Task result prediction method and device, equipment and storage medium

Abstract

The embodiment of the invention provides a task result prediction method and device, equipment and a storage medium. The method comprises the steps of obtaining input of a recommendation task, wherein the input comprises at least one feature corresponding to a plurality of feature items of a plurality of objects associated with the recommendation task; encoding the input by using a trained encoder model to obtain a first feature representation, the first feature representation representing at least one feature corresponding to the plurality of feature items; decoding the first feature representation by using a trained decoder model to obtain a second feature representation, the second feature representation representing a plurality of features corresponding to the plurality of feature items; and determining a first prediction result of the recommendation task by using a trained prediction model at least based on the second feature representation, the first prediction result indicating whether a first object in the plurality of objects is to be recommended to at least one second object in the plurality of objects.

Description

Technical Field

[0001] Example embodiments of the present disclosure generally relate to the field of computers, and in particular, to methods, apparatuses, devices, computer-readable storage media, and computer program products for task result prediction. Background Art

[0002] With the development of machine learning technology, machine learning technology has been widely applied in various industries. As a result, more and more applications provide services to users based on the prediction results of machine learning models to improve service quality. For example, in a recommendation system, recommendation content can be provided to users based on the prediction results of a machine learning model to improve the accuracy of the recommendation content. Also, for example, in a search engine, search content can be provided to users based on the prediction results of a machine learning model to improve the accuracy of the search content. Summary of the Invention

[0003] In a first aspect of the present disclosure, a method for task result prediction is provided. The method includes: obtaining an input of a recommendation task, the input including at least one feature corresponding to a plurality of feature items of a plurality of objects associated with the recommendation task; encoding the input by using a trained encoder model to obtain a first feature representation, the first feature representation characterizing at least one feature corresponding to the plurality of feature items; decoding the first feature representation by using a trained decoder model to obtain a second feature representation, the second feature representation characterizing a plurality of features corresponding to the plurality of feature items; and determining a first prediction result of the recommendation task by using a trained prediction model based at least on the second feature representation, the first prediction result indicating whether a first object among the plurality of objects is to be recommended to at least one second object among the plurality of objects.

[0004] In a second aspect of the present disclosure, an apparatus for task result prediction is provided. The apparatus includes: an obtaining module configured to obtain an input of a recommendation task, the input including at least one feature corresponding to a plurality of feature items of a plurality of objects associated with the recommendation task; an encoding module configured to encode the input by using a trained encoder model to obtain a first feature representation, the first feature representation characterizing at least one feature corresponding to the plurality of feature items; a decoding module configured to decode the first feature representation by using a trained decoder model to obtain a second feature representation, the second feature representation characterizing a plurality of features corresponding to the plurality of feature items; and a determining module configured to determine a first prediction result of the recommendation task by using a trained prediction model based at least on the second feature representation, the first prediction result indicating whether a first object among the plurality of objects is to be recommended to at least one second object among the plurality of objects.

[0005] In a third aspect of the present disclosure, an electronic device is provided. The device includes at least one processor; and at least one memory, the at least one memory being coupled to the at least one processor and storing instructions for execution by the at least one processor. The instructions, when executed by the at least one processor, cause the device to perform the method of the first aspect.

[0006] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided. The computer-readable storage medium stores computer-executable instructions that can be executed by a processor to implement the method of the first aspect.

[0007] In a fifth aspect of the present disclosure, a computer program product is provided, including computer-executable instructions, wherein the computer-executable instructions, when executed by a processor, implement the method according to the first aspect of the present disclosure.

[0008] It should be understood that the content described in this part is not intended to define the key features or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In combination with the accompanying drawings and with reference to the following detailed description, the above and other features, advantages, and aspects of the embodiments of the present disclosure will become more apparent. In the drawings, the same or similar reference numerals represent the same or similar elements, where:

[0010] Figure 1 A schematic diagram showing an example environment in which embodiments according to the present disclosure can be implemented;

[0011] Figure 2 A flowchart showing a process for task result prediction according to some embodiments of the present disclosure;

[0012] Figure 3 A schematic diagram showing an example architecture for task result prediction according to some embodiments of the present disclosure;

[0013] Figure 4 A schematic diagram showing an example architecture for task result prediction according to some embodiments of the present disclosure;

[0014] Figure 5 A schematic structural block diagram showing an example device for task result prediction according to other embodiments of the present disclosure; and

[0015] Figure 6 A block diagram showing an electronic device capable of implementing multiple embodiments of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0017] In the description of the embodiments of the present disclosure, the term "including" and its like should be understood as an open inclusion, that is, "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "an embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". There may also be other explicit and implicit definitions hereinafter.

[0018] In this article, unless expressly stated, performing a step "in response to A" does not mean that the step is performed immediately after "A", but may include one or more intermediate steps.

[0019] It can be understood that the data involved in the technical solution (including but not limited to the data itself, the acquisition or use of the data) should comply with the requirements of the corresponding laws, regulations and related provisions.

[0020] It can be understood that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, usage scopes, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner according to the relevant laws and regulations.

[0021] For example, when a user's active request is received, a prompt message is sent to the user to clearly prompt the user that the operation requested by the user will require the acquisition and use of the user's personal information, so that the user can autonomously choose whether to provide personal information to software or hardware such as an electronic device, an application program, a server or a storage medium that performs the operation of the technical solution of the present disclosure according to the prompt message.

[0022] As an optional but non-limiting implementation manner, the manner of sending a prompt message to the user in response to receiving the user's active request may be, for example, in the form of a pop-up window, and the prompt message may be presented in text in the pop-up window. In addition, the pop-up window may also carry a selection control for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0023] It can be understood that the above process of notifying and obtaining the user's authorization is only illustrative and does not limit the implementation manner of the present disclosure. Other manners that meet the relevant laws and regulations can also be applied to the implementation manner of the present disclosure.

[0024] As used herein, the term "model" can learn the corresponding association between inputs and outputs from training data, so that after training, for a given input, the corresponding output can be generated. The generation of the model can be based on machine learning techniques. Deep learning is a machine learning algorithm that processes inputs and provides corresponding outputs by using multiple layers of processing units. A neural network model is an example of a model based on deep learning. In this document, "model" can also be referred to as "machine learning model", "learning model", "machine learning network" or "learning network", and these terms are used interchangeably herein.

[0025] A "neural network" is a machine learning network based on deep learning. A neural network can process inputs and provide corresponding outputs, and it generally includes an input layer and an output layer and one or more hidden layers between the input layer and the output layer. Neural networks used in deep learning applications usually include many hidden layers, thus increasing the depth of the network. The layers of the neural network are connected in sequence, so that the output of the previous layer is provided as the input of the next layer, where the input layer receives the input of the neural network, and the output of the output layer is the final output of the neural network. Each layer of the neural network includes one or more nodes (also called processing nodes or neurons), and each node processes the input from the previous layer.

[0026] Generally, machine learning can roughly include three stages, namely, a training stage, a testing stage, and an application stage (also called an inference stage). In the training stage, a given model can be trained using a large amount of training data, and the parameter values are continuously iteratively updated until the model can obtain consistent inferences that meet the expected goals from the training data. Through training, the model can be considered to be able to learn the association from input to output (also called input-to-output mapping) from the training data. The parameter values of the trained model are determined. In the testing stage, the test inputs are applied to the trained model to test whether the model can provide correct outputs, so as to determine the performance of the model. In the application stage, the model can be used to process actual inputs based on the parameter values obtained from training and determine the corresponding outputs.

[0027] As mentioned above, with the development of machine learning technology, machine learning technology has been widely applied in various industries. Subsequently, more and more applications provide services to users based on the prediction results of machine learning models to improve service quality. For example, in a recommendation system, recommendation content can be provided to users based on the prediction results of a machine learning model to improve the accuracy of the recommendation content. Also for example, in a search engine, search content can be provided to users based on the prediction results of a machine learning model to improve the accuracy of the search content. However, in actual applications, it is easy for the model input of a machine learning model to lack features. For example, in a recommendation system, it is easy to lack one or more features associated with the recommendation content. This situation will affect the accuracy of the prediction results of the machine learning model.

[0028] In view of this, embodiments of the present disclosure propose an improved solution for task result prediction. In this solution, the input of a recommendation task is obtained. The input includes at least one feature corresponding to multiple feature items of multiple objects associated with the recommendation task. The input is encoded using a trained encoder model to obtain a first feature representation, which represents at least one feature corresponding to the multiple feature items. The first feature representation is decoded using a trained decoder model to obtain a second feature representation, which represents multiple features corresponding to the multiple feature items; and at least based on the second feature representation, a trained prediction model is used to determine a first prediction result of the recommendation task. The first prediction result can indicate whether a first object among the multiple objects is to be recommended to at least one second object among the multiple objects.

[0029] In embodiments of the present disclosure, if some features are missing in the input, using the decoder to decode the first feature representation to obtain the second feature representation can, to a certain extent, play a role in supplementing the missing features and can improve the representation ability of the input. Using the prediction model to determine the prediction result of the recommendation task based on the supplemented second feature representation can improve the accuracy of the prediction result of the recommendation task.

[0030] The following further describes various example implementations of this solution in detail with reference to the accompanying drawings.

[0031] Example environment

[0032] Figure 1 A schematic diagram of an example environment 100 in which embodiments of the present disclosure can be implemented is shown. In this example environment 100, an application 120 is installed in a terminal device 110. A user 140 can interact with the application 120 via the terminal device 110 and / or an attached device of the terminal device 110.

[0033] In some embodiments of the present disclosure, the application 120 can be any suitable application that can provide services to the user 140 based on the prediction results of the machine learning model 160. In some embodiments, the application 120 can present content to the user 140 based on the prediction results of the machine learning model 160. The content can include, but is not limited to, web pages, documents, images, audio, video, or content in other modalities.

[0034] In some embodiments, if the application 120 is active, the terminal device 110 can present the user interface 150 of the application 120. The user interface 150 can include various types of content that the application 120 can provide, such as a conversation page between the user and a digital assistant (where the current conversation and historical conversations, including text conversation content, can be presented), a presentation interface for text content, a playback interface for voice, a playback interface for video, and so on.

[0035] In some embodiments, the application 120 can utilize one or more machine learning models 160 (for example, it can include machine learning models 160-1, 160-2,..., 160-M, etc., where M is a positive integer. For the sake of convenience of description, one or more machine learning models are collectively referred to as the machine learning model 160 in this document) to support the interaction with the user 140. For example, the application 120 can utilize one or more machine learning models 160 to determine the content that matches the user 140.

[0036] In some embodiments, the terminal device 110 communicates with the server device 130 to implement the supply of services for the application 120. As Figure 1 shown, the server device 130 can invoke the machine learning model 160 to support the interaction between the application 120 and the user 140 based on the prediction results of the machine learning model 160. The terminal device 110 can be any type of mobile terminal, fixed terminal, or portable terminal, including mobile phones, desktop computers, laptop computers, notebook computers, netbook computers, tablet computers, media computers, multimedia tablets, personal communication system (PCS) devices, personal navigation devices, personal digital assistants (PDAs), audio / video players, digital cameras / cameras, positioning devices, television receivers, radio broadcast receivers, e-book devices, gaming devices, or any combination of the foregoing, including accessories and peripherals of these devices or any combination thereof. In some embodiments, the terminal device 110 can also support any type of interface for the user (such as a "wearable" circuit, etc.). The server device 130 can be various types of computing systems / servers that can provide computing capabilities, including but not limited to mainframes, edge computing nodes, computing devices in a cloud environment, and so on. The server device 130 can be implemented based on a cloud environment, for example.

[0037] The machine learning model 160 can be of different types. In some embodiments, one or more machine learning models 160 can be implemented as an encoder model, a decoder model, or a prediction model. In some embodiments, one or more machine learning models 150 can be constructed based on a Diffusion Model. The Diffusion Model, also known as the Diffusion Probability Model, is a type of generative model. This model generates data by simulating a diffusion process. This process is inspired by physical processes such as heat diffusion. The Diffusion Model includes a Forward Diffusion Process and a Reverse Diffusion Process. The Diffusion Model generates new data samples by simulating a forward diffusion process that gradually adds noise and then learning how to reverse this process.

[0038] In the forward diffusion process, noise is gradually added to the data, making the data more and more random through a series of steps until the data resembles pure noise. This process can be regarded as a Markov chain, and Gaussian noise is added to the data at each step. The forward diffusion process can be expressed as: where \(x\) t is the noisy data at the \(t\)-th step, and \(\alpha\) t is used to control the amount of noise added. The forward diffusion process is performed during model training, and the data with added noise is the training sample.

[0039] In the Reverse Diffusion Process (or reverse denoising process), the model learns how to reverse the steps of adding noise. Starting from pure noise, the diffusion model gradually removes the noise and generates data that matches the training distribution. The reverse diffusion process is usually simulated using a neural network that predicts the noise added at each step: where \(u\) θ and \(\sigma\) θ are the learned model parameters. After completing model training, the model that performs the reverse diffusion process can first sample from the noise distribution and use the model for iterative denoising until the desired data is obtained.

[0040] In the Diffusion Model, the Time Step refers to the number of steps of adding noise in the forward diffusion process. The total number of steps \(T\) is usually a preset value, indicating how many steps are required for the transformation from the original data to pure noise. At each time step \(t\), Gaussian noise is added to the data according to a predetermined noise scheme. This process is continuous, and each step depends on the result of the previous step.

[0041] When generating data, the inference step of the diffusion model refers to the number of steps required to recover from pure noise to the original data during the reverse diffusion process. The number of inference steps directly affects the quality and speed of the generated data. Generally, the more inference steps, the higher the quality of the generated data, but it will also increase the computational cost and time. In practical applications, the balance between generation quality and efficiency can be achieved by adjusting the number of inference steps. In some embodiments, the inference steps correspond to time steps, and each inference step can correspond to one or more time steps. For example, if the total number of time steps of the diffusion model is 1000 steps and the inference steps are set to 50 steps, then each inference step can correspond to 20 time steps.

[0042] It should be understood that the structures and functions of the various elements in the environment 100 are described only for exemplary purposes and do not imply any limitation on the scope of the present disclosure.

[0043] Some example embodiments of the present disclosure will be described below with continued reference to the accompanying drawings.

[0044] Example process

[0045] Some example embodiments of the present disclosure will be described below with continued reference to the accompanying drawings. Figure 2 A flowchart of a process 200 for task result prediction according to some embodiments of the present disclosure is shown. Part or all of the process 200 can be implemented by the server device 130, or can be implemented in cooperation with the server device 130 and the terminal device 110. Hereinafter, for the convenience of discussion, the execution of the process 200 is described from the perspective of the server device 130, but this is only exemplary.

[0046] In block 210, the server device 130 obtains the input of the recommendation task. In some embodiments, the recommendation task may include a plurality of associated objects, and the plurality of objects may include at least one first object and at least one second object. The prediction result of the recommendation task is to determine whether a first object among the plurality of objects is to be recommended to at least one second object among the plurality of objects. The input may include at least one feature corresponding to a plurality of feature items of the plurality of objects. As an example, the input can be represented as x = {x1, x2,..., x N}, x i represents the feature corresponding to the i-th feature item, i ∈ [1, N], and N represents the number of feature items. In practical applications, some feature items in the input may lack corresponding features, and the feature items lacking corresponding features can be set to 0 or an empty set. For example, x1 or x2 can be set to 0.

[0047] In some embodiments, the recommendation task may indicate determining the degree of association between the at least one first object and the at least one second object. The input may include at least one first feature corresponding to a plurality of first feature items of the at least one first object, and at least one second feature corresponding to a plurality of second feature items of the at least one second object. Here, the features of the first object and the features of the second object may be the same or different.

[0048] As an example, the at least one first object may include at least one recommended content, and the at least one second object may include at least one recommended audience. The task may indicate determining the degree of association between the at least one recommended content and the at least one recommended audience. The application 120 may be configured to present the recommended content to the user 140 according to the degree of association between the at least one recommended content and the at least one recommended audience. The input may include features corresponding to a plurality of feature items of the recommended content, and features corresponding to a plurality of feature items of the recommended audience. In practical applications, for some of the feature items required in the input that are related to the recommended content or some of the feature items related to the recommended audience, the corresponding features may not be collected. If the degree of association between the recommended content and the recommended audience is directly predicted based on the input lacking features, the accuracy of the prediction result will be low.

[0049] It should be noted that the process 200 for predicting the task result can be used for other appropriate tasks, not limited to the recommendation task. In some embodiments, the process 200 can also be used for the feature attribute determination task. The task may indicate analyzing the feature attributes of at least one object associated with the task. As an example, the at least one object may include at least one piece of content, and the task may indicate determining the quality score of the at least one piece of content. The input may include the theme, type, subject matter, style, etc. of the content. As another example, the object may include a physical object or an entity object, such as a commodity. The task may indicate determining the recommendation coefficient of at least one commodity, and the recommendation coefficient may indicate the degree to which the corresponding commodity is worthy of recommendation. The input may include the basic features (such as name, code, manufacturer) of the at least one commodity, physical features (such as material, color, size), and functional features (such as main function, auxiliary function). In practical applications, for some of the feature items of the commodity required in the input, the corresponding features may not be collected. If the recommendation coefficient of the commodity is directly predicted based on the input lacking features, the accuracy of the predicted recommendation coefficient may be low.

[0050] In some embodiments, process 200 can also be used for classification tasks. The task can indicate determining a classification label for at least one classification object of the classification task. The input can include at least one feature of a plurality of feature items of the at least one classification object. As an example, the at least one classification object can include the content provided by application 120, such as music content, text content, video content, or live content, etc. The task can indicate determining a classification label for the content, such as positive, negative, neutral, romantic, comedy, science fiction, etc. The input corresponding to the classification task can include metadata of the content (such as the publisher, publication time), structural features (such as the title, length), text features, semantic features, image features, audio features, etc.

[0051] It should be understood that the above tasks and the inputs corresponding to the tasks are all exemplary. In practical applications, other arbitrary appropriate tasks can also be included. Correspondingly, the input can include various appropriate features required to determine the prediction result of the corresponding task. Embodiments of the present disclosure do not limit the type of the task and the specific content of the features in the input.

[0052] In block 220, the server device 130 encodes the input using the trained encoder model to obtain a first feature representation. The first feature representation characterizes at least one feature corresponding to a plurality of feature items. The encoder model can be trained to perform encoding on the features in the input to obtain a feature representation that encodes the input into a specific feature space. In some embodiments, the server device 130 can use the trained encoder model to encode the input and obtain a first feature representation encoded into the latent feature space of the encoder model. In some embodiments, the encoder model can be constructed based on a deep network including an embedding layer. Of course, the encoder model can be constructed using any other appropriate model structure. Embodiments of the present disclosure do not limit the model structure of the encoder model.

[0053] In some embodiments, the input can include at least one first feature corresponding to a plurality of first feature items of a first object, and at least one second feature corresponding to a plurality of second feature items of a second object. The server device 130 can use the encoder model to perform encoding on the at least one first feature and the at least one second feature to obtain a first feature representation encoded into the latent feature space.

[0054] At block 230, the server device 130 decodes the first feature representation using the trained decoder model to obtain a second feature representation, which characterizes multiple features corresponding to multiple feature items. The decoder model can be trained to perform decoding through the first feature representation to supplement the feature representation corresponding to the missing features in the input and obtain the supplemented second feature representation. In this case, the process of decoding the first feature representation using the trained decoder model to obtain the second feature representation can be understood as a "feature completion" process, which can, to a certain extent, supplement the feature representation corresponding to the missing features in the input. For example, supplement the feature representation corresponding to some features associated with the recommended content, or supplement the feature representation corresponding to some features associated with the recommended audience.

[0055] In some embodiments, the decoder model can be trained to perform decoding on the first feature representation in a specific feature space to obtain the second feature representation in the specific feature space, so as to avoid changing the feature space of the feature representation while completing the features. As an example, the decoder model can be trained to perform decoding on the first feature representation in the latent feature space to obtain the second feature representation in the latent feature space. In some embodiments, the encoder model and the decoder model are jointly trained.

[0056] In some embodiments, the decoder model performs decoding based on the reverse diffusion process corresponding to the diffusion model. In this case, the decoder model can also be referred to as a decoder based on the diffusion model. If some feature items in the input are missing the corresponding features, the first feature representation generated based on the input can be regarded as a feature representation containing noise. Using the reverse diffusion process of the diffusion model to decode the first feature representation can be regarded as removing the noise from the feature representation containing noise to obtain the second feature representation after removing the noise. The second feature representation after removing the noise supplements, to a certain extent, the feature representation corresponding to the missing features. Compared with the first feature representation that can characterize at least one feature included in the input, the second feature representation can not only characterize at least one feature included in the input, but also characterize the determined features to a certain extent. In this way, the ability of the diffusion model to remove noise from the feature representation is utilized to achieve the purpose of completing the input with missing features, and a relatively complete feature representation can be obtained. It should be understood that using the diffusion model to implement the decoder model is only exemplary, and in actual applications, other model structures with denoising capabilities can also be selected according to actual needs to implement the decoder model. The embodiments of the present disclosure do not specifically limit the model structure of the decoder model.

[0057] In some embodiments, Figure 3 FIG. shows a schematic diagram of an example architecture 300 for task result prediction according to some embodiments of the present disclosure. As Figure 3As shown, the server device 130 can encode the input 302 using the encoder model 304 to obtain the first feature representation 306. The server device 130 can also determine the missing feature representation 308 corresponding to the input 302. The server device 130 can decode the first feature representation 306 using the decoder model 310 based on the missing feature representation 308 to obtain the second feature representation 312. The missing feature representation here can indicate the feature items among multiple feature items that are missing the corresponding features. By using the missing feature representation to indicate the feature items missing the corresponding features, the decoder model can specifically complete the missing features, which can improve the accuracy of the second feature representation.

[0058] As an example, the input 302 can be represented as x = {x1, x2,..., x N}, where x i represents features. The missing feature representation 308 can be represented as s = {s1, s2,..., s N}, where s i can indicate whether the feature item corresponding to x i is missing the corresponding feature. If the feature item corresponding to x i is missing the corresponding feature, s i can be set to 1. If the feature item corresponding to x i is not missing the corresponding feature, it can be set to 0. The encoder model 304 can be represented as h(·), and the first feature representation 306 can be represented as z = h(x). The denoising function of the decoder model 310 can be represented as g(·). In this case, the second feature representation 312 can be represented as z' = g([s, h(x)]). It should be understood that the specific data structure of the above missing feature representation is only exemplary. In practical applications, the missing feature representation can indicate the feature items missing the corresponding features in any appropriate way, and the embodiments of the present disclosure do not limit the specific data structure of the missing feature representation.

[0059] In block 240, the server device 130 determines a first prediction result of the recommendation task at least based on the second feature representation, using a trained prediction model. The prediction model can be trained to determine the prediction result of the recommendation task based on the feature representation corresponding to the input. In some embodiments, the prediction model can be implemented based on any appropriate model network that matches the recommendation task. For example, the prediction model can be implemented based on a generative model, or the prediction model can be implemented based on a classification model. Of course, the prediction model can also be implemented based on any other appropriate model network.

[0060] In some embodiments, the first prediction result indicates the degree of association between at least one first object and at least one second object. The server device 130 may determine whether the degree of association exceeds a threshold degree of association. If the degree of association between the first object and at least one second object exceeds the threshold degree of association, the server device 130 may determine that the first object is suitable for recommendation to the corresponding at least one second object. If the degree of association between the first object and at least one second object does not exceed the threshold degree of association, the server device 130 may determine that the first object is not suitable for recommendation to the corresponding at least one second object.

[0061] In some examples, the first prediction result may include a correlation probability or a correlation coefficient between at least one recommended content and at least one recommended audience, indicating the degree of association between the recommended content and the recommended audience through the correlation probability or the correlation coefficient. The application 120 may present the recommended content to the recommended audience based on the first prediction result.

[0062] As an example, the first prediction result may include multiple correlation probabilities, and each correlation probability may indicate the degree of association between the corresponding recommended content and the corresponding recommended audience. The server device 130 may determine whether the multiple correlation probabilities exceed a probability threshold. If the correlation probability exceeds the probability threshold, the server device 130 may push the corresponding recommended content to the terminal device 110 associated with the corresponding recommended audience. The terminal device 110 may present the corresponding recommended content through the user interface 150 of the application 120, such as a page, audio, video, image, etc. In an actual recommendation system, there often occurs a situation where a feature item related to the recommended content or a feature item related to the recommended audience cannot collect the corresponding feature. Using the decoder model can complete the feature representation corresponding to at least some of the missing features, can obtain a relatively complete feature representation, and further can accurately predict the degree of association between the recommended content and the recommended audience. On this basis, the recommended content that matches the recommended audience can be accurately recommended, which is beneficial to improving the quality of content recommendation.

[0063] As another example, the first prediction result may include multiple correlation coefficients, and each correlation coefficient may indicate the degree of association between the corresponding recommended content and the corresponding recommended audience. The server device 130 may sort the multiple correlation coefficients and select a predetermined number of correlation coefficients with higher rankings from the sorting. The server device 130 may send the matching recommended content to the terminal device 110 associated with each recommended audience based on the association relationship between the recommended content and the recommended audience indicated by the predetermined number of correlation coefficients. The terminal device 110 may use the application 120 to present the corresponding recommended content. Of course, the above recommendation strategies are only exemplary. In actual applications, any appropriate recommendation strategy may be selected according to actual needs to push the recommended content to the recommended audience. The embodiments of the present disclosure do not limit this.

[0064] In some embodiments, if process 200 is used to perform other tasks, the first prediction result may also include other content. In some examples, if process 200 is used to perform a feature attribute determination task, the first prediction result may also indicate the feature attributes of the object associated with the task (such as the quality score of the content, the recommendation coefficient of the commodity, etc.). In some examples, if process 200 is used to perform a classification task, the first prediction result may include the classification labels of multiple classification objects. As an example, the first prediction result may include the classification labels of multiple contents of application 120. The server device 130 may match each content to different sections of application 120 based on the classification labels, or the server device 130 may also push recommended content to different recommended audiences based on the classification labels. Using the decoder model can, to a certain extent, play a role in complementing features, be able to obtain a relatively complete feature representation, is beneficial to improving the accuracy of the classification labels, and thus can accurately allocate each content of application 120, which is beneficial to improving the content provision quality of application 120.

[0065] It can be understood that the above first prediction result is only exemplary. The first prediction result is matched with the task, and in different task situations, the specific content of the first prediction result may also be different. The embodiments of the present disclosure do not specifically limit the content of the first prediction result.

[0066] In some embodiments, as Figure 3 shown, at block 314, the server device 130 may concatenate the first feature representation 306 and the second feature representation 312 to obtain a third feature representation. After that, the server device 130 may use the prediction model 316 based on the third feature representation to determine the first prediction result 318 of the recommendation task. In this way, it is beneficial to improve the accuracy of the first prediction result.

[0067] The above introduces the specific process for task result prediction in the embodiments of the present disclosure. If some features are missing in the input, using the decoder to decode the first feature representation to obtain the second feature representation can, to a certain extent, play a role in supplementing the missing features and can improve the representation ability of the input. Using the prediction model to determine the prediction result of the recommendation task based on the supplemented second feature representation can improve the accuracy of the prediction result of the recommendation task.

[0068] The training processes of the encoder model, the decoder model, and the prediction model will be introduced below. It should be understood that such training processes can be executed by an appropriate training system. The training system may include, but is not limited to, the server device 130.

[0069] In some embodiments of the present disclosure, the training system may determine a first sample input and a second sample input for a sample recommendation task. The first sample input includes a plurality of sample features corresponding to a plurality of sample feature items of a sample object associated with the sample recommendation task, and the second sample input includes partial sample features corresponding to the plurality of sample feature items.

[0070] In some embodiments, the training system may obtain a first sample input for a sample recommendation task. Subsequently, the training system may remove the sample features corresponding to one or more sample feature items from the first sample input to obtain a second sample input. The input of the recommendation task includes a plurality of feature items, and this input is in a discrete data space. By removing the sample features corresponding to one or more sample feature items from the first sample input, it is possible to simulate the discrete feature missing manner in the real input, making the obtained second sample input closer to the input in the real application scenario, so as to improve the training effect of the model. In other words, the process of discarding the sample noise corresponding to one or more sample feature items from the first sample input can also be regarded as performing a feature masking operation on one or more sample feature items in the first sample input to simulate the feature missing manner in the real discrete feature space, and a better training effect can be achieved.

[0071] In some embodiments, the training system may determine a time step parameter corresponding to the first sample input. Based on the time step parameter, determine the number of sample feature items to be removed from the first sample input. Subsequently, the training system may remove the sample features corresponding to the number of sample feature items from the first sample input to obtain a second sample input. In some cases, the operation of removing sample features from the first sample may also be referred to as Feature Dropout. Removing sample features from the first sample input based on the time step parameter can simulate the discrete feature loss process in the real scenario.

[0072] In some examples, Figure 4 FIG. shows a schematic diagram of an example architecture 400 for task result prediction according to some embodiments of the present disclosure. As Figure 4 shown, the decoder model 310 may be a decoder model based on a diffusion model. The training system may pre-define to remove the sample features corresponding to one or more sample feature items with a uniform probability at each time step during the forward diffusion process of the diffusion model. Assume that the first sample input x0 (which may also be referred to as the original sample) contains N sample feature items, and the training system may randomly sample a time step parameter T within the value range (0, N) of a uniform distribution. Subsequently, the training system may determine the number of sample feature items to be removed from the first sample input x0 based on the time step parameter T.

[0073] For example, assume that the training system pre - defines removing the sample features corresponding to one sample feature item at each time step in the forward diffusion process of the diffusion model. If the value of the time - step parameter T is 1, the training system can remove the sample features of one sample feature item from the first sample input x0 to obtain the second sample input x1. If the value of the time - step parameter T is 2, the training system can remove the sample features of two sample feature items from the first sample input x0 to obtain the second sample input x2. If the value of the time - step parameter T is N - 1, the training system can remove the sample features of N - 1 sample feature items from the first sample input x0 to obtain the second sample input x N-1 . It should be understood that during the actual training process, for each first sample input x0, one or more time - step parameters T can be randomly determined. Correspondingly, one or more second sample inputs can be obtained, and the one or more second sample inputs may include any one or more of {x1, …, x N-1}. In some cases, the second sample input can also be represented as x T , x T ∈ {x1, …, x N-1}.

[0074] In some embodiments of the present disclosure, the training system can use an encoder model to encode the first sample input to obtain a first sample feature representation, and the first sample feature representation represents multiple sample features corresponding to multiple sample feature items. In some examples, as Figure 4 shown, the training system can use the encoder model 304 to encode the first sample input x0 to obtain the first sample feature representation z0.

[0075] In some embodiments of the present disclosure, the training system can use an encoder model to encode the second sample input to obtain a second sample feature representation, and the second sample feature representation represents partial sample features corresponding to multiple sample feature items. In some examples, as Figure 4 shown, the training system can use the decoder model 310 to decode the second sample input x T respectively to obtain the second sample feature representation. For example, one or more of the second sample feature representation z1, the second sample feature representation z2, …, the second sample feature representation z N-1 can be obtained. In some cases, the second sample feature representation can be represented as z T ∈ {z1, …, z N-1}.

[0076] In some embodiments of the present disclosure, the training system can use a decoder model to decode the second sample feature representation to obtain a third sample feature representation. The third sample feature representation represents multiple sample features corresponding to multiple sample feature items. In some examples, as Figure 4As shown, the training system can use the decoder model 310 to perform decoding on the second sample feature representation z based on the reverse diffusion process of the diffusion model T to obtain a third sample feature representation. For example, the third sample feature representations z′1, z′2, …, z′ N-1 can be obtained. In some cases, the third sample feature representation can be denoted as z′ T ∈{z′1, z′2, …, z′ N-1}. Alternatively, the third sample feature representation can also be uniformly denoted as z′0.

[0077] In some embodiments, the training system can determine the missing feature representation s of the second sample input x T . The missing feature representation can indicate the sample feature items in the multiple sample feature items of the second sample input x T that are missing the corresponding features. The training system can perform decoding on the corresponding second sample feature representation z T based on the missing feature representation s using the decoder model to obtain the third sample feature representation z′ T . If the decoder model 310 is denoted as g(·), then the third sample feature representation can be denoted as z′ T = g([s, z T ).

[0078] In some embodiments, the decoder model can be implemented based on the diffusion model. The first sample feature representation, the second sample feature representation, and the third sample feature representation can be located in the latent feature space of the prediction model. Specifically, the training system performs a forward diffusion process on the original sample input in the discrete data space of the original sample input. Then the training system performs a reverse diffusion process in an asymmetric manner (which can also be called asymmetric diffusion), and uses the decoder model based on the diffusion model to perform a reverse diffusion process in the latent feature space to reconstruct the latent feature representation of the original input, which can be regarded as a feature completion process. In this way, it is beneficial to reduce information loss and beneficial to retain the information in the latent feature representation (i.e., the third sample feature representation) to improve the information integrity in the latent feature representation. In some cases, the encoder model, the decoder model based on the diffusion model, and the prediction model can be applied to a recommendation system, and the encoder model, the decoder model, and the prediction model as a whole can also be called a recommendation model based on asymmetric diffusion.

[0079] In some embodiments of the present disclosure, the training system adjusts the parameters of the encoder model and the parameters of the decoder model based at least on a first difference between a first sample feature representation and a third sample feature representation to obtain a trained encoder model and a trained decoder model. In some examples, the training system may adjust the parameters of the encoder model and the parameters of the decoder model based on a first difference between one or more third sample feature representations and the first sample feature representation. As an example, the training system may adjust the parameters of the encoder model and the parameters of the decoder model based on the first difference using a reconstruction loss function as shown below:

[0080]

[0081] In some embodiments, the training system may obtain a sample task result corresponding to a first sample input in a sample recommendation task. The training system determines a second prediction result of the sample recommendation task using a prediction model based at least on the third sample feature representation. Thereafter, the training system may adjust the parameters of the encoder model and the parameters of the decoder model based on the first difference and a second difference between the sample task result and the second prediction result. The sample task result may also be referred to as a ground-truth task result, and the sample task result indicates the ground-truth corresponding to the respective first sample input.

[0082] As an example, as Figure 4 shown, the training system may obtain a sample task result corresponding to each first sample input x0. Each first sample input x0 may have a corresponding label y ∈ {0, 1}. For example, a label of y = 1 may indicate that the recommended object is associated or matched with the recommended audience, and a label of y = 0 may indicate that the recommended object is not relevant or not matched with the recommended audience. Assuming that the prediction model 316 is represented as f(·), the training system may also provide the third sample feature representation z′ T to the prediction model 316 to obtain a second prediction result f(z′ T ), for example, may obtain any one of the second prediction result f(z′1), the second prediction result f(z′2), …, the second prediction result f(z′ N-1 ).

[0083] The training system may adjust the parameters of the encoder model and the parameters of the decoder model using the reconstruction loss function shown in formula (1) and an auxiliary task-oriented loss function as shown below:

[0084]

[0085] In some embodiments, the training system may determine a third prediction result of the sample recommendation task based on the second sample feature representation and using the prediction model. Subsequently, the training system may adjust the parameters of the encoder model and the parameters of the decoder model based on the first difference and a third difference between the sample recommendation task result and the third prediction result.

[0086] As an example, as Figure 4 shown, the training system may also provide one or more second sample feature representations z T to the prediction model 316 to obtain a third prediction result. If the prediction model is represented as f(·), the third prediction result may be represented as The training system may adjust the parameters of the encoder model and the parameters of the decoder model using the reconstruction loss function shown in formula (1) and the cross-entropy loss function shown below:

[0087]

[0088] It can be understood that the training system may also adjust the parameters of the encoder model and the parameters of the decoder model based on and

[0089] In some embodiments, the prediction model may be jointly trained with the encoder model and the decoder model. That is, the training system may jointly train the encoder model, the decoder model, and the pre-trained model. As an example, the training system may also adjust the parameters of the encoder model, the parameters of the decoder model, and the parameters of the pre-trained model based on and to jointly train the encoder model, the decoder model, and the pre-trained model.

[0090] In this way, in the embodiments of the present disclosure, by performing feature dropout on the original sample input (i.e., the first sample input), the discrete feature loss process in the real scenario can be simulated, and the obtained second sample input is closer to the real sample in the discrete data space. By performing the forward diffusion process in the discrete data space and the asymmetric reverse diffusion process in the latent feature space, information loss can be reduced, which is beneficial to retaining the personalized information in the latent feature representation, can form a relatively complete and robust feature representation, and can enhance the representation learning ability of the decoder model.

[0091] Example device and equipment

[0092] Embodiments of the present disclosure also provide corresponding apparatuses for implementing the above methods or processes. Figure 5FIG. 0 shows a schematic structural block diagram of an example apparatus 500 for task result prediction according to certain embodiments of the present disclosure. The apparatus 500 may be implemented as or included in a server device 130. Each module / component in the apparatus 500 may be implemented by hardware, software, firmware, or any combination thereof.

[0093] As Figure 5 shown, the apparatus 500 includes: an acquisition module 510 configured to acquire an input of a recommendation task, the input including at least one feature corresponding to a plurality of feature items of a plurality of objects associated with the recommendation task; an encoding module 520 configured to encode the input by using a trained encoder model to obtain a first feature representation, the first feature representation characterizing at least one feature corresponding to the plurality of feature items; a decoding module 530 configured to decode the first feature representation by using a trained decoder model to obtain a second feature representation, the second feature representation characterizing a plurality of features corresponding to the plurality of feature items; and a determination module 540 configured to determine a first prediction result of the recommendation task by using a trained prediction model based at least on the second feature representation, the first prediction result indicating whether a first object among the plurality of objects is to be recommended to at least one second object among the plurality of objects.

[0094] In some embodiments, the determination module 540 is further configured to: concatenate the first feature representation and the second feature representation to obtain a third feature representation; and determine the first prediction result of the recommendation task based on the third feature representation by using the prediction model.

[0095] In some embodiments, the decoding module 530 is further configured to: determine a missing feature representation corresponding to the input, the missing feature representation indicating a feature item among the plurality of feature items that is missing a corresponding feature; and decode the first feature representation by using the decoder model based on the missing feature representation to obtain the second feature representation.

[0096] In some embodiments, the first prediction result includes a degree of association between at least one first object and at least one second object among the plurality of objects, and wherein in the case where the degree of association exceeds a threshold degree of association, the first object is to be recommended to at least one second object.

[0097] In some embodiments, the apparatus 500 may further include a training module configured to train the encoder model and the decoder model by: determining a first sample input and a second sample input for a sample recommendation task, the first sample input including a plurality of sample features corresponding to a plurality of sample feature items of a sample object associated with the sample recommendation task, and the second sample input including partial sample features corresponding to the plurality of sample feature items; encoding the first sample input using the encoder model to obtain a first sample feature representation, the first sample feature representation characterizing the plurality of sample features corresponding to the plurality of sample feature items; encoding the second sample input using the encoder model to obtain a second sample feature representation, the second sample feature representation characterizing the partial sample features corresponding to the plurality of sample feature items; decoding the second sample feature representation using the decoder model to obtain a third sample feature representation, the third sample feature representation characterizing the plurality of sample features corresponding to the plurality of sample feature items; and adjusting parameters of the encoder model and parameters of the decoder model at least based on a first difference between the first sample feature representation and the third sample feature representation to obtain a trained encoder model and a trained decoder model.

[0098] In some embodiments, the training module is further configured to: remove sample features corresponding to one or more sample feature items from the first sample input to obtain the second sample input.

[0099] In some embodiments, the training module is further configured to: determine a time step parameter corresponding to the first sample input; based on the time step parameter, determine the number of sample feature items to be removed from the first sample input; and remove sample features corresponding to the number of sample feature items from the first sample input to obtain the second sample input.

[0100] In some embodiments, the training module is further configured to: obtain a sample task result corresponding to the first sample input in the sample recommendation task; determine a second prediction result of the sample recommendation task using a prediction model at least based on the third sample feature representation; and adjust parameters of the encoder model and parameters of the decoder model based on the first difference and a second difference between the sample task result and the second prediction result.

[0101] In some embodiments, the training module is further configured to: determine a third prediction result of the sample recommendation task using a prediction model based on the second sample feature representation; and adjust parameters of the encoder model and parameters of the decoder model based on the first difference and a third difference between the sample recommendation task result and the third prediction result.

[0102] In some embodiments, the prediction model is jointly trained with the encoder model and the decoder model.

[0103] In some embodiments, the decoder model performs decoding based on the reverse diffusion process corresponding to the diffusion model.

[0104] The units and / or modules included in apparatus 500 can be implemented in various ways, including software, hardware, firmware, or any combination thereof. In some embodiments, one or more units and / or modules can be implemented using software and / or firmware, such as machine-executable instructions stored on a storage medium. In addition to or in place of the machine-executable instructions, some or all of the units and / or modules in apparatus 500 can be implemented at least partially by one or more hardware logic components. By way of example and not limitation, exemplary types of hardware logic components that can be used include field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system on a chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0105] Figure 6 A block diagram of an electronic device 600 in which one or more embodiments of the present disclosure can be implemented is shown. It should be understood that Figure 6 The electronic device 600 shown is merely exemplary and should not impose any limitation on the functionality and scope of the embodiments described herein. Figure 6 The electronic device 600 shown can include or be implemented as Figure 1 a server device 130 of Figure 5 the apparatus 500 of

[0106] As Figure 6 shown, the electronic device 600 is in the form of a general-purpose electronic device. The components of the electronic device 600 can include, but are not limited to, one or more processors 610, a memory 620, a storage device 630, one or more communication units 640, one or more input devices 650, and one or more output devices 660. The processor 610 can be an actual or virtual processor and can perform various processes according to executable instructions stored in the memory 620. In a multi-processor system, multiple processors execute computer-executable instructions in parallel to improve the parallel processing ability of the electronic device 600.

[0107] The electronic device 600 generally includes multiple computer storage media. Such media can be any accessible media to which the electronic device 600 can gain access, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 620 can be volatile memory (such as registers, caches, random access memory (RAM)), non-volatile memory (such as read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. The storage device 630 can be removable or non-removable media and can include machine-readable media, such as flash drives, magnetic disks, or any other media that can be capable of storing information and / or data and can be accessed within the electronic device 600.

[0108] The electronic device 600 can further include additional removable / non-removable, volatile / non-volatile storage media. Although not shown in Figure 6 it, a disk drive for reading from or writing to a removable, non-volatile magnetic disk (such as a "floppy disk") and an optical disk drive for reading from or writing to a removable, non-volatile optical disk can be provided. In these cases, each drive can be connected to a bus (not shown) by one or more data media interfaces. The memory 620 can include a computer program product 625 having one or more executable instruction modules that are configured to perform the various methods or actions of the various embodiments of the present disclosure.

[0109] The communication unit 640 enables communication with other electronic devices via a communication medium. Additionally, the functions of the components of the electronic device 600 can be implemented in a single computing cluster or multiple computer machines that are capable of communicating via a communication connection. Thus, the electronic device 600 can operate in a networked environment using a logical connection with one or more other servers, network personal computers (PCs), or another network node.

[0110] The input device 650 can be one or more input devices, such as a mouse, keyboard, trackball, etc. The output device 660 can be one or more output devices, such as a display, speaker, printer, etc. The electronic device 600 can also communicate with one or more external devices (not shown) as needed via the communication unit 640, such as storage devices, display devices, etc., communicate with one or more devices that enable a user to interact with the electronic device 600, or communicate with any device that enables the electronic device 600 to communicate with one or more other electronic devices (such as a network card, modem, etc.). Such communication can be performed via an input / output (I / O) interface (not shown).

[0111] According to an exemplary implementation of the present disclosure, a computer-readable storage medium is provided, on which computer-executable instructions are stored, and the computer-executable instructions are executed by a processor to implement the method described above. According to an exemplary implementation of the present disclosure, a computer-executable instruction product is also provided, the computer-executable instruction product being tangibly stored on a non-transitory computer-readable medium and including computer-executable instructions, and the computer-executable instructions being executed by a processor to implement the method described above.

[0112] Aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of methods, apparatuses, devices, and computer-executable instruction products implemented according to the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and the combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer-readable executable instructions.

[0113] These computer-executable instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine such that when these instructions are executed by the processor of the computer or other programmable data processing device, a device is produced that implements the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-executable instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other devices to work in a specific manner. Thus, the computer-readable medium storing the instructions includes a manufactured article that includes instructions for implementing various aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0114] The computer-executable instructions can be loaded onto a computer, other programmable data processing device, or other device, such that a series of operating steps are executed on the computer, other programmable data processing device, or other device to produce a computer-implemented process, so that the instructions executed on the computer, other programmable data processing device, or other device implement the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0115] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer-executable instruction products according to various implementations of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, an executable instruction, or a portion of an instruction that contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions noted in the blocks may occur in a different order than noted in the accompanying drawings. For example, two consecutive blocks may in fact be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified functions or actions, or by a combination of dedicated hardware and computer instructions.

[0116] The various implementations of the present disclosure have been described above. The above description is exemplary, not exhaustive, and is not limited to the disclosed implementations. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The choice of terms used herein is intended to best explain the principles of the implementations, the practical application, or improvements made to the technology in the marketplace, or to enable other ordinary skill in the art to understand the various implementations disclosed herein.

Claims

1. A method for predicting task results, comprising: Acquire an input of a recommendation task, the input comprising at least one feature corresponding to multiple feature items of multiple objects associated with the recommendation task; Encoding the input using the trained encoder model to obtain a first feature representation, wherein the first feature representation represents the at least one feature corresponding to the plurality of feature items; Decoding the first feature representation using a trained decoder model to obtain a second feature representation, wherein the second feature representation represents a plurality of features corresponding to the plurality of feature items; as well as Based at least on the second feature representation, a first prediction result of the recommendation task is determined using a trained prediction model, the first prediction result indicating whether a first object among the multiple objects is to be recommended to at least one second object among the multiple objects.

2. The method according to claim 1, wherein determining the first prediction result of the recommended task comprises: Cascading the first feature representation and the second feature representation to obtain a third feature representation; as well as Based on the third feature representation, the first prediction result of the recommended task is determined using the prediction model.

3. The method of claim 1, wherein decoding the first feature representation using a trained decoder model to obtain the second feature representation comprises: Determine a missing feature representation corresponding to the input, the missing feature representation indicating a feature item in the plurality of feature items that lacks a corresponding feature; as well as Based on the missing feature representation, the first feature representation is decoded using the decoder model to obtain the second feature representation.

4. The method according to claim 1, wherein the first prediction result includes a degree of association between at least one first object among the multiple objects and the at least one second object, and wherein when the degree of association exceeds a threshold degree of association, the first object is to be recommended to the at least one second object.

5. The method of claim 1, wherein the encoder model and the decoder model are trained by: Determine a first sample input and a second sample input for a sample recommendation task, wherein the first sample input includes a plurality of sample features corresponding to a plurality of sample feature items of a sample object associated with the sample recommendation task, and the second sample input includes a portion of sample features corresponding to the plurality of sample feature items; Encoding the first sample input using the encoder model to obtain a first sample feature representation, wherein the first sample feature representation represents the plurality of sample features corresponding to the plurality of sample feature items; Encoding the second sample input using the encoder model to obtain a second sample feature representation, wherein the second sample feature representation represents the partial sample features corresponding to the multiple sample feature items; Decoding the second sample feature representation using the decoder model to obtain a third sample feature representation, wherein the third sample feature representation represents the plurality of sample features corresponding to the plurality of sample feature items; and Based at least on a first difference between the first sample feature representation and the third sample feature representation, parameters of the encoder model and parameters of the decoder model are adjusted to obtain the trained encoder model and the decoder model.

6. The method of claim 5, wherein the second sample input is determined by: The sample features corresponding to one or more sample feature items are removed from the first sample input to obtain the second sample input.

7. The method according to claim 6, wherein removing sample features corresponding to one or more sample feature items from the first sample input to obtain the second sample input comprises: determining a time step parameter corresponding to the first sample input; Determining a number of sample feature items to be removed from the first sample input based on the time step parameter; as well as The sample features corresponding to the number of sample feature items are removed from the first sample input to obtain the second sample input.

8. The method of claim 5, wherein adjusting parameters of the encoder model and parameters of the decoder model comprises: Obtaining a sample task result corresponding to the first sample input in the sample recommendation task; Determining a second prediction result of the sample recommendation task using the prediction model based at least on the third sample feature representation; as well as Based on the first difference and a second difference between the sample task result and the second prediction result, parameters of the encoder model and parameters of the decoder model are adjusted.

9. The method of claim 5, wherein adjusting parameters of the encoder model and parameters of the decoder model comprises: Based on the second sample feature representation, using the prediction model, determining a third prediction result of the sample recommendation task; as well as Based on the first difference and a third difference between the sample task result and the third prediction result, parameters of the encoder model and parameters of the decoder model are adjusted.

10. The method of claim 5, wherein the prediction model is jointly trained with the encoder model and the decoder model.

11. The method according to claim 1, wherein the decoder model performs decoding based on a back-diffusion process corresponding to a diffusion model.

12. A device for predicting task results, comprising: an acquisition module configured to acquire an input of a recommendation task, wherein the input includes at least one feature corresponding to multiple feature items of multiple objects associated with the recommendation task; an encoding module, configured to encode the input using a trained encoder model to obtain a first feature representation, wherein the first feature representation represents the at least one feature corresponding to the plurality of feature items; A decoding module, configured to decode the first feature representation using a trained decoder model to obtain a second feature representation, wherein the second feature representation represents a plurality of features corresponding to the plurality of feature items; as well as A determination module is configured to determine a first prediction result of the recommendation task using a trained prediction model based at least on the second feature representation, wherein the first prediction result indicates whether a first object among the multiple objects is to be recommended to at least one second object among the multiple objects.

13. An electronic device comprising: at least one processor; as well as At least one memory, the at least one memory is coupled to the at least one processor and stores instructions for execution by the at least one processor, the instructions, when executed by the at least one processor, causing the electronic device to perform the method according to any one of claims 1 to 11.

14. A computer-readable storage medium having computer-executable instructions stored thereon, wherein the computer-executable instructions can be executed by a processor to implement the method according to any one of claims 1 to 11.

15. A computer program product comprising computer executable instructions, wherein the computer executable instructions, when executed by a processor, implement the method according to any one of claims 1 to 11.

Images