characteristics, and negative migration phenomenon can occur among the plurality of tasks, namely, the plurality of tasks mutually generate negative influence, so that the prediction accuracy of each task is reduced, and the learning efficiency of the multi-task model is reduced.
Currently, subjective experience is mainly relied on to select multiple tasks and construct a multi-task model. The multiple tasks selected based on subjective experience have poor correlation, so that the learning efficiency of the multi-task model is reduced.
In view of this, embodiments of the present disclosure provide a method of modeling a multitasking model, which may be performed by an electronic device. The electronic device may be a server. The server may be a cloud server, for example, a central server in a central cloud computing cluster, or an edge server in an edge cloud computing cluster. Of course, the server may also be a server in a local data center. The local data center refers to a data center directly controlled by a user.
The modeling method of the multi-task model comprises the steps that electronic equipment obtains characteristics for constructing tasks, an initial task set is constructed according to the characteristics for constructing the tasks, then mutual information among different tasks in the initial task set is determined, a relevant task set is obtained based on the mutual information among the different tasks, the mutual information in the relevant task set meets a first preset condition, a sample of the multi-task model is generated according to the characteristics corresponding to each task in the relevant task set, and model training is conducted by the sample of the multi-task model to obtain the multi-task model.
Therefore, in the modeling method of the multi-task model provided by the embodiment of the disclosure, the electronic device filters the initial task set, so that the correlation between tasks in the obtained related task set is stronger, and compared with a plurality of tasks which are selected by simply relying on subjective experience, the correlation between a plurality of tasks corresponding to the multi-task model is improved. In this way, the learning efficiency of the multi-task model can be improved in the multi-task learning process.
The multi-task model obtained by the modeling method of the multi-task model provided by the embodiment of the disclosure can be applied to various scenes. For example, in a promotional content push scenario, a multitasking model may be used to predict the conversion rate of promotional content, the duration of the playback of promotional content (the duration of time that the promotional content has elapsed since it began to be played until the user closed the promotional content). The electronic equipment can input the attribute of the behavior of the user on the popularization content into the multi-task model so as to obtain the reasoning result. The electronic device may then adjust the promotion policy for the promotion content based on the inference results. For example, when the conversion rate of the promotion content is greater than or equal to a conversion rate threshold value and when the playing time of the promotion content is greater than or equal to a time length threshold value, the promotion times of the promotion content are increased, and when the conversion rate of the promotion content is less than the conversion rate threshold value and when the playing time of the promotion content is less than the time length threshold value, the promotion times of the promotion content are reduced.
In order to make the technical solution of the present disclosure clearer and easier to understand, the modeling method of the multitasking model provided by the embodiments of the present disclosure is described below in terms of electronic devices. As shown in fig. 1, the method is a flowchart of a modeling method of a multitasking model according to an embodiment of the disclosure, and the method includes:
s101, the electronic equipment acquires the characteristics for constructing the task, and constructs an initial task set according to the characteristics for constructing the task.
The electronic device may obtain initial data that includes a plurality of features. For example, in the promotion content push scenario, the initial data may include the conversion rate of the promotion content, the play duration of the promotion content, the presentation type of the promotion content, the information of the promotion object in the promotion content, the generation time of the promotion content, and other features.
The electronic device may obtain features for the build task based on the initial data. In some examples, the electronic device may receive a plurality of features configured by a developer as features for the build task. For example, the electronic device may present a plurality of candidate features, such as the features included in the initial data described above, to a user via a display device (e.g., a display screen), and then take the feature selected by the developer as the feature for the build task according to a selection operation of the plurality of candidate features by the developer. Next, the electronic device builds an initial set of tasks based on the features used to build the tasks.
S102, the electronic equipment determines mutual information among different tasks in the initial task set.
Mutual information (mutual information, MI) refers to the degree of association between two random variables, i.e., the degree of attenuation of the uncertainty of one random variable given the other. For example, when the mutual information value is 0 (minimum value), it indicates that there is no relation to determining another random variable given one random variable, and when the mutual information value is entropy (maximum value) of the random variable, it indicates that the uncertainty of another random variable can be completely eliminated given one random variable.
As shown in fig. 2, the electronic device may calculate mutual information between different tasks in the initial task set, respectively. For example, the electronic device may calculate mutual information between different tasks based on characteristics of the tasks in the initial task set. Specifically, the electronic device may calculate mutual information between different tasks in the initial task set based on the following formula:
Wherein I (X; Y) is mutual information between task X and task Y in the initial task set, X is a feature corresponding to task X, Y is a feature corresponding to task Y, p (X, Y) is probability of simultaneous occurrence of task X and task Y, p (X) is probability of occurrence of task X, and p (Y) is probability of occurrence of task Y.
It should be noted that, mutual information between different tasks in the initial task set is non-negative, i.e., I (X; Y) is not less than 0, and I (Y; X) =i (X; Y) each represents mutual information between task X and task Y in the initial task set.
S103, the electronic equipment obtains a related task set according to the mutual information among different tasks, wherein the mutual information of the tasks included in the related task set meets a first preset condition.
As described above, the mutual information may represent uncertainty of one random variable in the plurality of random variables to another random variable, based on which the electronic device may filter the initial task set based on mutual information between different tasks in the initial task set, thereby obtaining a related task set, where the mutual information of the tasks included in the related task set meets a first preset condition. For example, the first preset condition may be that mutual information between different tasks is greater than or equal to a mutual information threshold.
In some examples, the electronic device may receive a developer configured mutual information threshold to determine a set of related tasks based on the configured mutual information threshold, such that the developer adjusts tasks in the set of related tasks within a certain range.
The embodiment of the disclosure is not particularly limited to the mutual information threshold, and as described above, the mutual information threshold may be a numerical value configured by a developer, or may be a default value. For example, the mutual information threshold may be 0.1, 0.2, etc.
S104, the electronic equipment generates a sample of the multi-task model according to the corresponding characteristics of each task in the related task set, and performs model training by using the sample of the multi-task model to obtain the multi-task model.
After determining the set of related tasks, the electronic device may construct a multi-task model based on the tasks in the set of related tasks. Because the mutual information among the tasks in the related task set meets the first preset condition, the correlation among the tasks in the related task set is strong. Therefore, after the multi-task model is built based on the task with strong correlation, the learning efficiency of the multi-task model is high.
The electronic device may generate a sample of the multitasking model based on features corresponding to the individual tasks in the set of related tasks. The samples of the multitasking model include feature vectors of the samples and labels of the feature vectors of the samples. As described above, the initial data includes a plurality of features, the features corresponding to the tasks in the relevant task set are part of the features in the plurality of features of the initial data, based on which the electronic device may use the features corresponding to the tasks in the relevant task set as labels of feature vectors of the samples, and remove the features corresponding to the tasks in the relevant task set from the initial data, and then use the obtained initial data as the feature vectors of the samples.
For example, the initial data includes feature 1, feature 2, feature 3, feature 4, and feature 5, and task 1 corresponds to feature 1 and task 2 corresponds to feature 2 in the related task set. The electronic device may take feature 1 as a label for monitoring the feature vector of the sample of task 1, take feature 2 as a label for monitoring the feature vector of the sample of task 2, remove feature 1 and feature 2 from the initial data to obtain initial data including feature 3, feature 4 and feature 5, and take the initial data including feature 3, feature 4 and feature 5 as the feature vector of the sample of the multitasking model.
Then, the electronic device may perform model training using the samples of the multitasking model to obtain the multitasking model. As shown in fig. 3, the figure is a schematic diagram of a multitasking model provided in an embodiment of the disclosure. The multitasking model includes a task exclusive network 310 and an output network 330, and a shared network 320, each corresponding to a plurality of tasks in a set of related tasks. The task exclusive network 310 may be a deep neural network (deep neural networks, DNN), a convolutional neural network (convolutional neural network, CNN), or a self-attention network, and the shared network 320 may be a deep neural network, a convolutional neural network, or a self-attention network.
It should be noted that the present application is not particularly limited to the types of the task independent network 310 and the shared neural network 320. In the image scene, in order to build invariance of content information, a convolutional neural network can be used for building a shared network 320 and a task exclusive network 310, in the popularization content scene, in order to meet the requirement of feature crossing, a deep neural network can be used for building the shared network 320 and the task exclusive network 310, in the text scene, in order to meet the requirement of parallel processing of the time-series information, a self-attention network can be used for building the shared network 320 and the task exclusive network 310. Output network 330 may use sigmoid, relu, etc. activation functions to obtain outputs (e.g., classification values, regression values) corresponding to the respective tasks.
In some examples, the electronic device may input feature vectors of samples of the multitasking model into shared network 320, resulting in shared components, which are then input into the task exclusive networks of each task in the relevant task set, resulting in outputs of each task exclusive network. The electronic device may then determine a loss value based on the tag value of the feature vector of the sample of the multi-tasking model and the output of the task-exclusive network, update the weight of the task-exclusive network based on the loss value, and perform model training to obtain the multi-tasking model.
Based on the above description, the embodiment of the disclosure provides a modeling method of a multi-task model, which has stronger correlation between tasks in a related task set obtained by filtering an initial task set by electronic equipment compared with a plurality of tasks selected by simply relying on subjective experience. In this way, the learning efficiency of the multi-task model can be improved in the multi-task learning process. Further, the business requirements under various scenes are met, and more accurate prediction of various contents is achieved.
The embodiment of the disclosure also provides a modeling method of the multi-task model, which is based on the embodiment shown in fig. 2, and further groups and aggregates the tasks in the related task set, so as to balance the number of positive samples in a plurality of task groups in the aggregated related task set. As shown in fig. 4, the modeling method of the multitasking model further includes:
s401, the electronic equipment aggregates tasks meeting a second preset condition in the related task set.
The second preset condition may be a plurality of tasks in the set of related tasks having task relevance greater than a relevance threshold. The electronic device may aggregate the tasks in the related task set based on the second preset condition, where the aggregated related task set includes a plurality of task groups.
With continued reference to fig. 3, the electronic device may obtain correlations between outputs of the task exclusive network 310 corresponding to a plurality of tasks in the related task set, and then aggregate the plurality of tasks in the related task set based on the correlations between the outputs of the task exclusive network 310. Specifically, the electronic device may calculate the correlation between the outputs of the task exclusive network 310 corresponding to the plurality of tasks in the related task set based on the following formula:
Wherein r mn is the correlation between task m and task n in the related task set, O im is the output of sample i on the task exclusive network 310 corresponding to task m, For the average value of the outputs of the task exclusive network 310 corresponding to task m to all samples, O in for the output of sample i at the task exclusive network 310 corresponding to task n,For the average value of the outputs of the task exclusive network 310 for all samples corresponding to task n, K is the total number of samples.
In some examples, the electronic device may compare the correlation r mn between the task m and the task n with a correlation threshold, and aggregate the task m and the task n into one group when r mn is greater than or equal to the correlation threshold, so that the electronic device may obtain that the aggregate correlated task set includes a plurality of task groups.
S402, the electronic equipment determines a plurality of target tasks according to the positive sample number of the tasks in the task groups aiming at each task group.
Wherein the sum of the positive sample numbers of the plurality of target tasks is greater than a positive sample number threshold that is the minimum of the sum of the positive sample numbers in each task group. For example, the aggregated related task set includes a plurality of task groups, and the task groups include a task group G1 and a task group G2, where the task group G1 includes a task T1, a task T2, and a task T3, the number of positive samples of the task T1 is 100, the number of positive samples of the task T2 is 200, and the number of positive samples of the task T3 is 300, and the task group G2 includes a task T4, a task T5, and a task T6, the number of positive samples of the task T4 is 200, the number of positive samples of the task T5 is 400, and the number of positive samples of the task T6 is 500. It can be seen that the sum of the positive sample numbers of task group G1 is 600 and the sum of the positive sample numbers of task group G2 is 1100. The electronic device may take the sum of the positive sample numbers 600 of task group G1 as the positive sample number threshold.
In some examples, the electronic device may order tasks T4, T5, and T6 in task group G2, such as adding up one by one in order of the number of positive samples from small to large until the sum of the number of positive samples for the tasks in task group G2 is greater than or equal to positive sample number threshold 600 described above. The electronic device may determine that the plurality of tasks in the task group are a plurality of target tasks when the sum of the positive sample numbers of the tasks in the task group is greater than or equal to a positive sample number threshold. In this way, after summing the positive sample number of the task T4 and the positive sample number of the task T5, the electronic device may take the task T4 and the task T5 as multiple target tasks in the task group, where the sum result is equal to the positive sample number threshold.
In other examples, the electronic device may randomly select the positive sample number of the plurality of tasks at task group G2 to sum until the positive sample number of the plurality of tasks is greater than or equal to the positive sample number threshold, and then determine the randomly selected plurality of tasks as the plurality of target tasks. Continuing the above example, the electronic device may sum the positive sample numbers of the task T5 and the task T6 in the task group G2, where the sum result is greater than the positive sample number threshold, so that the electronic device may determine the task T5 and the task T6 as multiple target tasks.
Note that the sum of the positive sample numbers of the task T4 and the task T5 in the task group G2 being greater than or equal to the positive sample number threshold means that when the positive sample number of either the task T4 or the task T5 in the task group G2 is subtracted, the sum of the positive sample numbers of the remaining tasks is made smaller than the positive sample number threshold.
After the electronic device determines a plurality of target tasks of the task group, tasks except the target tasks in the task group can be removed. Taking the example of a plurality of target tasks in the task group G2 as the task T4 and the task T5, the electronic device may remove the task T6 from the task group G2.
In some embodiments, the electronic device may generate samples of the multitasking model based on features corresponding to the multiple target tasks in the task group. Taking the task group G2 including the target task as the task T4 and the task T5 as an example, the label of the feature vector of the positive sample corresponding to the task T4 is taken as the label of the task group G2, and the label of the feature vector of the positive sample corresponding to the task T5 is taken as the label of the task group G2 for the task group G2. In this way, in the present embodiment, the electronic device can perform balanced configuration on the number of positive samples included in the plurality of task groups, so as to reduce the situation that the learning efficiency of the multi-task model is reduced due to a large difference in the number of positive samples.
Based on the description, the modeling method of the multi-task model realizes the aggregation of high-correlation tasks by analyzing statistical information of large-scale data containing a plurality of characteristics and determining a correlation task set based on correlation. The high-correlation tasks are utilized to train the multi-task model, so that negative migration can be reduced, and the learning efficiency of the multi-task model is improved.
And performing screening and combination of each grouping task by taking the positive sample number corresponding to the task grouping with the minimum positive sample number as a positive sample number threshold value, thereby realizing the balance of the positive sample numbers of different tasks. The number of positive samples among different tasks after grouping and aggregation can be completely equal in proportion, so that the multi-task model finishes the multi-task learning process without deviation.
The embodiment of the disclosure also provides a promotion content processing method, as shown in fig. 5, which includes:
S501, the electronic equipment acquires the attribute of the behavior of the user on the promotion content.
Attributes of the user's behavior on the promotional content may include the length of time the user viewed the promotional content, whether the user clicked on the promotional content, the type of presentation of the promotional content clicked on by the user (e.g., video type, picture type, etc.), whether the user was converted by the promotional content, and so forth.
It should be noted that, the electronic device needs to obtain the authorization of the user in advance, and after obtaining the authorization of the user to use the corresponding data (such as the attribute of the user to the behavior of the promotion content), the electronic device can obtain the data such as the attribute of the user to the behavior of the promotion content.
S502, the electronic equipment obtains an inference result of the multitask model according to the attribute of the behavior of the user on the promotion content and the multitask model.
The multi-task model is obtained based on a sample of the multi-task model generated by the characteristics corresponding to each task in the related task set, the related task set is obtained based on mutual information among different tasks, the different tasks are tasks in an initial task set constructed by the characteristics used for constructing the tasks, and the reasoning result comprises a plurality of types of conversion rate of promotion contents, play duration of the promotion contents, presentation types of the promotion contents or information of promotion objects in the promotion contents. The process of training the multi-task model can be found in the above embodiments, and will not be described here.
And S503, the electronic equipment adjusts the popularization strategy for the popularization content according to the reasoning result.
In some examples, the electronic device increases the number of popularizes of the promotional content when the inference result indicates that the conversion rate of the promotional content is greater than the conversion rate threshold and the playing time of the promotional content is greater than the duration threshold, and decreases the number of popularizes of the promotional content when the conversion rate of the promotional content is less than the conversion rate threshold and the playing time of the promotional content is less than the duration threshold. In other examples, the electronic device increases the number of popularizes of the promotional content when the presentation type of the promotional content is a preset type (e.g., video type) and the information of the promotional object in the promotional content is preset information (e.g., indicating that the promotional object is a game, virtual object, or physical object). Therefore, invalid release of popularization content is reduced, and resource waste is reduced.
FIG. 6 is a schematic diagram of a modeling apparatus of a multi-tasking model according to an exemplary disclosed embodiment, as shown in FIG. 6, the modeling apparatus 600 of the multi-tasking model includes:
The acquisition module 601 is configured to acquire a feature for constructing a task, and construct an initial task set according to the feature for constructing a task, where the initial task set includes at least two of a conversion rate of a promotion content, a play duration of the promotion content, a presentation type of the promotion content, and information of a promotion object in the promotion content;
A mutual information determining module 602, configured to determine mutual information between different tasks in the initial task set;
The related task determining module 603 is configured to obtain a related task set according to mutual information between the different tasks, where the mutual information of the tasks included in the related task set meets a first preset condition;
and the training module 604 is configured to generate a sample of a multitasking model according to the features corresponding to each task in the related task set, and perform model training by using the sample of the multitasking model to obtain the multitasking model.
Optionally, the related task determining module 603 is further configured to aggregate tasks in the related task set that meet a second preset condition.
Optionally, the related task determining module 603 is specifically configured to determine a correlation between outputs of task exclusive networks corresponding to a plurality of tasks in the related task set, and aggregate the plurality of tasks in the related task set according to the correlation.
Optionally, the related task determining module 603 is further configured to determine, for each task group, a plurality of target tasks according to the positive sample number of the tasks in the task group, where a sum of the positive sample numbers of the plurality of target tasks is greater than a positive sample number threshold, and the positive sample number threshold is a minimum value of the sum of the positive sample numbers in each task group;
the training module 604 is specifically configured to generate a sample of the multitasking model according to the feature corresponding to the target task in the task group.
Optionally, the related task determining module 603 is specifically configured to add up the positive sample numbers of the tasks in the task group from small to large until the sum of the positive sample numbers of the tasks in the task group is greater than or equal to the positive sample number threshold value, and determine that the plurality of tasks in the task group are a plurality of target tasks when the sum of the positive sample numbers of the tasks in the task group is greater than or equal to the positive sample number threshold value
Optionally, the training module 604 is specifically configured to input the feature vector of the sample of the multitasking model into a shared network to obtain a shared component, input the shared component into a task exclusive network of each task in the related task set to obtain an output corresponding to the task exclusive network of each task, and train the multitasking model according to the tag value of the feature vector of the sample of the multitasking model and the output of the task exclusive network.
Optionally, the task exclusive network comprises a deep neural network, a convolutional neural network or a self-attention network, and the shared network comprises a deep neural network, a convolutional neural network or a self-attention network.
Fig. 7 is a schematic diagram of a promotional content processing device according to an exemplary disclosed embodiment, as shown in fig. 7, the promotional content processing device 700 includes:
An obtaining module 701, configured to obtain an attribute of a behavior of a user on promotion content;
The reasoning module 702 is used for obtaining a reasoning result of the multitask model according to the attribute of the behavior of the user on the promotion content and the multitask model, wherein the multitask model is obtained based on a sample of the multitask model generated by the characteristics corresponding to each task in a related task set, the related task set is obtained based on mutual information among different tasks, and the different tasks are tasks in an initial task set constructed by the characteristics used for constructing the tasks;
and the processing module 703 is used for adjusting the promotion policy of the promotion content according to the reasoning result.
The functions of the above modules are described in detail in the method steps in the above embodiment, and are not described herein.
Referring now to fig. 8, a schematic diagram of a configuration of an electronic device 800 suitable for use in implementing embodiments of the present disclosure for implementing functions corresponding to the modeling apparatus 600 of the multitasking model shown in fig. 6, or for implementing functions corresponding to the promotional content processing apparatus 700 shown in fig. 7, is shown. The electronic device shown in fig. 8 is merely an example and should not be construed to limit the functionality and scope of use of the disclosed embodiments.
As shown in fig. 8, the electronic device 800 may include a processing means (e.g., a central processor, a graphics processor, etc.) 801, which may perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 802 or a program loaded from a storage means 808 into a Random Access Memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the electronic device 800 are also stored. The processing device 801, the ROM 802, and the RAM 803 are connected to each other by a bus 804. An input/output (I/O) interface 805 is also connected to the bus 804.
In general, devices may be connected to I/O interface 805 including input devices 806 such as a touch screen, touch pad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc., output devices 807 including a Liquid Crystal Display (LCD), speaker, vibrator, etc., storage devices 808 including magnetic tape, hard disk, etc., and communication devices 809. The communication means 809 may allow the electronic device 800 to communicate wirelessly or by wire with other devices to exchange data. While fig. 8 shows an electronic device 800 having various means, it is to be understood that not all of the illustrated means are required to be implemented or provided. More or fewer devices may be implemented or provided instead.
In particular, according to embodiments of the present disclosure, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a non-transitory computer readable medium, the computer program comprising program code for performing the method shown in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via communication device 809, or installed from storage device 808, or installed from ROM 802. The above-described functions defined in the methods of the embodiments of the present disclosure are performed when the computer program is executed by the processing device 801.
It should be noted that the computer readable medium described in the present disclosure may be a computer readable signal medium or a computer readable storage medium, or any combination of the two. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples of a computer-readable storage medium may include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present disclosure, however, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to electrical wiring, fiber optic cable, RF (radio frequency), and the like, or any suitable combination of the foregoing.
In some embodiments, the clients, servers may communicate using any currently known or future developed network protocol, such as HTTP (HyperText Transfer Protocol ), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), the internet (e.g., the internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future developed networks.
The computer readable medium may be included in the electronic device or may exist alone without being incorporated into the electronic device.
The computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to:
The method comprises the steps of obtaining characteristics for constructing a task, and constructing an initial task set according to the characteristics for constructing the task, wherein the initial task set comprises at least two of conversion rate of promotion content, play time of the promotion content, presentation type of the promotion content and information of promotion objects in the promotion content;
determining mutual information among different tasks in the initial task set;
Acquiring a related task set according to the mutual information of different tasks, wherein the mutual information of the tasks included in the related task set meets a first preset condition;
Generating a sample of a multi-task model according to the corresponding characteristics of each task in the related task set, and performing model training by using the sample of the multi-task model to obtain the multi-task model, or
Acquiring the attribute of the behavior of the user on the promotion content;
Obtaining an inference result of the multitask model according to the attribute of the behavior of the user on the promotion content and the multitask model, wherein the multitask model is obtained based on a sample of the multitask model generated by the characteristics corresponding to each task in a related task set, the related task set is obtained based on mutual information among different tasks, the different tasks are tasks in an initial task set constructed by the characteristics used for constructing the tasks, and the inference result comprises a plurality of types of conversion rate of the promotion content, play time of the promotion content, presentation type of the promotion content or information of promotion objects in the promotion content;
and adjusting the popularization strategy for the popularization content according to the reasoning result.
Computer program code for carrying out operations of the present disclosure may be written in one or more programming languages, including, but not limited to, an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider).
The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The modules described in the embodiments of the present disclosure may be implemented in software or hardware. The name of a module is not limited to the module itself in some cases, and for example, the first acquisition module may also be described as "a module that acquires at least two internet protocol addresses".
The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic that may be used include Field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems-on-a-chip (SOCs), complex Programmable Logic Devices (CPLDs), and the like.
In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
According to one or more embodiments of the present disclosure, example 1 provides a modeling method of a multitasking model, wherein characteristics for constructing a task are obtained, an initial task set is constructed according to the characteristics for constructing the task, the initial task set includes at least two of conversion rate of promotion content, play time of the promotion content, presentation type of the promotion content and information of promotion objects in the promotion content, mutual information between different tasks in the initial task set is determined, a relevant task set is obtained according to the mutual information between the different tasks, the mutual information of tasks included in the relevant task set meets a first preset condition, a sample of the multitasking model is generated according to the characteristics corresponding to each task in the relevant task set, and model training is performed by using the sample of the multitasking model to obtain the multitasking model.
In accordance with one or more embodiments of the present disclosure, example 2 provides the method of example 1, the method further comprising:
and aggregating the tasks meeting the second preset condition in the related task set.
According to one or more embodiments of the present disclosure, example 3 provides the method of example 2, wherein aggregating the tasks meeting the second preset condition in the related task set includes:
Determining the correlation between the outputs of task exclusive networks corresponding to a plurality of tasks in the related task set;
And aggregating a plurality of tasks in the related task set according to the relevance.
In accordance with one or more embodiments of the present disclosure, example 4 provides the method of example 3, the aggregated set of related tasks comprising a plurality of task groups, the method further comprising:
For each task group, determining a plurality of target tasks according to the positive sample number of the tasks in the task group, wherein the sum of the positive sample numbers of the target tasks is larger than a positive sample number threshold, and the positive sample number threshold is the minimum value of the sum of the positive sample numbers in each task group;
the generating a sample of the multi-task model according to the characteristics corresponding to each task in the related task set includes:
and generating a sample of the multi-task model according to the characteristics corresponding to the target task in the task group.
Example 5 provides the method of example 4, according to one or more embodiments of the present disclosure, the determining a plurality of target tasks from the number of positive samples of tasks in the task group, comprising:
adding the positive samples of the tasks in the task group one by one according to the sequence from small to large until the sum of the positive samples of the tasks in the task group is greater than or equal to the threshold value of the positive samples;
And when the sum of the positive sample numbers of the tasks in the task group is greater than or equal to the threshold value of the positive sample number, the tasks in the task group are multiple target tasks.
Example 6 provides the method of example 1, according to one or more embodiments of the present disclosure, the model training using samples of the multitasking model, comprising
Inputting the feature vector of the sample of the multitasking model into a sharing network to obtain a sharing component;
Inputting the shared component to the task exclusive network of each task in the related task set to obtain the output corresponding to the task exclusive network of each task;
training the multi-task model according to the label value of the characteristic vector of the sample of the multi-task model and the output of the task exclusive network.
Example 7 provides the method of example 6, wherein the task exclusive network comprises a deep neural network, a convolutional neural network, or a self-attention network, and wherein the shared network comprises a deep neural network, a convolutional neural network, or a self-attention network, according to one or more embodiments of the present disclosure.
According to one or more embodiments of the present disclosure, example 8 provides a promotion content processing method, which includes obtaining an attribute of a user's behavior on a promotion content, obtaining an inference result of a multitask model according to the attribute of the user's behavior on the promotion content and the multitask model, wherein the multitask model is obtained based on a sample of the multitask model generated by features corresponding to tasks in a related task set, the related task set is obtained based on mutual information among different tasks, the different tasks are tasks in an initial task set constructed by features for constructing tasks, the inference result includes a plurality of types of conversion rate of promotion content, play time of the promotion content, presentation type of promotion content or information of promotion objects in the promotion content, and adjusting a promotion policy for the promotion content according to the inference result.
The foregoing description is only of the preferred embodiments of the present disclosure and description of the principles of the technology being employed. It will be appreciated by persons skilled in the art that the scope of the disclosure referred to in this disclosure is not limited to the specific combinations of features described above, but also covers other embodiments which may be formed by any combination of features described above or equivalents thereof without departing from the spirit of the disclosure. Such as those described above, are mutually substituted with the technical features having similar functions disclosed in the present disclosure (but not limited thereto).
Moreover, although operations are depicted in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order. In certain circumstances, multitasking and parallel processing may be advantageous. Likewise, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the present disclosure. Certain features that are described in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable subcombination.
Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are example forms of implementing the claims. The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.
Claims (10)
1. A method of modeling a multitasking model, comprising:
The method comprises the steps of obtaining characteristics for constructing a task, and constructing an initial task set according to the characteristics for constructing the task, wherein the initial task set comprises at least two of conversion rate of promotion content, play time of the promotion content, presentation type of the promotion content and information of promotion objects in the promotion content;
determining mutual information among different tasks in the initial task set;
Acquiring a related task set according to the mutual information of different tasks, wherein the mutual information of the tasks included in the related task set meets a first preset condition;
generating a sample of a multi-task model according to the corresponding characteristics of each task in the related task set, and performing model training by using the sample of the multi-task model to obtain the multi-task model;
Determining the correlation between the outputs of task exclusive networks corresponding to a plurality of tasks in the related task set;
The aggregated set of related tasks includes a plurality of task groupings, the method further comprising:
For each task group, determining a plurality of target tasks according to the positive sample number of the tasks in the task group, wherein the sum of the positive sample numbers of the target tasks is larger than a positive sample number threshold, and the positive sample number threshold is the minimum value of the sum of the positive sample numbers in each task group;
the generating a sample of the multi-task model according to the characteristics corresponding to each task in the related task set includes:
and generating a sample of the multi-task model according to the characteristics corresponding to the target task in the task group.
2. The method of claim 1, wherein the determining a plurality of target tasks based on the number of positive samples of tasks in the task group comprises:
adding the positive samples of the tasks in the task group one by one according to the sequence from small to large until the sum of the positive samples of the tasks in the task group is greater than or equal to the threshold value of the positive samples;
And when the sum of the positive sample numbers of the tasks in the task group is greater than or equal to the threshold value of the positive sample number, the tasks in the task group are multiple target tasks.
3. The method of claim 1, wherein the model training using samples of the multitasking model comprises
Inputting the feature vector of the sample of the multitasking model into a sharing network to obtain a sharing component;
Inputting the shared component to the task exclusive network of each task in the related task set to obtain the output corresponding to the task exclusive network of each task;
training the multi-task model according to the label value of the characteristic vector of the sample of the multi-task model and the output of the task exclusive network.
4. The method of claim 3, wherein the task exclusive network comprises a deep neural network, a convolutional neural network, or a self-attention network, and wherein the shared network comprises a deep neural network, a convolutional neural network, or a self-attention network.
5. A promotional content processing method, comprising:
acquiring the attribute of the behavior of the user on the promotion content;
Obtaining an inference result of the multitask model according to the attribute of the behavior of the user on the promotion content and the multitask model; the task clustering method comprises the steps that a multitask model is obtained based on samples of the multitask model generated by characteristics corresponding to target tasks in task groups, a relevant task set is obtained based on mutual information among different tasks, the task groups are obtained by aggregation of the relevant task sets, the relevant task sets are obtained by aggregation of a plurality of tasks in the relevant task sets according to relevance, the relevance is the relevance among outputs of task exclusive networks corresponding to the plurality of tasks in the relevant task sets, the target tasks are determined according to the positive sample number of the tasks in each task group, the sum of the positive sample number of the plurality of target tasks is larger than a positive sample number threshold, the positive sample number threshold is the minimum value of the sum of the positive sample number in each task group, and the different tasks are tasks in an initial task set constructed by the characteristics used for constructing the tasks;
and adjusting the popularization strategy for the popularization content according to the reasoning result.
6. A modeling apparatus for a multitasking model, comprising:
the acquisition module is used for acquiring characteristics for constructing a task and constructing an initial task set according to the characteristics for constructing the task, wherein the initial task set comprises at least two of conversion rate of promotion content, play time of the promotion content, presentation type of the promotion content and information of promotion objects in the promotion content;
The mutual information determining module is used for determining mutual information among different tasks in the initial task set;
the related task determining module is used for obtaining a related task set according to the mutual information among different tasks, wherein the mutual information of the tasks included in the related task set meets a first preset condition;
The training module is used for generating samples of a multi-task model according to the characteristics corresponding to each task in the related task set, and carrying out model training by utilizing the samples of the multi-task model to obtain the multi-task model;
The related task determining module is also used for determining the correlation between the outputs of the task exclusive network corresponding to the tasks in the related task set;
The related task determining module is further configured to determine, for each task group, a plurality of target tasks according to the positive sample numbers of the tasks in the task group, where a sum of the positive sample numbers of the plurality of target tasks is greater than a positive sample number threshold, and the positive sample number threshold is a minimum value of the sum of the positive sample numbers in each task group;
the training module is specifically configured to generate a sample of the multitasking model according to the features corresponding to the target tasks in the task group.
7. A promotional content processing device, comprising:
the acquisition module is used for acquiring the attribute of the behavior of the user on the promotion content;
The reasoning module is used for obtaining a reasoning result of the multitasking model according to the attribute of the behavior of the user on the popularization content and the multitasking model; the task model is obtained based on samples of the multi-task model generated by characteristics corresponding to target tasks in task groups, the task groups are obtained by aggregating related task sets, the related task sets are obtained by aggregating a plurality of tasks in the related task sets according to correlation, the correlation is the correlation between outputs of task exclusive networks corresponding to the tasks in the related task sets, the target tasks are determined according to positive sample numbers of the tasks in the task groups for each task group, the sum of the positive sample numbers of the target tasks is larger than a positive sample number threshold, the positive sample number threshold is a minimum value of the sum of the positive sample numbers in each task group, the related task sets are obtained based on mutual information among different tasks, and the different tasks are tasks in an initial task set constructed by the characteristics for constructing the tasks;
and the processing module is used for adjusting the popularization strategy for the popularization content according to the reasoning result.
8. An electronic device, comprising:
a storage device having a computer program stored thereon;
Processing means for executing said computer program in said storage means to carry out the steps of the method according to any one of claims 1 to 4 or to carry out the steps of the method according to claim 5.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the program when being executed by a processing device implements the steps of the method according to any one of claims 1 to 4, or the program when being executed by a processing device implements the steps of the method according to claim 5.
10. A computer program product for causing a computer to perform the steps of the method as claimed in any one of claims 1 to 4 when said computer program product is run on a computer or for causing a computer to perform the steps of the method as claimed in claim 5 when said computer program product is run on a computer.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111573367.9A CN114240506B (en) | 2021-12-21 | 2021-12-21 | Multi-task model modeling method, promotion content processing method and related device |
| PCT/CN2022/124765 WO2023116138A1 (en) | 2021-12-21 | 2022-10-12 | Modeling method for multi-task model, promotional content processing method, and related apparatuses |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111573367.9A CN114240506B (en) | 2021-12-21 | 2021-12-21 | Multi-task model modeling method, promotion content processing method and related device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114240506A CN114240506A (en) | 2022-03-25 |
| CN114240506B true CN114240506B (en) | 2025-04-11 |
Family
ID=80760649
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111573367.9A Active CN114240506B (en) | 2021-12-21 | 2021-12-21 | Multi-task model modeling method, promotion content processing method and related device |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN114240506B (en) |
| WO (1) | WO2023116138A1 (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114240506B (en) * | 2021-12-21 | 2025-04-11 | 北京有竹居网络技术有限公司 | Multi-task model modeling method, promotion content processing method and related device |
| CN114860411B (en) * | 2022-05-17 | 2023-05-05 | 北京百度网讯科技有限公司 | Multi-task learning method, device, electronic equipment and storage medium |
| CN116992134B (en) * | 2023-07-07 | 2025-09-16 | 中国平安人寿保险股份有限公司 | Information prediction method and device based on multitasking model, electronic equipment and medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378229A (en) * | 2019-06-19 | 2019-10-25 | 浙江大学 | A kind of electronic nose data characteristics selection method based on filter-wrapper frame |
| CN112559007A (en) * | 2020-12-14 | 2021-03-26 | 北京百度网讯科技有限公司 | Parameter updating method and device of multitask model and electronic equipment |
Family Cites Families (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070218998A1 (en) * | 2005-09-12 | 2007-09-20 | Arbogast Christopher P | Download and configuration method for gaming machines |
| US8924315B2 (en) * | 2011-12-13 | 2014-12-30 | Xerox Corporation | Multi-task learning using bayesian model with enforced sparsity and leveraging of task correlations |
| US9378065B2 (en) * | 2013-03-15 | 2016-06-28 | Advanced Elemental Technologies, Inc. | Purposeful computing |
| CN107045506A (en) * | 2016-02-05 | 2017-08-15 | 阿里巴巴集团控股有限公司 | Evaluation index acquisition methods and device |
| CN109165997A (en) * | 2018-07-19 | 2019-01-08 | 阿里巴巴集团控股有限公司 | It does shopping under a kind of line the generation method and device of recommendation |
| CN109784476B (en) * | 2019-01-12 | 2022-08-16 | 福州大学 | Method for improving DSOD network |
| CN109977402B (en) * | 2019-03-11 | 2022-11-11 | 北京明略软件系统有限公司 | Named entity identification method and system |
| US11403668B2 (en) * | 2019-10-30 | 2022-08-02 | Target Brands, Inc. | Multitask transfer learning for optimization of targeted promotional programs |
| CN111178981B (en) * | 2020-01-02 | 2023-11-28 | 众安在线财产保险股份有限公司 | Advertisement putting method and device, computer equipment and storage medium |
| CN111951249A (en) * | 2020-08-13 | 2020-11-17 | 浙江理工大学 | Visual detection method of mobile phone light guide plate defects based on multi-task learning network |
| CN112202901A (en) * | 2020-09-30 | 2021-01-08 | 中国银行股份有限公司 | Promotion information pushing method, device, equipment and medium |
| CN112163676B (en) * | 2020-10-13 | 2024-04-05 | 北京百度网讯科技有限公司 | Method, device, equipment and storage medium for training multitasking service prediction model |
| CN112561077B (en) * | 2020-12-14 | 2022-06-21 | 北京百度网讯科技有限公司 | Training method, device and electronic device for multi-task model |
| CN112967086B (en) * | 2021-02-27 | 2022-08-02 | 北京橙色风暴数字技术有限公司 | Intelligent marketing promotion method and device and electronic equipment |
| CN113205362A (en) * | 2021-04-30 | 2021-08-03 | 北京有竹居网络技术有限公司 | Method, apparatus, device, storage medium and program product for determining a promoter |
| CN113344086B (en) * | 2021-06-16 | 2022-07-01 | 深圳市商汤科技有限公司 | Man-machine loop method, device, system, electronic equipment and storage medium |
| CN114240506B (en) * | 2021-12-21 | 2025-04-11 | 北京有竹居网络技术有限公司 | Multi-task model modeling method, promotion content processing method and related device |
-
2021
- 2021-12-21 CN CN202111573367.9A patent/CN114240506B/en active Active
-
2022
- 2022-10-12 WO PCT/CN2022/124765 patent/WO2023116138A1/en not_active Ceased
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378229A (en) * | 2019-06-19 | 2019-10-25 | 浙江大学 | A kind of electronic nose data characteristics selection method based on filter-wrapper frame |
| CN112559007A (en) * | 2020-12-14 | 2021-03-26 | 北京百度网讯科技有限公司 | Parameter updating method and device of multitask model and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114240506A (en) | 2022-03-25 |
| WO2023116138A1 (en) | 2023-06-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114240506B (en) | Multi-task model modeling method, promotion content processing method and related device | |
| US20230206132A1 (en) | Method and Apparatus for Training AI Model, Computing Device, and Storage Medium | |
| WO2019180433A1 (en) | Predicting using digital twins | |
| CN114282681A (en) | Multitask processing and model training method, device, medium and equipment | |
| JP7652916B2 (en) | Method and apparatus for pushing information - Patents.com | |
| CN111340220B (en) | Method and apparatus for training predictive models | |
| Xu et al. | Joint foundation model caching and inference of generative AI services for edge intelligence | |
| WO2022001887A1 (en) | Method and apparatus for training item coding model | |
| WO2024152686A1 (en) | Method and apparatus for determining recommendation index of resource information, device, storage medium and computer program product | |
| CN118331716A (en) | Intelligent migration method for calculation force under heterogeneous calculation force integrated system | |
| CN119902904A (en) | Intelligent computing center model development method and device for inclusive computing power | |
| CN109598344B (en) | Model generation method and device | |
| CN117009631A (en) | Method, device, equipment and storage medium for screening put objects | |
| CN114611610A (en) | A video popularity prediction method, device, device and medium | |
| CN113641930A (en) | Information output method and device and electronic equipment | |
| CN114021010B (en) | A training method, device and equipment for information recommendation model | |
| CN110689117A (en) | Information processing method and device based on neural network | |
| CN113837808B (en) | Promotion information pushing method, device, equipment, medium and product | |
| CN116594762A (en) | Method and device for advanced scheduling of dynamic computational graph | |
| CN110633596A (en) | Method and device for predicting vehicle direction angle | |
| CN114201674A (en) | Model training method, promotion content processing method and related device | |
| CN114461911B (en) | Multi-task model training method, promotion content processing method and related devices | |
| CN113472550A (en) | Distributed management method and system, and management system | |
| CN119109978B (en) | Message pushing method, device, equipment and medium | |
| CN110633707A (en) | Method and device for predicting speed |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |