Recommendation information generation method, apparatus, device, medium, and program product

Abstract

Embodiments of the present disclosure disclose a recommendation information generation method, device, equipment, medium and program product. A specific implementation of the method includes: obtaining a historical creative image sequence, a user sparse feature information set, a main image and an item sparse feature information set; performing graph encoding processing on the main image to obtain a main graph encoding vector, and performing graph encoding processing on each historical creative image to generate a historical image encoding vector; performing information encoding on each user sparse feature information to generate a user sparse feature vector, and performing information encoding on each item sparse feature information to generate an item sparse feature vector; adjusting visual preferences of each historical image encoding vector to obtain an adjusted historical image encoding vector sequence; and generating a recommendation image set corresponding to a target recommendation item to be pushed to a target user. The implementation is related to artificial intelligence, generates accurate recommendation information, and achieves good recommendation effect.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of computer technology, and more specifically to recommendation information generation methods, apparatus, devices, media, and program products. Background Technology

[0002] Currently, products often have multiple pre-designed materials (e.g., main product image, creative images, etc.). To select materials that the target user might like from these multiple materials, the common approach is as follows: First, input the multiple materials into a pre-trained user material preference model to generate a score set for each material. Then, using the score set, select at least one material that the target user might like from the multiple materials. Finally, recommend at least one material to the target user.

[0003] However, the inventors discovered that when using the above method to filter out materials that the target user might like from multiple materials, the following technical problems often arise:

[0004] Training a user content preference model solely using historical creative image sequences viewed by the target user limits the feature information it can learn, resulting in inaccurate modeling and consequently poor recommendation performance.

[0005] The information disclosed in this background section is only intended to enhance the understanding of the background of the inventive concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0007] Some embodiments of this disclosure provide methods, apparatus, devices, media, and program products for generating recommendation information to address the technical problems mentioned in the background section above.

[0008] In a first aspect, some embodiments of this disclosure provide a method for generating recommendation information, including: acquiring a sequence of historical browsing creative images of a target user, a user sparse feature information set of the target user, a main image of a target recommended item, and an item sparse feature information set of the target recommended item, wherein the item sparse feature information set includes a set of creative image feature information for the target recommended item; performing graph encoding processing on the main image to obtain a main image encoding vector, and performing graph encoding processing on each historical browsing creative image in the sequence of historical browsing creative images to generate a historical image encoding vector, thereby obtaining a sequence of historical image encoding vectors; and performing graph encoding processing on each user sparse feature information set of the target recommended item. Feature information is encoded to generate user sparse feature vectors, resulting in a user sparse feature vector set. Similarly, each item sparse feature information in the aforementioned item sparse feature information set is encoded to generate an item sparse feature vector, resulting in an item sparse feature vector set. Visual preference adjustments are made to each historical image encoding vector in the aforementioned historical image encoding vector sequence, resulting in an adjusted historical image encoding vector sequence. Based on the adjusted historical image encoding vector sequence, the aforementioned main image encoding vector, the aforementioned user sparse feature vector set, and the aforementioned item sparse feature vector set, a multi-head attention mechanism model is used to generate a set of recommended images corresponding to the aforementioned target recommended items to be pushed to the aforementioned target user.

[0009] Optionally, the above-mentioned adjustment of visual preferences for each historical image encoding vector in the above-mentioned historical image encoding vector sequence to obtain an adjusted historical image encoding vector sequence includes: determining the visual preference information corresponding to the above-mentioned historical image encoding vector sequence as target visual preference information; and adjusting each historical image encoding vector in the above-mentioned historical image encoding vector sequence according to the target visual preference information to obtain an adjusted historical image encoding vector sequence.

[0010] Optionally, the above-mentioned generation of a recommended image set corresponding to the target recommended item to be pushed to the target user, based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set, using a multi-head attention mechanism model, includes: concatenating the user sparse feature vector set and the item sparse feature vector set to obtain a concatenated sparse feature vector; inputting the adjusted historical image encoding vector sequence, the main image encoding vector, and the concatenated sparse feature vector into the multi-head attention mechanism model to obtain a feature information fusion vector; inputting the concatenated sparse feature vector into a fully connected model to obtain an output vector; concatenating the feature information fusion vector and the output vector to obtain a concatenated vector; and generating the recommended image set using a preset loss function.

[0011] Optionally, the above-mentioned graph coding processing of each historical browsing creative image in the above-mentioned historical browsing creative image sequence to generate a historical image coding vector includes: inputting the above-mentioned historical browsing creative image into a pre-trained graph coding model to generate a historical image coding vector.

[0012] Optionally, the graph coding model includes a residual network model and multiple fully connected layers; and the above-mentioned inputting the above-mentioned historical browsing creative images into the pre-trained graph coding model to generate historical image coding vectors includes: inputting the above-mentioned historical browsing creative images into the above-mentioned residual network model to obtain the model output result; and inputting the above-mentioned model output result into the above-mentioned multiple fully connected layers to obtain the above-mentioned historical image coding vectors.

[0013] Optionally, generating the recommended image set using a preset loss function includes: generating a creative image score set for the spliced ​​vector using the preset loss function; generating a recommended item score corresponding to the target recommended item using a multi-head attention mechanism model based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set; in response to determining that the recommended item score is greater than a predetermined recommended item value, determining the creative image scores in the creative image score set that are greater than the predetermined recommended creative image value, and obtaining a subset of creative image scores; and determining the creative image set corresponding to the subset of creative image scores as the recommended image set.

[0014] Optionally, the above method further includes: generating a recommended main image score corresponding to the target recommended item using a multi-head attention mechanism model based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set; and pushing the main image to the terminal corresponding to the target user in response to determining that the recommended item score is greater than a predetermined recommended item value and the recommended main image score is greater than a predetermined recommended main image value.

[0015] Secondly, some embodiments of this disclosure provide a recommendation information generation apparatus, comprising: an acquisition unit configured to acquire a sequence of historical browsing creative images of a target user, a user sparse feature information set of the target user, a main image of a target recommended item, and an item sparse feature information set of the target recommended item, wherein the item sparse feature information set includes a set of creative image feature information for the target recommended item; a graph encoding unit configured to perform graph encoding processing on the main image to obtain a main image encoding vector, and to perform graph encoding processing on each historical browsing creative image in the sequence of historical browsing creative images to generate a historical image encoding vector, thereby obtaining a sequence of historical image encoding vectors; and an information encoding unit configured to encode information in the user sparse feature information set... Each user's sparse feature information is encoded to generate a user sparse feature vector, resulting in a user sparse feature vector set. Similarly, each item's sparse feature information in the aforementioned item sparse feature information set is encoded to generate an item sparse feature vector, resulting in an item sparse feature vector set. An adjustment unit is configured to adjust the visual preferences of each historical image encoding vector in the aforementioned historical image encoding vector sequence, resulting in an adjusted historical image encoding vector sequence. A generation unit is configured to generate a set of recommended images corresponding to the target recommended items to be pushed to the target user, based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set, using a multi-head attention mechanism model.

[0016] Optionally, the adjustment unit can be configured to: determine the visual preference information corresponding to the above-mentioned historical image coding vector sequence as the target visual preference information; and adjust each historical image coding vector in the above-mentioned historical image coding vector sequence according to the above-mentioned target visual preference information to obtain the adjusted historical image coding vector sequence.

[0017] Optionally, the generation unit can be configured to: concatenate the user sparse feature vector set and the item sparse feature vector set to obtain a concatenated sparse feature vector; input the adjusted historical image encoding vector sequence, the main image encoding vector, and the concatenated sparse feature vector into the multi-head attention mechanism model to obtain a feature information fusion vector; input the concatenated sparse feature vector into a fully connected model to obtain an output vector; concatenate the feature information fusion vector and the output vector to obtain a concatenated vector; and generate the recommended image set using a preset loss function.

[0018] Optionally, the graph coding unit can be configured to input the aforementioned historical browsing creative images into a pre-trained graph coding model to generate historical image coding vectors.

[0019] Optionally, the graph coding model includes a residual network model and multiple fully connected layers; and the graph coding unit can be configured to: input the above-mentioned historical browsing creative image into the above-mentioned residual network model to obtain the model output result; input the above-mentioned model output result into the above-mentioned multiple fully connected layers to obtain the above-mentioned historical image coding vector.

[0020] Optionally, the generation unit can be configured to: generate a set of creative image scores for the spliced ​​vector using the aforementioned preset loss function; generate a recommended item score corresponding to the target recommended item using a multi-head attention mechanism model based on the aforementioned adjusted historical image encoding vector sequence, the aforementioned main image encoding vector, the aforementioned user sparse feature vector set, and the aforementioned item sparse feature vector set; in response to determining that the aforementioned recommended item score is greater than a predetermined recommended item value, determine the creative image scores in the aforementioned creative image score set that are greater than the predetermined recommended creative image value, and obtain a subset of creative image scores; and determine the creative image set corresponding to the aforementioned subset of creative image scores as the aforementioned recommended image set.

[0021] Optionally, the device further includes: generating a recommended main image score corresponding to the target recommended item using a multi-head attention mechanism model based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set; and pushing the main image to the terminal corresponding to the target user in response to determining that the recommended item score is greater than a predetermined recommended item value and the recommended main image score is greater than a predetermined recommended main image value.

[0022] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, such that when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any implementation of the first aspect.

[0023] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method as described in any implementation of the first aspect.

[0024] Fifthly, some embodiments of this disclosure provide a computer program product, including a computer program that, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0025] The above embodiments of this disclosure have the following beneficial effects: the recommendation information generation method of some embodiments of this disclosure generates accurate recommendation information and achieves better recommendation results. Specifically, the reason for the poor recommendation results is that the user material preference model is trained only using the historical browsing creative image sequence of the target user, which limits the feature information that the user material preference model can learn, resulting in an inaccurate user material preference model. This indirectly leads to poor recommendation results. Based on this, the recommendation information generation method of some embodiments of this disclosure first obtains the historical browsing creative image sequence of the target user, the user sparse feature information set of the target user, the main image of the target recommended item, and the item sparse feature information set of the target recommended item, wherein the item sparse feature information set includes the creative image feature information set of the target recommended item. This is used to obtain more feature information in the future, which is convenient for generating more accurate recommendation information (i.e., multiple recommended creative image scores and recommended main image scores). Then, the main image is graph encoded to obtain a main image encoding vector, and each historical browsing creative image in the historical browsing creative image sequence is graph encoded to generate a historical image encoding vector, resulting in a historical image encoding vector sequence. Here, graph encoding is performed on the main image to extract its feature information. Furthermore, compared to generating recommendation information from the main image itself, using the main image encoding vector effectively solves the problem of high computational cost due to the large pixel dimension of the image. Similarly, compared to adjusting visual preferences using historical creative image sequences, adjusting visual preferences using historical image encoding vector sequences effectively solves the problem of high computational cost. Next, each user sparse feature in the aforementioned user sparse feature information set is encoded into vector form for easy use. Similarly, each item sparse feature in the aforementioned item sparse feature information set is encoded into vector form for easy use. Finally, visual preference adjustment is performed on each historical image encoding vector in the aforementioned historical image encoding vector sequence to make the adjusted historical image encoding vector sequence more clearly reflect visual preferences, which helps in generating more accurate recommendation information. Finally, by employing a multi-head attention mechanism model, multifaceted feature information can be extracted from the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set. Consequently, the generated recommendation information (i.e., the recommended image set) becomes more accurate. Attached Figure Description

[0026] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0027] Figures 1-2 This is a schematic diagram illustrating an application scenario of a recommendation information generation method according to some embodiments of the present disclosure;

[0028] Figure 3 This is a flowchart of some embodiments of the recommendation information generation method according to this disclosure;

[0029] Figure 4 These are flowcharts of some other embodiments of the recommendation information generation method according to this disclosure;

[0030] Figure 5 These are schematic diagrams illustrating the structure of some embodiments of the recommendation information generation apparatus according to this disclosure;

[0031] Figure 6 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0032] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0033] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0034] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0035] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0036] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0037] Before performing any of the operations involving the collection, storage, or use of creative materials (such as creative images and main images) disclosed in this disclosure, the relevant organizations or individuals shall fulfill their obligations, including conducting a security impact assessment of the materials, informing the material creator, and obtaining prior authorization and consent from the material creator.

[0038] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0039] Figures 1-2 This is a schematic diagram illustrating an application scenario of a recommendation information generation method according to some embodiments of the present disclosure.

[0040] exist Figures 1-2In this application scenario, firstly, the electronic device 101 can acquire the historical browsing creative image sequence 103 of the target user 102, the user sparse feature information set 104 of the target user 102, the main image 106 of the target recommended item 105, and the item sparse feature information set 107 of the target recommended item 105. The item sparse feature information set 107 includes the creative image feature information set for the target recommended item 105. In this application scenario, the target user 102 can be “Li**”. The historical browsing creative image sequence 103 can include: historical browsing creative image 1031 and historical browsing creative image 1032. The user sparse feature information set 104 can be: {“Gender: 1”, “Age: 18”, “Height: 184”}. The target recommended item 105 can be “Apple”. The item sparse feature information set 107 can be: {“Click-through rate: 0.4”, “Price: 5”, “Origin: 02”}. Here, gender “1” can represent male. Origin “02” can represent Shanghai. Then, the electronic device 101 can perform graph encoding processing on the main image 106 to obtain the main image encoding vector 110, and perform graph encoding processing on each historical browsing creative image in the historical browsing creative image sequence 103 to generate a historical image encoding vector, resulting in a historical image encoding vector sequence 108. In this application scenario, the historical image encoding vector sequence 108 may include: the historical image encoding vector 1081 corresponding to the historical browsing creative image 1031 and the historical image encoding vector 1082 corresponding to the historical browsing creative image 1032. Next, the electronic device 101 can perform information encoding on each user sparse feature information in the user sparse feature information set 104 to generate a user sparse feature vector, resulting in a user sparse feature vector set 109, and perform information encoding on each item sparse feature information in the item sparse feature information set 107 to generate an item sparse feature vector, resulting in an item sparse feature vector set 111. In this application scenario, the user sparse feature vector set 109 may include: user sparse feature vector 1091 corresponding to "Gender: 1", user sparse feature vector 1092 corresponding to "Age: 18", and user sparse feature vector 1093 corresponding to "Height: 184". The item sparse feature vector set 111 may include: item sparse feature vector 1111 corresponding to "Click-through rate: 0.4", item sparse feature vector 1112 corresponding to "Price: 5", and item sparse feature vector 1113 corresponding to "Place of origin: 02". Furthermore, the electronic device 101 can perform visual preference adjustment on each historical image encoding vector in the above-mentioned historical image encoding vector sequence 108 to obtain the adjusted historical image encoding vector sequence 112.In this application scenario, the adjusted historical image encoding vector sequence 112 includes: the adjusted historical image encoding vector 1121 corresponding to the historical image encoding vector 1081 and the adjusted historical image encoding vector 1122 corresponding to the historical image encoding vector 1082. Finally, the electronic device 101 can generate a recommended image set 114 corresponding to the target recommended item 105 to be pushed to the target user 102, based on the adjusted historical image encoding vector sequence 112, the main image encoding vector 110, the user sparse feature vector set 109, and the item sparse feature vector set 111, using a multi-head attention mechanism model 113. The recommended image set 114 is a subset of the images in the creative image set corresponding to the creative image feature information set. In this application scenario, the recommended image set 114 includes: recommended image 1141 and recommended image 1142.

[0041] It should be noted that the aforementioned electronic device 101 can be either hardware or software. When the electronic device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device. When the electronic device is software, it can be installed in the hardware devices listed above. It can be implemented as, for example, multiple software programs or software modules used to provide distributed services, or as a single software program or software module. No specific limitations are made here.

[0042] It should be understood that Figures 1-2 The number of electronic devices shown is merely illustrative. Any number of electronic devices can be used depending on the implementation requirements.

[0043] Continue to refer to Figure 3 The diagram illustrates a flow 300 of some embodiments of a recommendation information generation method according to the present disclosure. This recommendation information generation method includes the following steps:

[0044] Step 301: Obtain the target user's historical browsing creative image sequence, the target user's user sparse feature information set, the main image of the target recommended item, and the target recommended item's item sparse feature information set.

[0045] In some embodiments, the execution entity of the above-described recommendation information generation method (e.g.) Figure 1The electronic device 101 shown can acquire, via wired or wireless connection, a sequence of historically viewed creative images of the target user, a set of sparse user feature information of the target user, the main image of the target recommended item, and a set of sparse item feature information of the target recommended item. The set of sparse item feature information includes a set of creative image feature information for the target recommended item. The creative image feature information can be the ID identifier information of the creative image of the target recommended item. There is a one-to-one correspondence between the ID identifier information and the creative image. The target user can be the user of the material to be recommended. In practice, the material can be in the form of an image or a video. For video material, keyframes extracted from the video can be used as content recommended to the user. For e-commerce scenarios, the material can be the main image of the item or a creative image of the item. The creative image can be a pre-designed image based on the characteristics of the item. The sequence of historically viewed creative images can be a sequence of creative images viewed by the target user on various applications (Apps) within a historical time period. For example, the current time is April 2022. The historical browsing creative image sequence can be a sequence of creative images viewed by users from January 2021 to April 2021. The main image of the target recommended item can be the main image of the target recommended item. The sparse feature information of the target user can be the numerical information of the sparse features of the target user. The sparse features of the target user can be ID-type features of the target user. These ID-type features can be unique ID identifiers of user features or identifiers of user identity. For example, the sparse features of the target user can be one of the following: the target user's gender ID, the target user's age ID, the target user's national ID card ID, or the target user's height ID. For example, a gender ID of "1" indicates that the target user is male, and a gender ID of "0" indicates that the target user is female. The sparse feature information of the target recommended item can be the numerical information of the sparse features of the target recommended item. The sparse features of the target recommended item can be ID-type features of the target recommended item. ID-type features can be unique ID identifiers of item features or identifiers of item identity. For example, the sparse features of the target recommended item can be one of the following: the target recommended item's click-through rate, the target recommended item's price, or the target recommended item's place of origin. For example, the place of origin ID is "02", which indicates that the target recommended item is produced in Shanghai.

[0046] It should be noted that the information that can be recommended to the target user may include, but is not limited to, one of the following: creative images of the target recommended item, and the main image of the target recommended item.

[0047] Step 302: Perform graph encoding processing on the main image to obtain the main image encoding vector, and perform graph encoding processing on each historical browsing creative image in the historical browsing creative image sequence to generate a historical image encoding vector, thus obtaining a historical image encoding vector sequence.

[0048] In some embodiments, the execution entity may perform graph encoding processing on the main image to obtain a main image encoding vector, and perform graph encoding processing on each historical browsing creative image in the historical browsing creative image sequence to generate a historical image encoding vector, thus obtaining a historical image encoding vector sequence. The main image encoding vector can represent the image feature information of the main image. The historical image encoding vector can represent the image feature information of the historical browsing creative images.

[0049] As an example, firstly, the aforementioned execution entity can input the main image into multiple serially connected convolutional neural networks (CNNs) to obtain the output of the first model. Then, the output of the first model is input into the BERT encoding model to obtain the main image encoding vector.

[0050] Similarly, firstly, the aforementioned execution entity can input each historical creative image in the historical browsing creative image sequence into multiple serially connected convolutional neural networks to generate the second model output result, resulting in a second model output result sequence. Then, each second model output result in the second model output result sequence is input into the BERT encoding model to obtain a historical image encoding vector sequence.

[0051] Here, using the main image encoding vector to subsequently generate multiple recommended creative image scores and recommended main image scores is computationally less than using the main image itself. Furthermore, adjusting visual preferences based on historical image encoding vector sequences effectively addresses the issue of high computational cost compared to adjusting visual preferences through historical creative image sequences.

[0052] In some alternative implementations of certain embodiments, the aforementioned execution entity may input the aforementioned historical browsing creative images into a pre-trained graph coding model to generate historical image coding vectors.

[0053] Graph coding models can be models that encode creative images to generate encoded vectors. For example, a graph coding model can be multiple sequentially connected convolutional neural networks.

[0054] Optionally, the graph coding model described above includes a residual network model and multiple fully connected layers. The process of inputting the aforementioned historical browsing creative images into a pre-trained graph coding model to generate historical image coding vectors may include the following steps:

[0055] The first step is to input the aforementioned historical creative images into the aforementioned Residual Networks (ResNets) model to obtain the model output results.

[0056] The second step involves inputting the output of the model into the aforementioned fully connected layers to obtain the historical image encoding vector. These fully connected layers can be sequentially connected.

[0057] Step 303: Encode each user sparse feature information in the above user sparse feature information set to generate a user sparse feature vector, and obtain a user sparse feature vector set; and encode each item sparse feature information in the above item sparse feature information set to generate an item sparse feature vector, and obtain an item sparse feature vector set.

[0058] In some embodiments, the execution entity may encode each user sparse feature information in the user sparse feature information set to generate a user sparse feature vector, thus obtaining a user sparse feature vector set; and encode each item sparse feature information in the item sparse feature information set to generate an item sparse feature vector, thus obtaining an item sparse feature vector set. The user sparse feature vector represents the feature information of the user sparse feature information. The item sparse feature vector represents the feature information of the item sparse feature information.

[0059] As an example, the aforementioned execution entity can input each user sparse feature information from the user sparse feature information set into the BERT encoding model to generate a user sparse feature vector, thus obtaining a user sparse feature vector set. Similarly, the aforementioned execution entity can input each item sparse feature information from the aforementioned item sparse feature information set into the BERT encoding model to generate an item sparse feature vector, thus obtaining an item sparse feature vector set.

[0060] Step 304: Perform visual preference adjustment on each historical image encoding vector in the above historical image encoding vector sequence to obtain the adjusted historical image encoding vector sequence.

[0061] In some embodiments, the executing entity may perform visual preference adjustment on each historical image encoding vector in the aforementioned historical image encoding vector sequence to obtain an adjusted historical image encoding vector sequence. The visual preference information corresponding to each adjusted historical image encoding vector in the adjusted historical image encoding vector sequence is the same as the visual preference information corresponding to each historical image encoding vector in the overall historical image encoding vector sequence. The visual preference information can characterize the user's visual preference features. Specifically, the user's video preference features can be, but are not limited to, one of the following: a user's preference for aesthetically pleasing videos, or a user's preference for humorous videos. The visual preference information corresponding to each adjusted historical image encoding vector can be the user's visual preference features reflected in each adjusted historical image encoding vector. The number of historical image encoding vectors included in the aforementioned historical image encoding vector sequence is the same as the number of adjusted historical image encoding vectors included in the adjusted historical image encoding vector sequence.

[0062] As an example, the aforementioned execution entity can directly input each historical image encoding vector in the historical image encoding vector sequence into the Transformer model to generate an adjusted historical image encoding vector sequence.

[0063] It should be noted that the visual preference features reflected by each adjusted historical image coding vector in the adjusted historical image coding vector sequence are stronger than those reflected by each historical image coding vector in the historical image coding vector sequence.

[0064] In some optional implementations of certain embodiments, the above-mentioned visual preference adjustment of each historical image coding vector in the historical image coding vector sequence to obtain an adjusted historical image coding vector sequence may include the following steps:

[0065] The first step is to determine the visual preference information corresponding to the above-mentioned historical image encoding vector sequence, which will be used as the target visual preference information.

[0066] As an example, the aforementioned execution entity can input the aforementioned historical image encoded vector sequence into a Seq2Seq (Sequence to Sequence) model to output target visual preference information.

[0067] The second step is to adjust each historical image encoding vector in the above historical image encoding vector sequence according to the target visual preference information, so as to obtain the adjusted historical image encoding vector sequence.

[0068] As an example, the aforementioned execution entity can input the target visual preference information and each historical image encoding vector in the aforementioned historical image encoding vector sequence into the Generative Adversarial Network (GAN) model to obtain the adjusted historical image encoding vector sequence.

[0069] Step 305: Based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set, a multi-head attention mechanism model is used to generate a set of recommended images corresponding to the target recommended items to be pushed to the target user.

[0070] In some embodiments, the executing entity may, based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set, utilize a multi-head attention mechanism to generate a set of recommended images corresponding to the target recommended items to be pushed to the target user. The recommended image set is a subset of the creative image set corresponding to the creative image feature information set. The diverse feature information that the multi-head attention mechanism model can learn may include, but is not limited to, at least one of the following: vector relationships between various adjusted historical image encoding vectors in the adjusted historical image encoding vector sequence; vector relationships between various user sparse feature vectors in the user sparse feature vector set; vector relationships between various item sparse feature vectors in the item sparse feature vector set; vector relationships between the adjusted historical image encoding vector sequence and the main image encoding vector; vector relationships between the adjusted historical image encoding vector sequence and the adjusted historical image encoding vector sequence; vector relationships between the adjusted historical image encoding vector sequence and the item sparse feature vector set; vector relationships between the main image encoding vector and the user sparse feature vector set; vector relationships between the main image encoding vector and the item sparse feature vector set; and vector relationships between the user sparse feature vector set and the item sparse feature vector set.

[0071] As an example, the aforementioned execution entity can input the adjusted historical image encoding vector sequence, the aforementioned main image encoding vector, the aforementioned user sparse feature vector set, and the aforementioned item sparse feature vector set into the multi-head attention mechanism model to generate a set of recommended images corresponding to the aforementioned target recommended items to be pushed to the aforementioned target user.

[0072] The above embodiments of this disclosure have the following beneficial effects: the recommendation information generation method of some embodiments of this disclosure generates accurate recommendation information and achieves better recommendation results. Specifically, the reason for the poor recommendation results is that the user material preference model is trained only using the historical browsing creative image sequence of the target user, which limits the feature information that the user material preference model can learn, resulting in an inaccurate user material preference model. This indirectly leads to poor recommendation results. Based on this, the recommendation information generation method of some embodiments of this disclosure first obtains the historical browsing creative image sequence of the target user, the user sparse feature information set of the target user, the main image of the target recommended item, and the item sparse feature information set of the target recommended item, wherein the item sparse feature information set includes the creative image feature information set of the target recommended item. This is used to obtain more feature information in the future, which is convenient for generating more accurate recommendation information (i.e., multiple recommended creative image scores and recommended main image scores). Then, the main image is graph encoded to obtain a main image encoding vector, and each historical browsing creative image in the historical browsing creative image sequence is graph encoded to generate a historical image encoding vector, resulting in a historical image encoding vector sequence. Here, graph encoding is performed on the main image to extract its feature information. Furthermore, compared to generating recommendation information from the main image itself, using the main image encoding vector effectively solves the problem of high computational cost due to the large pixel dimension of the image. Similarly, compared to adjusting visual preferences using historical creative image sequences, adjusting visual preferences using historical image encoding vector sequences effectively solves the problem of high computational cost. Next, each user sparse feature in the aforementioned user sparse feature information set is encoded into vector form for easy use. Similarly, each item sparse feature in the aforementioned item sparse feature information set is encoded into vector form for easy use. Finally, visual preference adjustment is performed on each historical image encoding vector in the aforementioned historical image encoding vector sequence to make the adjusted historical image encoding vector sequence more clearly reflect visual preferences, which helps in generating more accurate recommendation information. Finally, by employing a multi-head attention mechanism model, multifaceted feature information can be extracted from the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set. Consequently, the generated recommendation information (i.e., the recommended image set) becomes more accurate.

[0073] Further reference Figure 4 The diagram illustrates a flow 400 of another embodiment of the recommendation information generation method according to the present disclosure. This recommendation information generation method includes the following steps:

[0074] Step 401: Obtain the target user's historical browsing creative image sequence, the target user's sparse feature information set, the main image of the target recommended item, and the target recommended item's sparse feature information set.

[0075] Step 402: Perform graph encoding processing on the main image to obtain the main image encoding vector, and perform graph encoding processing on each historical browsing creative image in the historical browsing creative image sequence to generate a historical image encoding vector, thus obtaining a historical image encoding vector sequence.

[0076] Step 403: Encode each user sparse feature information in the above user sparse feature information set to generate a user sparse feature vector, and obtain a user sparse feature vector set; and encode each item sparse feature information in the above item sparse feature information set to generate an item sparse feature vector, and obtain an item sparse feature vector set.

[0077] Step 404: Visual preference adjustment is performed on each historical image encoding vector in the above historical image encoding vector sequence to obtain the adjusted historical image encoding vector sequence.

[0078] In some embodiments, the specific implementation of steps 401-404 and the resulting technical effects can be found in [reference needed]. Figure 3 Steps 301-304 in the corresponding embodiments will not be repeated here.

[0079] Step 405: Concatenate the above user sparse feature vector set and the above item sparse feature vector set to obtain the concatenated sparse feature vector.

[0080] In some embodiments, the executing entity (e.g. Figure 1 The electronic device 101 shown can concatenate the above-mentioned user sparse feature vector set and the above-mentioned item sparse feature vector set to obtain a concatenated sparse feature vector.

[0081] Step 406: Input the adjusted historical image encoding vector sequence, the main image encoding vector, and the spliced ​​sparse feature vector into the multi-head attention mechanism model to obtain the feature information fusion vector.

[0082] In some embodiments, the execution entity may input the adjusted historical image encoding vector sequence, the main image encoding vector, and the concatenated sparse feature vector into the multi-head attention mechanism model to obtain a feature information fusion vector. The feature information fusion vector is obtained by fusing multiple feature information sources. Specifically, the feature information fusion vector may include, but is not limited to, at least one of the following: vector relationship information between the various historical image encoding vectors in the adjusted historical image encoding vector sequence; vector relationship information between the adjusted historical image encoding vector sequence and the main image encoding vector; vector relationship information between the concatenated sparse feature vector and the adjusted historical image encoding vector sequence; and vector relationship information between the concatenated sparse feature vector and the main image encoding vector.

[0083] In addition, the multi-head attention mechanism model can learn a variety of features, including: the importance of sparse feature information (i.e., the user sparse feature vector set and the item sparse feature vector set) and the importance of material content information (i.e., the adjusted historical image encoding vector sequence and the main image encoding vector).

[0084] Step 407: Input the above-mentioned concatenated sparse feature vector into the fully connected model to obtain the output vector.

[0085] In some embodiments, the execution entity can input the concatenated sparse feature vector into a fully connected model to obtain an output vector. The fully connected model may include multiple fully connected layers.

[0086] Step 408: Generate the recommended image set using a preset loss function. In some embodiments, the executing entity may use the preset loss function to generate the recommended image set in various ways. For example, the preset loss function may be a quadratic loss function.

[0087] In some optional implementations of certain embodiments, generating the recommended image set using a preset loss function may include the following steps:

[0088] The first step involves generating a set of creative image scores for the concatenated vector using the aforementioned preset loss function. Each creative image score represents the degree of interest the target user has in the creative images of the recommended items. There is a one-to-one correspondence between the creative image scores in the set of creative image scores and the creative images in the set of creative image scores.

[0089] As an example, the aforementioned execution entity can input the splicing vector into a preset loss function to obtain a creative image score set for the splicing vector.

[0090] The second step involves generating recommended item scores for the target user based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set, using a multi-head attention mechanism model. The recommended item score represents the target user's preference for the recommended items.

[0091] As an example, the aforementioned execution entity can input the adjusted historical image encoding vector sequence, the aforementioned main image encoding vector, the aforementioned user sparse feature vector set, and the aforementioned item sparse feature vector set into the multi-head attention mechanism model to generate the recommended item score corresponding to the aforementioned target user.

[0092] Third, in response to determining that the score of the recommended item is greater than the predetermined recommended item value, the executing entity can determine the creative image scores in the creative image score set that are greater than the predetermined recommended creative image value, and obtain a subset of creative image scores.

[0093] For example, the recommended item score is 75.

[0094] Fourth, the aforementioned implementing entity can determine the creative image set corresponding to the aforementioned creative image score subset as the aforementioned recommended image set.

[0095] Optionally, the steps also include:

[0096] The first step involves generating a recommended main image score for the target recommended item based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set. This score represents the target user's level of interest in the main image.

[0097] As an example, the aforementioned execution entity can input the adjusted historical image encoding vector sequence, the aforementioned main image encoding vector, the aforementioned user sparse feature vector set, and the aforementioned item sparse feature vector set into the multi-head attention mechanism model to generate the recommended main image score corresponding to the aforementioned target recommended item.

[0098] The second step involves, in response to determining that the score of the recommended item is greater than the predetermined value for recommended items and the score of the recommended main image is greater than the predetermined value for recommended main images, pushing the main image to the terminal corresponding to the target user. The terminal corresponding to the target user can be a display terminal.

[0099] For example, the value for the pre-order recommended item could be 70. The value for the pre-order recommended main image could be 75.

[0100] from Figure 4 It can be seen from this that, with Figure 3Compared to the description of some corresponding embodiments, Figure 4 In some corresponding embodiments, the recommendation information generation method process 400 utilizes a multi-head attention mechanism model to learn diverse feature information regarding the importance of sparse feature information (i.e., user sparse feature vector set and item sparse feature vector set) and the importance of material content information (i.e., the adjusted historical image encoding vector sequence and the main image encoding vector), thereby generating a more accurate creative image set.

[0101] Further reference Figure 5 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a recommendation information generation apparatus, which are similar to... Figure 3 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0102] like Figure 5 As shown, a recommendation information generation device 500 includes: an acquisition unit 501, an image encoding unit 502, an information encoding unit 503, an adjustment unit 504, and a generation unit 505. The acquisition unit 501 is configured to acquire a sequence of historical browsing creative images of a target user, a set of user sparse feature information of the target user, a main image of a target recommended item, and a set of item sparse feature information of the target recommended item, wherein the set of item sparse feature information includes a set of creative image feature information for the target recommended item. The image encoding unit 502 is configured to perform image encoding processing on the main image to obtain a main image encoding vector, and to perform image encoding processing on each historical browsing creative image in the sequence of historical browsing creative images to generate a historical image encoding vector, resulting in a sequence of historical image encoding vectors. The information encoding unit 503 is configured to process each user sparse feature information in the set of user sparse feature information. Information encoding is performed to generate user sparse feature vectors, resulting in a user sparse feature vector set. Information encoding is also performed on each item sparse feature information in the aforementioned item sparse feature information set to generate item sparse feature vectors, resulting in an item sparse feature vector set. Adjustment unit 504 is configured to perform visual preference adjustment on each historical image encoding vector in the aforementioned historical image encoding vector sequence, resulting in an adjusted historical image encoding vector sequence. Generation unit 505 is configured to generate a set of recommended images corresponding to the target recommended items to be pushed to the target user, using a multi-head attention mechanism model, based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set.

[0103] In some optional implementations of some embodiments, the adjustment unit 504 in the above-described device 500 may be further configured to: determine the visual preference information corresponding to the above-described historical image coding vector sequence as target visual preference information; and adjust each historical image coding vector in the above-described historical image coding vector sequence according to the above-described target visual preference information to obtain an adjusted historical image coding vector sequence.

[0104] In some optional implementations of certain embodiments, the generation unit 505 in the above-mentioned device 500 may be further configured to: concatenate the user sparse feature vector set and the item sparse feature vector set to obtain a concatenated sparse feature vector; input the adjusted historical image encoding vector sequence, the main image encoding vector, and the concatenated sparse feature vector into the multi-head attention mechanism model to obtain a feature information fusion vector; input the concatenated sparse feature vector into a fully connected model to obtain an output vector; concatenate the feature information fusion vector and the output vector to obtain a concatenated vector; and generate the above-mentioned recommended image set using a preset loss function.

[0105] In some optional implementations of some embodiments, the graph coding model includes a residual network model and multiple fully connected layers; the graph coding unit 502 in the device 500 can be further configured to: input the historical browsing creative image into the residual network model to obtain the model output result; input the model output result into the multiple fully connected layers to obtain the historical image coding vector.

[0106] In some optional implementations of certain embodiments, the generation unit 505 in the above-described apparatus 500 may be further configured to: generate a set of creative image scores for the spliced ​​vector using the preset loss function; generate a recommended item score corresponding to the target recommended item using a multi-head attention mechanism model based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set; in response to determining that the recommended item score is greater than a predetermined recommended item value, determine the creative image scores in the creative image score set that are greater than the predetermined recommended creative image value, and obtain a subset of creative image scores; and determine the creative image set corresponding to the subset of creative image scores as the recommended image set.

[0107] In some optional implementations of certain embodiments, the generation unit 505 in the above-mentioned device 500 may be further configured to: generate a recommended main image score corresponding to the target recommended item using a multi-head attention mechanism model based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set; and push the main image to the terminal corresponding to the target user in response to determining that the recommended item score is greater than a predetermined recommended item value and the recommended main image score is greater than a predetermined recommended main image value.

[0108] It is understandable that the units described in the device 500 are related to the reference. Figure 3 The steps in the described method correspond accordingly. Therefore, the operations, features, and beneficial effects described above for the method also apply to the device 500 and the units contained therein, and will not be repeated here.

[0109] The following is for reference. Figure 6 It illustrates electronic devices suitable for implementing some embodiments of this disclosure (e.g., Figure 1 A schematic diagram of the structure of electronic device 101)600 in the middle. Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0110] like Figure 6 As shown, electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processor, etc.) 601, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 602 or a program loaded from storage device 608 into random access memory (RAM) 603. RAM 603 also stores various programs and data required for the operation of electronic device 600. Processing device 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0111] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 608 including, for example, magnetic tapes, hard disks, etc.; and communication devices 609. Communication device 609 allows electronic device 600 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 An electronic device 600 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 6 Each box shown can represent a device or multiple devices as needed.

[0112] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 609, or installed from a storage device 608, or installed from a ROM 602. When the computer program is executed by the processing device 601, it performs the functions defined above in the methods of some embodiments of this disclosure.

[0113] It should be noted that, in some embodiments of this disclosure, the computer-readable medium described above may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. 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 disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0114] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and end-to-end networks (e.g., ad hoc end-to-end networks), as well as any currently known or future-developed networks.

[0115] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: acquire a sequence of historical browsing creative images of a target user, a set of user sparse feature information of the target user, a main image of a target recommended item, and a set of item sparse feature information of the target recommended item, wherein the set of item sparse feature information includes a set of creative image feature information for the target recommended item; perform graph encoding processing on the main image to obtain a main image encoding vector; and perform graph encoding processing on each historical browsing creative image in the aforementioned sequence of historical browsing creative images to generate a historical image encoding vector, obtaining a sequence of historical image encoding vectors; and perform graph encoding processing on the aforementioned user sparse feature information... Information encoding is performed on each user sparse feature information in the information set to generate a user sparse feature vector, resulting in a user sparse feature vector set. Information encoding is also performed on each item sparse feature information in the item sparse feature information set to generate an item sparse feature vector, resulting in an item sparse feature vector set. Visual preference adjustment is performed on each historical image encoding vector in the historical image encoding vector sequence to obtain an adjusted historical image encoding vector sequence. Based on the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set, and the item sparse feature vector set, a multi-head attention mechanism model is used to generate a set of recommended images corresponding to the target recommended items to be pushed to the target user.

[0116] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0117] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0118] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including an acquisition unit, an image encoding unit, an information encoding unit, an adjustment unit, and a generation unit. The names of these units do not necessarily limit the specific unit itself; for example, the acquisition unit may also be described as "a unit that acquires a sequence of historical creative images browsed by a target user, a set of sparse user feature information of the target user, a main image of a target recommended item, and a set of sparse item feature information of the target recommended item."

[0119] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0120] Some embodiments of this disclosure also provide a computer program product, including a computer program that, when executed by a processor, implements any of the above-described recommendation information generation methods.

[0121] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A recommendation information generation method, comprising: obtaining a historical creative image sequence of a target user, a user sparse feature information set of the target user, a main image of a target recommendation item, and an item sparse feature information set of the target recommendation item, wherein the item sparse feature information set comprises a creative image feature information set for the target recommendation item; performing graph encoding processing on the main image to obtain a main graph encoding vector, and performing graph encoding processing on each historical creative image in the historical creative image sequence to generate a historical image encoding vector, thereby obtaining a historical image encoding vector sequence; performing information encoding on each user sparse feature information in the user sparse feature information set to generate a user sparse feature vector, thereby obtaining a user sparse feature vector set, and performing information encoding on each item sparse feature information in the item sparse feature information set to generate an item sparse feature vector, thereby obtaining an item sparse feature vector set; adjusting each historical image encoding vector in the historical image encoding vector sequence according to a visual preference, thereby obtaining an adjusted historical image encoding vector sequence; generating, by using a multi-head attention mechanism model, a recommendation image set corresponding to the target recommendation item to be pushed to the target user according to the adjusted historical image encoding vector sequence, the main graph encoding vector, the user sparse feature vector set, and the item sparse feature vector set, wherein the recommendation image set is a subset of the creative image set corresponding to the creative image feature information set.

2. The method of claim 1, wherein, The adjusting each historical image encoding vector in the historical image encoding vector sequence according to a visual preference, thereby obtaining an adjusted historical image encoding vector sequence, comprises: determining visual preference information corresponding to the historical image encoding vector sequence as target visual preference information; adjusting each historical image encoding vector in the historical image encoding vector sequence according to the target visual preference information, thereby obtaining an adjusted historical image encoding vector sequence.

3. The method of claim 1, wherein, The generating, by using a multi-head attention mechanism model, a recommendation image set corresponding to the target recommendation item to be pushed to the target user according to the adjusted historical image encoding vector sequence, the main graph encoding vector, the user sparse feature vector set, and the item sparse feature vector set, comprises: splicing the user sparse feature vector set and the item sparse feature vector set to obtain a spliced sparse feature vector; inputting the adjusted historical image encoding vector sequence, the main graph encoding vector, and the spliced sparse feature vector into the multi-head attention mechanism model to obtain a feature information fusion vector; inputting the spliced sparse feature vector into a full connection model to obtain an output vector; splicing the feature information fusion vector and the output vector to obtain a spliced vector; generating the recommendation image set by using a preset loss function.

4. The method of claim 1, wherein, The performing graph encoding processing on each historical creative image in the historical creative image sequence to generate a historical image encoding vector, comprises: inputting the historical browsed creative image into a pre-trained graph encoding model to generate a historical image encoding vector.

5. The method of claim 4, wherein, The graph encoding model comprises a residual network model and a plurality of fully connected layers; and The inputting the historical browsed creative image into a pre-trained graph encoding model to generate a historical image encoding vector comprises: inputting the historical browsed creative image into the residual network model to obtain a model output result; inputting the model output result into the plurality of fully connected layers to obtain the historical image encoding vector.

6. The method of claim 3, wherein, The generating the recommended image set using the preset loss function comprises: generating a set of creative image scores for the splicing vector using the preset loss function; generating a recommended main image score corresponding to the target recommended item using a multi-head attention mechanism model according to the adjusted historical image encoding vector sequence, the main graph encoding vector, the set of user sparse feature vectors, and the set of item sparse feature vectors; in response to determining that the recommended item score is greater than a predetermined recommended item value, determining a creative image score in the set of creative image scores that is greater than a predetermined recommended creative image value to obtain a subset of creative image scores; determining a set of creative images corresponding to the subset of creative image scores as the recommended image set.

7. The method of claim 6, wherein, The method further comprises: generating a recommended main image score corresponding to the target recommended item using a multi-head attention mechanism model according to the adjusted historical image encoding vector sequence, the main graph encoding vector, the set of user sparse feature vectors, and the set of item sparse feature vectors; in response to determining that the recommended item score is greater than a predetermined recommended item value and the recommended main image score is greater than a predetermined recommended main image value, pushing the main image to a terminal corresponding to the target user.

8. A recommended information generation apparatus, comprising: an acquisition unit configured to acquire a sequence of historical browsed creative images of a target user, a set of user sparse feature information of the target user, a main image of a target recommended item, and a set of item sparse feature information of the target recommended item, wherein the set of item sparse feature information comprises a set of creative image feature information of the target recommended item; a graph encoding unit configured to perform graph encoding processing on the main image to obtain a main graph encoding vector, and perform graph encoding processing on each historical browsed creative image in the sequence of historical browsed creative images to generate a historical image encoding vector, thereby obtaining a sequence of historical image encoding vectors; an information encoding unit configured to perform information encoding on each user sparse feature information in the set of user sparse feature information to generate a user sparse feature vector, thereby obtaining a set of user sparse feature vectors, and perform information encoding on each item sparse feature information in the set of item sparse feature information to generate an item sparse feature vector, thereby obtaining a set of item sparse feature vectors; an adjustment unit configured to perform visual preference adjustment on each historical image encoding vector in the sequence of historical image encoding vectors to obtain an adjusted sequence of historical image encoding vectors; an adjustment unit configured to perform visual preference adjustment on each historical image encoding vector in the sequence of historical image encoding vectors to obtain an adjusted sequence of historical image encoding vectors; The generating unit is configured to generate, according to the adjusted historical image encoding vector sequence, the main image encoding vector, the user sparse feature vector set and the item sparse feature vector set, a recommendation image set corresponding to the target recommendation item to be pushed to the target user by using a multi-head attention mechanism model. 9.An electronic device, comprising: one or more processors; a memory device having stored thereon one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1-7.

10. A computer readable medium having stored thereon a computer program, wherein, The program is executed by the processor to implement the method of any one of claims 1-7. 11.A computer program product comprising a computer program which, when executed by a processor, implements the method of any one of claims 1-7.

Images