Method and device for generating interactive interface, electronic equipment and storage medium

Abstract

The application discloses a kind of generation method and device of interactive interface, electronic equipment, storage medium, it is related to artificial intelligence field, wherein, the method comprises: user function vector and attribute feature vector are parallelly connected and handled, obtain combined feature vector, combined feature vector is input into fully connected neural network, and combined feature vector is handled using fully connected neural network Feature fusion processing and dimension reduction processing are carried out, and fusion feature vector is obtained, and fusion feature vector and random Gaussian noise are input into layout generator, and interface layout picture is output, based on interface layout picture, the interactive interface to be displayed on the terminal screen of user terminal is laid out, and target interactive interface is generated.The application solves the technical problems that user interactive interface is single in the related art, and cannot be customized for different customers.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, in particular to a method and device for generating an interactive interface, an electronic device and a storage medium. BACKGROUND

[0002] With the continuous development of information technology and the upgrading of user terminals, the aesthetics of users gradually improves, the demand for interactive interfaces is more and more complex, and the design of interactive interfaces, as an important means to attract customers, also needs to be updated rapidly to cater to the needs of the public. Different people have different needs when using interactive interfaces, for example, young people want interactive interfaces to be more dynamic and colorful, middle-aged people want interactive interfaces to be more concise and clear, and the elderly want interactive interfaces to be simpler and easier to understand. The functions contained in the user interface are also important, and users will want the interface to contain only the functions they commonly use, and the functions they do not commonly use or do not use will be hidden as much as possible.

[0003] In related technologies, the design of the style of the interactive interface requires the interactive interface designer to design the interface separately according to the user's needs, the developer needs to develop the corresponding page, and the operation and maintenance personnel needs to maintain multiple sets of content. This way of designing the interactive interface has the following disadvantages: first, the user interface design is relatively single and cannot be customized for different users, and the user function buttons are also arranged on the interactive interface according to a uniform size and shape, resulting in a monotonous interface; second, the interactive interface designer needs to design the interactive interface separately for the needs of each user, which is time-consuming and labor-intensive.

[0004] At present, there is no effective solution to the above problems. SUMMARY

[0005] The embodiments of the present application provide a method and device for generating an interactive interface, an electronic device and a storage medium to at least solve the technical problem that the user interface in related technologies is single and cannot be customized for different customers.

[0006] According to an aspect of an embodiment of the present application, a method for generating an interactive interface is provided, including: processing a user function vector and an attribute feature vector in parallel to obtain a combined feature vector, wherein the user function vector corresponds to a function required by a financial account selected by a user terminal, and the attribute feature vector corresponds to an account attribute of an associated financial account selected by the user terminal; inputting the combined feature vector into a fully connected neural network, performing feature fusion processing and dimension reduction processing on the combined feature vector by using the fully connected neural network, and obtaining a fusion feature vector; inputting the fusion feature vector and random Gaussian noise into a layout generator, and outputting an interface layout picture; and performing layout on an interactive interface to be displayed on a terminal screen of the user terminal based on the interface layout picture, and generating a target interactive interface.

[0007] Optionally, before processing the user function vector and the attribute feature vector in parallel, the method further includes: receiving a function requirement set input by the user terminal, wherein the function requirement set at least includes N functions required by financial accounts selected by the user terminal, each function required by a financial account corresponds to a function icon, and N is a positive integer greater than or equal to 1; randomly combining function icons of the N functions required by financial accounts; pasting the function icons of the N functions required by financial accounts after random combination onto a specified background template to generate an initial function picture; and inputting the initial function picture into a picture classification model to output the user function vector, wherein the picture classification model includes a multi-layer convolutional neural network.

[0008] Optionally, before processing the user function vector and the attribute feature vector in parallel, the method further includes: receiving an attribute selection set input by the user terminal, wherein the attribute selection set at least includes M account attributes of associated financial accounts selected by the user terminal, each account attribute corresponds to an attribute embedding word table, and M is a positive integer greater than or equal to 1; converting a category to which an account attribute selected by the user terminal belongs into an attribute feature based on an attribute embedding word table corresponding to each account attribute; inputting each attribute feature into a multi-layer perceptron, performing fusion processing and encoding processing on all attribute features by using the multi-layer perceptron, and obtaining the attribute feature vector.

[0009] Optionally, further comprising: obtaining a plurality of historical interface layout pictures and historical fusion feature vectors to obtain a training set; inputting the historical interface layout pictures and the historical fusion feature vectors in the training set into a layout encoder to output a latent space vector, wherein the layout encoder comprises a multi-layer convolutional neural network; inputting the historical fusion feature vectors and the random Gaussian noise into the layout generator to output a to-be-pushed interface layout picture, wherein the layout generator comprises a multi-layer deconvolutional neural network; grouping the to-be-pushed interface layout picture and the random Gaussian noise as a first type of binary tuple, and grouping the historical interface layout picture and the latent space vector as a second type of binary tuple; using a discriminator to distinguish real pictures in the first type of binary tuple and the second type of binary tuple to generate a loss function value; adjusting the structure of each deconvolutional neural network in the layout generator based on the loss function value; and determining that the training of the layout generator is completed in a case where the loss function value is less than a preset function threshold.

[0010] Optionally, after the interface layout picture is used to layout an interactive interface to be displayed on a terminal screen of the user terminal, the method further comprises: receiving an interface illustration generation request, wherein the interface illustration generation request at least contains an illustration generation position and illustration text description information, the illustration generation position being an arbitrary click position or an arbitrary selected region of the user terminal on the interface layout picture; in response to the interface illustration generation request, inputting the illustration text description information into a layout interface illustration generation model to output at least one to-be-selected illustration; receiving an illustration selection operation of the user terminal; and based on the illustration selection operation, replacing a target to-be-selected illustration with an interface layout picture on an interactive interface at the illustration generation position to generate a target interface illustration.

[0011] Optionally, the layout interface illustration generation model is trained by using the following method: collecting a plurality of historical interactive pictures to generate a picture training set; using a pre-trained picture encoder to encode each historical interactive picture in the picture training set into a latent space vector, wherein the picture encoder comprises a multi-layer convolutional neural network; adding the random Gaussian noise to the latent space vector in K steps to obtain a target random noise, wherein K is a positive integer greater than or equal to 2; using a specified semantic segmentation model to perform restoration processing on the target random noise to output an illustration corresponding to the historical interactive picture, wherein the specified semantic segmentation model incorporates a text condition vector generated by a text encoder.

[0012] Optionally, the step of restoring the target random noise by using the specified semantic segmentation model comprises: integrating the text condition vector into the specified semantic segmentation model by using a preset text classification strategy; estimating the added noise corresponding to each step in the K steps by using the specified semantic segmentation model to obtain a noise estimation result; and restoring the target random noise based on the noise estimation result and the text condition vector.

[0013] According to another aspect of the embodiment of the present application, a generation device of an interactive interface is further provided, comprising: a parallel unit configured to parallel process a user function vector and an attribute feature vector to obtain a combined feature vector, wherein the user function vector corresponds to a function required by a financial account selected by a user terminal, and the attribute feature vector corresponds to an account attribute of an associated financial account selected by the user terminal; an input unit configured to input the combined feature vector into a fully connected neural network, and perform feature fusion processing and dimension reduction processing on the combined feature vector by using the fully connected neural network to obtain a fusion feature vector; an output unit configured to input the fusion feature vector and random Gaussian noise into a layout generator, and output an interface layout picture; and a layout unit configured to perform layout on an interactive interface to be displayed on a terminal screen of the user terminal based on the interface layout picture, and generate a target interactive interface.

[0014] Optionally, the generation device of the interactive interface further comprises: a first receiving module configured to receive a function requirement set input by the user terminal, wherein the function requirement set at least contains functions required by N financial accounts selected by the user terminal, each function required by a financial account corresponds to a function icon, and N is a positive integer greater than or equal to 1; a first combination module configured to randomly combine the function icons of the functions required by the N financial accounts; a first generation module configured to paste the function icons of the functions required by the N financial accounts after random combination onto a specified background template to generate an initial function picture; and a first input module configured to input the initial function picture into a picture classification model to output the user function vector, wherein the picture classification model contains a multi-layer convolutional neural network.

[0015] Optionally, the generation apparatus of the interactive interface further comprises: a second receiving module configured to receive an attribute selection set input by the user terminal, wherein the attribute selection set at least includes M account attributes of the associated financial account selected by the user terminal, each account attribute corresponds to an attribute embedding vocabulary, and M is a positive integer greater than or equal to 1; a first conversion module configured to convert a category of the account attribute selected by the user terminal into an attribute feature based on the attribute embedding vocabulary corresponding to each account attribute; and a first processing module configured to input each attribute feature into a multi-layer perception machine, and perform fusion processing and encoding processing on all attribute features by using the multi-layer perception machine to obtain the attribute feature vector.

[0016] Optionally, the generation apparatus of the interactive interface further comprises: a first obtaining module configured to obtain a plurality of historical interface layout pictures and historical fusion feature vectors to obtain a training set; a second input module configured to input the historical interface layout pictures and the historical fusion feature vectors in the training set into a layout encoder to output a latent space vector, wherein the layout encoder comprises a multi-layer convolutional neural network; a third input module configured to input the historical fusion feature vectors and the random Gaussian noise into the layout generator to output a to-be-pushed interface layout picture, wherein the layout generator comprises a multi-layer deconvolutional neural network; a first composition module configured to compose the to-be-pushed interface layout picture and the random Gaussian noise into a first type of binary tuple, and compose the historical interface layout picture and the latent space vector into a second type of binary tuple; a first discrimination module configured to discriminate a real picture in the first type of binary tuple and the second type of binary tuple by using a discriminator to generate a loss function value; a first adjustment module configured to adjust a structure of each deconvolutional neural network in the layout generator based on the loss function value; and a first determination module configured to determine that the training of the layout generator is completed in a case where the loss function value is less than a preset function threshold.

[0017] Optionally, the layout unit comprises: a third receiving module configured to receive an interface illustration generation request, wherein the interface illustration generation request at least includes an illustration generation position and illustration text description information, and the illustration generation position is an arbitrary click position or an arbitrary selected region of the user terminal on the interface layout picture; a first response module configured to respond to the interface illustration generation request, input the illustration text description information into a layout interface illustration generation model, and output at least one to-be-selected illustration; a fourth receiving module configured to receive an illustration selection operation of the user terminal; and a first replacement module configured to replace, based on the illustration selection operation, a target to-be-selected illustration with an interface layout picture on the interactive interface at the illustration generation position to generate a target interface illustration.

[0018] Optionally, the generation apparatus of the interactive interface further comprises: a first acquisition module, configured to acquire a plurality of historical interaction pictures to generate a picture training set; a first encoding module, configured to encode each historical interaction picture in the picture training set into a latent space vector by using a pre-trained picture encoder, wherein the picture encoder comprises a multi-layer convolutional neural network; a first adding module, configured to add the random Gaussian noise to the latent space vector in K steps to obtain a target random noise, wherein K is a positive integer greater than or equal to 2; and a first output module, configured to restore the target random noise by using a specified semantic segmentation model to output an illustration corresponding to the historical interaction picture, wherein the specified semantic segmentation model incorporates a text condition vector generated by a text encoder.

[0019] Optionally, the first output module comprises: a first incorporation submodule, configured to incorporate the text condition vector into the specified semantic segmentation model by using a preset text classification strategy; a first estimation submodule, configured to estimate the added noise corresponding to each of the K steps by using the specified semantic segmentation model to obtain a noise estimation result; and a first restoration submodule, configured to restore the target random noise based on the noise estimation result and the text condition vector.

[0020] According to another aspect of the embodiments of the present application, a computer readable storage medium is also provided, which comprises a stored computer program, wherein the computer program, when executed, controls a device in which the computer readable storage medium is located to perform any one of the generation methods of the interactive interface.

[0021] According to another aspect of the embodiments of the present application, an electronic device is also provided, which comprises one or more processors and a memory, the memory being configured to store one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement any one of the generation methods of the interactive interface.

[0022] In the present disclosure, the following steps are adopted: firstly, a user function vector and an attribute feature vector are processed in parallel to obtain a combined feature vector, wherein the user function vector corresponds to a function required by a financial account selected by a user terminal, and the attribute feature vector corresponds to an account attribute of an associated financial account selected by the user terminal; secondly, the combined feature vector is input into a fully connected neural network, and the fully connected neural network is used to perform feature fusion processing and dimension reduction processing on the combined feature vector to obtain a fusion feature vector; thirdly, the fusion feature vector and a random Gaussian noise are input into a layout generator to output an interface layout picture; and finally, based on the interface layout picture, an interactive interface to be displayed on a terminal screen of the user terminal is laid out to generate a target interactive interface.

[0023] In the present disclosure, a user can select attributes and functions required by an interface on a terminal interface, and the system customizes a personalized interface for the user based on the attributes and functions selected by the user, thereby realizing personalized definition of an interactive interface, improving user experience, reducing workload of interactive interface designers, and solving the technical problem of single user interactive interface and inability to customize for different customers in related technologies. BRIEF DESCRIPTION OF DRAWINGS

[0024] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:

[0025] Figure 1 is a flow chart of an optional generation method of an interactive interface according to an embodiment of the present application;

[0026] Figure 2 is a flow chart of an optional acquisition of a user function vector according to an embodiment of the present application;

[0027] Figure 3 is a schematic diagram of generation of a picture based on a text description according to an embodiment of the present application;

[0028] Figure 4 is a structural schematic diagram of an attribute feature generation model according to an embodiment of the present application;

[0029] Figure 5 is a schematic diagram of an optional generation device of an interactive interface according to an embodiment of the present application;

[0030] Figure 6 is a hardware structural block diagram of an electronic device (or a mobile device) of a generation method of an interactive interface according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] In order to enable persons skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by persons skilled in the art without creative labor should fall within the scope of protection of the present application.

[0032] It should be noted that the terms "first", "second", and the like in the description and claims of the application and the above drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0033] It should be noted that the generation method and device of the interactive interface in the present disclosure can be used in the field of artificial intelligence in the case of customized design of the user interactive interface, and can also be used in any field other than the field of artificial intelligence in the case of generating personalized interactive interface. The application field of the generation method and device of the interactive interface in the present disclosure is not limited.

[0034] It should be noted that the related information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, analyzed data, etc.) involved in the present disclosure are all information and data authorized by the user or authorized by all parties. For example, an interface is provided between the system and the related user or institution. Before obtaining the related information, the interface needs to send a request to the aforementioned user or institution, and after receiving the consent information fed back by the aforementioned user or institution, the related information is obtained.

[0035] The present application can be applied to various interactive interface generation devices / apparatus / systems (for example, various mobile terminals, PC terminals, financial APPs), and the user can select the attributes and functions in the interface to be designed on the terminal interface. The system customizes a personalized interface for the user based on the attributes and functions selected by the user, thereby realizing personalized definition of the interactive interface, reducing the workload of the interactive interface designer while improving the user experience.

[0036] The present application will be described in detail below in conjunction with various embodiments.

[0037] Example One

[0038] According to the embodiments of the present application, a method for generating an interactive interface is provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0039] Figure 1 is a flowchart of an optional method for generating an interactive interface according to an embodiment of the present application, as shown in the figure, the method comprises the following steps: Figure 1

[0040] In step S101, the user function vector and the attribute feature vector are processed in parallel to obtain a combined feature vector, wherein the user function vector corresponds to the functions required by the financial account selected by the user terminal, and the attribute feature vector corresponds to the account attributes of the associated financial account selected by the user terminal.

[0041] In step S102, the combined feature vector is input into a fully connected neural network, and the fully connected neural network is used to perform feature fusion processing and dimension reduction processing on the combined feature vector to obtain a fused feature vector.

[0042] In step S103, the fused feature vector and random Gaussian noise are input into a layout generator to output an interface layout picture.

[0043] In step S104, based on the interface layout picture, the interactive interface to be displayed on the terminal screen of the user terminal is laid out to generate a target interactive interface.

[0044] Through the above steps, the user function vector and the attribute feature vector are first processed in parallel to obtain a combined feature vector, wherein the user function vector corresponds to the functions required by the financial account selected by the user terminal, and the attribute feature vector corresponds to the account attributes of the associated financial account selected by the user terminal. Then, the combined feature vector is input into a fully connected neural network, and the fully connected neural network is used to perform feature fusion processing and dimension reduction processing on the combined feature vector to obtain a fused feature vector. The fused feature vector and random Gaussian noise are input into a layout generator to output an interface layout picture. Finally, based on the interface layout picture, the interactive interface to be displayed on the terminal screen of the user terminal is laid out to generate a target interactive interface. In this embodiment, the user can select the required attributes and functions of the interface on the terminal interface, and the system customizes a personalized interface for the user based on the selected attributes and functions, thereby realizing personalized definition of the interactive interface, improving the user experience, and reducing the workload of the interactive interface design personnel, thereby solving the technical problem that the user interactive interface is single and cannot be customized for different customers in related technologies.

[0045] ​It should be noted that the embodiment of the present application provides a generation scheme / model of user interaction interface automatic design based on a generative adversarial network, realizes personalized design of the user interaction interface, meets personalized needs of different users, and in particular, for software such as a financial APP, each interaction page contains a large amount of function information, and different users have different needs, for example, for middle-aged and old people, a more concise interaction interface is preferred, and the functions used are less, and through the embodiment, the user can design page functions and page layouts according to own needs.

[0046] It should be noted that before the user function vector and the attribute feature vector are processed in parallel, the user function vector also needs to be obtained, and how to obtain the user function vector is first explained in the embodiment. Figure 2 is an optional flowchart for obtaining a user function vector according to the embodiment of the present application, as shown in Figure 2 The steps of obtaining the user function vector include:

[0047] Step S201, receiving a function requirement set input by a user terminal, wherein the function requirement set at least contains functions required by N financial accounts selected by the user terminal, each function required by the financial account corresponds to a function icon, and N is a positive integer greater than or equal to 1;

[0048] Step S202, randomly combining the function icons of the functions required by the N financial accounts;

[0049] Step S203, pasting the function icons of the functions required by the N financial accounts after random combination to a specified background template to generate an initial function picture;

[0050] Step S204, inputting the initial function picture into a picture classification model to output a user function vector, wherein the picture classification model contains a multi-layer convolutional neural network.

[0051] It should be noted that the user interface of the financial APP contains functions such as transfer, balance viewing, finance, credit, investment, expenditure flow viewing, and income flow viewing, these functions will be displayed in the form of small icons on the interaction page, the user can view and select the functions needed on the terminal page, after the user selects, the selected icons are randomly combined and pasted to a background template of a fixed size (the color and type of the background template can be customized, for example, the background template is selected as a pure black background), and the functions not needed can be hidden and not displayed on the interface.

[0052] In the embodiment of the present application, before the user function vector and the attribute feature vector are processed in parallel, the attribute feature vector also needs to be obtained. When obtaining the attribute feature vector, the following steps are included: receiving an attribute selection set input by the user terminal, wherein the attribute selection set at least includes M account attributes of the associated financial account selected by the user terminal, each account attribute corresponds to an attribute embedding word table, and M is a positive integer greater than or equal to 1; converting the category to which the account attribute selected by the user terminal belongs into an attribute feature based on the attribute embedding word table corresponding to each account attribute; inputting each attribute feature into a multi-layer perception machine, and using the multi-layer perception machine to perform fusion processing and encoding processing on all attribute features to obtain an attribute feature vector.

[0053] It should be noted that the attribute features include but are not limited to age, gender, occupation, and preference. The category to which the account attribute belongs can be presented in various ways on the user interface, such as an up interface, a virtual frame interface, a drop-down interface, a sliding interface, and the like. For example, the category to which the account attribute belongs can be presented in the form of a drop-down interface, and the user can select or fill in through the drop-down interface module on the terminal interface. The user can set each attribute value, and each attribute will have a corresponding embedding word table. For example, the age category will be set to multiple ranges: childhood (0-6 years old), adolescence (7-17 years old), youth (18-40 years old), middle age (35-59 years old), and old age (60 years old and above). The model will convert the age category selected by the user into a vector with a specified dimension according to the age embedding word table. Other attributes of the user will also be converted and merged in the same way to form attribute features, and the generated attribute features will be input into a multi-layer perception machine for feature fusion and encoding. Finally, an attribute feature vector with a specified dimension is output. The size of the attribute feature vector needs to be customized, for example, 1024 dimensions.

[0054] After obtaining the user function vector and the attribute feature vector, the different vectors need to be fused to obtain a fused feature vector, which is used to represent the preferences and use requirements of the user.

[0055] The specific application scenarios of the present embodiment will be described in detail in combination with the above-mentioned implementation steps.

[0056] Step S101, the user function vector and the attribute feature vector are processed in parallel to obtain a combined feature vector, wherein the user function vector corresponds to the functions required by the financial account selected by the user terminal, and the attribute feature vector corresponds to the account attributes of the associated financial account selected by the user terminal.

[0057] The financial account mentioned in the embodiment can refer to a card number registered or applied in a financial institution, such as a savings card or a credit card, or a personal virtual account registered or applied by a user through a financial APP, or an account applied by a user through a terminal and allowed to be used as a financial account.

[0058] In step S102, the combined feature vector is input into the fully connected neural network, the fully connected neural network is used for feature fusion processing and dimension reduction processing on the combined feature vector, and a fusion feature vector is obtained.

[0059] The embodiment provides a virtual feature fusion module, which realizes parallel processing of a user function vector and an attribute feature vector, the feature virtual module is composed of a fully connected neural network, the fully connected neural network receives the combined feature vector after parallel processing, inputs the combined feature vector into the fully connected neural network for feature fusion and dimension reduction, and finally outputs a fusion feature vector of a specified dimension.

[0060] In step S103, the fusion feature vector and random Gaussian noise are input into the layout generator, and an interface layout picture is output.

[0061] In the embodiment of the application, before the fusion feature vector and the random Gaussian noise are input into the layout generator, the layout generator (or the layout generation model) needs to be constructed and trained and adjusted, and the steps include: obtaining a plurality of historical interface layout pictures and historical fusion feature vectors to obtain a training set; inputting the historical interface layout pictures and the historical fusion feature vectors in the training set into a layout encoder to output a hidden space vector, wherein the layout encoder comprises a multi-layer convolutional neural network; inputting the historical fusion feature vector and the random Gaussian noise into the layout generator to output a to-be-pushed interface layout picture, wherein the layout generator comprises a multi-layer deconvolutional neural network; the to-be-pushed interface layout picture and the random Gaussian noise are composed into a first type of binary tuple, and the historical interface layout picture and the hidden space vector are composed into a second type of binary tuple; a discriminator is used to discriminate the real picture in the first type of binary tuple and the second type of binary tuple to generate a loss function value; the structure of each deconvolutional neural network in the layout generator is adjusted based on the loss function value; and in the case that the loss function value is less than a preset function threshold, it is determined that the layout generator is trained.

[0062] It should be noted that in the embodiment, the layout generator is trained and adjusted by the layout generation model, the layout generation model uses a conditional generative adversarial network as a basic model, and at least includes a layout encoder, a layout generator and a discriminator three parts, wherein the layout encoder is composed of a multi-layer convolutional neural network, the input is the layout picture x (corresponding to the historical interface layout picture) and the condition vector c (corresponding to the fusion feature vector) in the training data set, and the output is the hidden space vector z' after encoding; the layout generator is composed of a multi-layer deconvolutional neural network, the input is random Gaussian noise z and the condition vector c (corresponding to the historical fusion feature vector), and the output is the layout picture x' (corresponding to the to-be-pushed interface layout picture); the discriminator is composed of a multi-layer convolutional neural network, the input is two binary tuples (x', z) and (x, z'), the two binary tuples respectively represent the real layout picture and the generated layout picture, the discriminator will judge which binary tuple is the real picture and which binary tuple is the generated picture according to the input, so as to generate a loss function value, which can guide the layout generator to generate more real layout pictures.

[0063] In step S104, based on the interface layout picture, the interactive interface to be displayed on the terminal screen of the user terminal is laid out, and a target interactive interface is generated.

[0064] Through the above embodiment, the interactive interface can be generated based on the required function options and attribute options selected by the user.

[0065] In the related art, the monotonous interactive page lacks personalized illustrations, which causes user aesthetic fatigue and affects user experience. In the embodiment of the present application, in order to solve the problem of monotonous interactive page illustrations, a layout interface illustration generation model is introduced to solve the problem of page background pictures. In the embodiment of the present application, after laying out the interactive interface to be displayed on the terminal screen of the user terminal based on the interface layout picture, the method further comprises: receiving an interface illustration generation request, wherein the interface illustration generation request at least includes: an illustration generation position and illustration text description information, the illustration generation position is an arbitrary click position or an arbitrary selected area of the user terminal on the interface layout picture; in response to the interface illustration generation request, inputting the illustration text description information into the layout interface illustration generation model, and outputting at least one to-be-selected illustration; receiving an illustration selection operation of the user terminal; based on the illustration selection operation, replacing the interface layout picture on the interactive interface in the illustration generation position with the selected target to-be-selected illustration, and generating a target interface illustration.

[0066] When the user generates a suitable page layout using the page layout generation model, the user can customize personalized page illustrations for different parts. The types of the page illustrations are various, such as page background pictures, function button stickers, welcome page stickers, etc.

[0067] It should be noted that when the user is laying out the page illustrations, the user generally first selects the illustration position that needs to be customized, for example, a background page, a cover page, a cover top, a cover bottom, etc., and then the user needs to manually input a text description of the picture, for example, an astronaut riding a horse in space, a rabbit squatting on the roadside eating grass, etc. Figure 3 is a schematic view of generating a picture based on a text description according to an embodiment of the present application, as shown in Figure 3 If the user manually inputs the text description of the picture, the page will present multiple pictures related to the text theme for the user to select, and the picture in the figure illustrates an astronaut riding a horse in space. The user only needs to provide a text description, and the layout page illustration generation model can generate multiple pictures for the user to select based on the text description.

[0068] In the embodiment of the present application, the layout interface illustration generation model is trained by the following method: a plurality of historical interaction pictures are collected to generate a picture training set; each historical interaction picture in the picture training set is encoded into a latent space vector by using a pre-trained picture encoder, wherein the picture encoder comprises a plurality of layers of convolutional neural networks; random Gaussian noise is added to the latent space vector in K steps to obtain target random noise, wherein K is a positive integer greater than or equal to 2; a specified semantic segmentation model is used to restore the target random noise to output an illustration corresponding to the historical interaction picture, wherein the specified semantic segmentation model incorporates a text condition vector generated by a text encoder.

[0069] It should be noted that the layout interface illustration generation model is based on a diffusion generation model, based on the principle of non-equilibrium thermodynamics, a Markov chain of a diffusion step is defined, random noise is gradually added to the data, and then an inverse diffusion process is learned to construct the required data samples from the noise. The training process of the diffusion generation model is more stable and the dimension of the latent variable is higher, which is more helpful to improve the effect of picture generation.

[0070] The layout page illustration generation model in the embodiment of the present application converts the original calculation process at the pixel level to the latent space level by introducing an autoencoder module, greatly reduces the calculation resources, and thus facilitates use. The autoencoder module includes a picture encoder and a picture decoder, the picture encoder is composed of multiple layers of convolutional neural networks and can encode the input picture into a latent space vector z, the picture decoder is composed of multiple layers of deconvolutional neural networks and can restore the latent space vector z to the original picture, and the layout interface illustration generation model is pre-trained using a large picture dataset and the weights are saved.

[0071] It should be noted that when the layout interface illustration generation model is trained, the main structure of the model is a U-net structure (corresponding to the specified semantic segmentation model described above), a conditional vector is generated by a pre-trained Chinese CLIP model (Contrastive Language-Image Pre-training, a pre-trained model based on a contrastive text-image pair) text encoder, the training data set of the model is a large picture data set, the model first encodes the input picture into a hidden space vector z using a pre-trained picture encoder, and then adds random Gaussian noise to the hidden space vector z in multiple steps, each time the noise is added to make the picture more blurred, until the last time the noise is added, and the vector z becomes completely random noise z T .

[0072] In the embodiment of the application, the step of restoring the random noise by using the specified semantic segmentation model includes: fusing the text condition vector into the specified semantic segmentation model by using a preset text classification strategy; estimating the added noise corresponding to each step in K steps by using the specified semantic segmentation model to obtain a noise estimation result; and restoring the random noise based on the noise estimation result and the text condition vector.

[0073] After obtaining the random noise z T , the vector needs to be restored, and the U-net model is used to perform noise reduction operation on the random noise z T , and the U-net model will predict the noise added in each step in turn to restore z T . In the restoration process, the text condition vector c generated by the text encoder in the pre-trained Chinese CLIP model will be used as a conditional vector and fused into the U-net model in a cross-attention manner (corresponding to the preset text classification strategy described above) to guide the noise reduction process of the model on z T .

[0074] Through the above embodiment, a personalized interface can be customized for a user based on the attributes and functions selected by the user, the workload of an interactive interface designer can be reduced while the user experience is improved, a diffusion model is introduced to generate a layout page illustration, the original calculation process at the pixel level is converted to the hidden space level, the calculation resources are greatly reduced, the user has more picture selection, and the user experience is improved.

[0075] The application will be described in detail below with reference to a specific embodiment.

[0076] Example Two

[0077] The customized interactive page of the embodiment of the application is mainly realized based on two models, namely a page layout generation model and a layout page illustration generation model.

[0078] Page layout generation model: the page layout generation model is used to generate the layout format of the user interaction interface, and the model can automatically generate multiple page layouts according to the related attributes selected by the user for the user to select. The model is composed of an attribute feature generation model and a layout generation model.

[0079] Figure 4 is a structural schematic diagram of an attribute feature generation model according to an embodiment of the application, as Figure 4 shown, the attribute feature generation model includes four different modules: a function decoding module 41, an attribute decoding module 42, a feature fusion module 43, and a layout generation module 44.

[0080] The function decoding module 41 uses a pre-trained picture classification model (such as the Resnet series, VGG details, and the ViT series), which is composed of a multi-layer convolutional neural network and is pre-trained on the imagenet (a large visual database for visual object recognition software research) dataset. The trained model weight will be saved and directly used.

[0081] The function decoding module 41 inputs the commonly used functions selected by the user according to their own needs (such as transferring money, checking balance, investment and finance, etc.), which will be provided to the user in the form of small icons for selection. After the user selects, the selected icons will be randomly combined and pasted on a fixed-size black background (corresponding to the above specified background template), and the newly generated picture will be input into the pre-trained picture classification model. The function decoding module will extract the feature vector output by the second-to-last layer of the picture classification model as the user function vector.

[0082] The attribute decoding module 42 uses a multi-layer perceptron (MLP) as the basic model, and uses an attribute embedding module to convert attributes into fixed-length feature vectors. Multiple attribute feature vectors will be combined together to form attribute features, which will be input into the multi-layer perceptron and output a fixed-length attribute feature vector. The input of the attribute decoding module is a pre-set attribute (such as age range, gender, favorite font, font size, etc.), and the user can set the value of each attribute. Each attribute will have a corresponding embedding vocabulary, for example, the age category will be set to multiple ranges: childhood (0-6 years old), adolescence (7-17 years old), youth (18-40 years old), middle age (35-59 years old), and old age (60 years old and above). The model will convert the user-selected age category into a 1024-dimensional vector according to the age embedding vocabulary. Other attributes of the user will be converted and combined in the same way to form attribute features, and the generated attribute features will be input into the multi-layer perceptron for feature fusion and encoding, and finally output a 1024-dimensional attribute feature vector.

[0083] The feature fusion module 43 fuses the feature vectors output by the function decoding module and the attribute decoding module and inputs them into the layout generation module as a condition vector to guide the generation of the page layout. The feature fusion module is composed of a fully connected neural network. The fully connected neural network first accepts the feature vectors output by the function decoding module and the attribute decoding module and performs parallel connection. The combined feature vector after parallel connection is input into the fully connected neural network for feature fusion and dimension reduction. The final model outputs a fused feature vector c with a length of 1024 dimensions.

[0084] The layout generation module 44 uses a conditional generative adversarial network as a basic model. The layout generation module 44 includes at least three parts: a layout encoder, a layout generator, and a discriminator. During the training stage, all three modules are used. The overall process is as follows:

[0085] The layout encoder is composed of a multi-layer convolutional neural network. Its input is the layout picture x in the training data set (corresponding to the historical interface layout picture described above) and the condition vector c (corresponding to the fused feature vector described above). The output is the hidden space vector z’ after encoding.

[0086] The layout generator is composed of a multi-layer deconvolutional neural network. Its input is random Gaussian noise z and the condition vector c. Its output is the layout picture x’ (corresponding to the to-be-pushed interface layout picture described above).

[0087] The discriminator is composed of a multi-layer convolutional neural network. Its input is two binary tuples (x’, z) and (x, z’). The two binary tuples represent the real layout picture and the generated layout picture, respectively. The discriminator determines which binary tuple is the real picture and which binary tuple is the generated picture based on the input, thereby generating a loss function value, which can guide the layout generator to generate more realistic layout pictures.

[0088] Layout page illustration generation model: The introduction of the layout page illustration generation model is mainly to meet the needs of user personalization layout and personalized layout page illustration. Traditional monotonous page illustrations lack personalization, which leads to user aesthetic fatigue and affects user experience. To solve the above problems, a diffusion model is introduced to solve the problem of page background pictures. After the user generates a suitable page layout using the page layout generation model, the user can customize personalized page illustrations for different parts, such as page background pictures, function button stickers, welcome page stickers, etc. The user can choose default pictures, self-selected pictures, and generated pictures, etc.

[0089] The diffusion generative model is based on the principle of non-equilibrium thermodynamics, first defines a Markov chain of diffusion steps, gradually adds random noise to the data, and then learns the inverse diffusion process to construct the required data samples from the noise. Compared with the traditional generative model of generative adversarial network, the training process of the diffusion generative model is more stable and the dimension of the hidden variable is higher, which is more helpful to improve the effect of picture generation. The diffusion model includes de l l-E 2, imagen and stable diffusion, etc. The embodiment of the present application takes the stable diffusion model as an example to illustrate the layout page illustration generation model, which overcomes the problem of excessive calculation caused by adding and denoising noise at the pixel level in the previous model. By introducing an autoencoder structure of a picture encoder-decoder model, the original calculation process at the pixel level is converted to the hidden space level, greatly reducing the calculation resources, thereby facilitating use.

[0090] The autoencoder module includes a picture encoder and a picture decoder. The picture encoder is composed of a multi-layer convolutional neural network and can encode the input picture into a hidden space vector z. The picture decoder is composed of a multi-layer deconvolutional neural network and can restore the hidden space vector z to the original picture. The model is pre-trained using a large picture dataset and the weights are saved.

[0091] The main structure of the diffusion generative model is a U-net structure, and the conditional vector is generated by a text encoder in a pre-trained Chinese CLIP model. The training data set of the model is a large picture data set. The model first encodes the input picture into a hidden space vector z using the pre-trained picture encoder in the previous step, then adds random Gaussian noise to the hidden space vector z in 1000 steps (the specific number of steps is not limited in the embodiment of the present application, and 1000 steps are taken as an example in the embodiment of the present application), each time the noise is added, the picture becomes more blurred, and the last time the noise is added, the vector z is completely changed into random noise z T . Then the model uses the U-net model to denoise the random noise z T , and the U-net model will predict the noise added at each step in turn to restore z T . In the restoration process, the text condition vector c generated by the text encoder in the pre-trained Chinese CLIP model will be used as a conditional vector and fused into the U-net model using cross-attention to guide the denoising process of the model for z T .

[0092] After the layout page illustration generation model is constructed, the user first selects the picture position that needs to be customized, and then manually inputs the text description of the picture that the user wants, for example: an astronaut riding a horse is running in space, the layout page illustration generation model generates multiple pictures based on the text description, the user selects the picture that the user likes and replaces it into the layout page, and repeatedly repeats the above process until the user is satisfied, that is, the custom design of the user interaction page is completed.

[0093] Through the above embodiments, the page layout generation model and the layout page illustration generation model are introduced, so that the user can design and define the personalized interaction interface, thereby improving the user experience and reducing the workload of the interaction interface designer. By introducing the diffusion model, the originally pixel-level calculation process is converted to the latent space level, which greatly reduces the calculation resources.

[0094] The application will be described below in conjunction with another alternative embodiment.

[0095] Example Three

[0096] The embodiment provides an interaction interface generation device, each implementation unit of the interaction interface generation device corresponds to each implementation step in the first embodiment.

[0097] Figure 5 is a schematic diagram of an alternative interaction interface generation device according to an embodiment of the application, as Figure 5 shown, the interaction interface generation device comprises a parallel unit 51, an input unit 52, an output unit 53 and a layout unit 54, wherein,

[0098] The parallel unit 51 is configured to perform parallel processing on the user function vector and the attribute feature vector to obtain a combined feature vector, wherein the user function vector corresponds to the function required by the financial account selected by the user terminal, and the attribute feature vector corresponds to the account attribute of the associated financial account selected by the user terminal;

[0099] The input unit 52 is configured to input the combined feature vector into a fully connected neural network, and perform feature fusion processing and dimension reduction processing on the combined feature vector by using the fully connected neural network to obtain a fusion feature vector;

[0100] The output unit 53 is configured to input the fusion feature vector and random Gaussian noise into a layout generator, and output an interface layout picture;

[0101] The layout unit 54 is configured to perform layout on the interaction interface to be displayed on the terminal screen of the user terminal based on the interface layout picture, and generate a target interaction interface.

[0102] The generation apparatus of the interactive interface in the embodiment generates the target interactive interface by parallel processing the user function vector and the attribute feature vector through the parallel unit 51, inputting the combined feature vector into the fully connected neural network through the input unit 52, performing feature fusion processing and dimension reduction processing on the combined feature vector by the fully connected neural network, inputting the fusion feature vector and the random Gaussian noise into the layout generator through the output unit 53, and outputting the interface layout picture, and performing layout on the interactive interface to be displayed on the terminal screen of the user terminal based on the interface layout picture through the layout unit 54. In the embodiment, the user can select the attributes and functions required by the interface on the terminal interface, and the system customizes the personalized interface for the user based on the selected attributes and functions of the user, thereby realizing the personalized definition of the interactive interface, improving the user experience, reducing the workload of the interactive interface designer, and solving the technical problem that the user interactive interface is single and cannot be customized for different customers in the related art.

[0103] Optionally, the generation apparatus of the interactive interface further includes: a first receiving module configured to receive a function requirement set input by a user terminal, wherein the function requirement set at least includes functions required by N financial accounts selected by the user terminal, each function required by a financial account corresponds to a function icon, and N is a positive integer greater than or equal to 1; a first combination module configured to randomly combine the function icons of the functions required by the N financial accounts; a first generation module configured to paste the function icons of the functions required by the N financial accounts after random combination on a specified background template to generate an initial function picture; and a first input module configured to input the initial function picture into a picture classification model to output a user function vector, wherein the picture classification model includes a multi-layer convolutional neural network.

[0104] Optionally, the generation apparatus of the interactive interface further includes: a second receiving module configured to receive an attribute selection set input by a user terminal, wherein the attribute selection set at least includes M account attributes of associated financial accounts selected by the user terminal, each account attribute corresponds to an attribute embedding vocabulary, and M is a positive integer greater than or equal to 1; a first conversion module configured to convert a category to which an account attribute selected by the user terminal belongs into an attribute feature based on an attribute embedding vocabulary corresponding to each account attribute; and a first processing module configured to input each attribute feature into a multi-layer perceptron to perform fusion processing and encoding processing on all attribute features by the multi-layer perceptron to obtain an attribute feature vector.

[0105] Optionally, the generation apparatus of the interactive interface further comprises: a first acquisition module, configured to acquire a plurality of historical interface layout pictures and historical fusion feature vectors to obtain a training set; a second input module, configured to input the historical interface layout pictures and the historical fusion feature vectors in the training set to a layout encoder to output a latent space vector, wherein the layout encoder comprises a multi-layer convolutional neural network; a third input module, configured to input the historical fusion feature vectors and random Gaussian noise to a layout generator to output a to-be-pushed interface layout picture, wherein the layout generator comprises a multi-layer deconvolutional neural network; a first composition module, configured to compose the to-be-pushed interface layout picture and the random Gaussian noise into a first type of binary tuple and compose the historical interface layout picture and the latent space vector into a second type of binary tuple; a first discrimination module, configured to discriminate real pictures in the first type of binary tuple and the second type of binary tuple by using a discriminator to generate a loss function value; a first adjustment module, configured to adjust structures of the deconvolutional neural networks in the layout generator based on the loss function value; and a first determination module, configured to determine that the training of the layout generator is completed in a case where the loss function value is less than a preset function threshold.

[0106] Optionally, the layout unit 54 comprises: a third receiving module, configured to receive an interface illustration generation request, wherein the interface illustration generation request at least contains an illustration generation position and illustration text description information, and the illustration generation position is an arbitrary click position or an arbitrary selected region of the user terminal on the interface layout picture; a first response module, configured to respond to the interface illustration generation request, input the illustration text description information into a layout interface illustration generation model, and output at least one to-be-selected illustration; a fourth receiving module, configured to receive an illustration selection operation of the user terminal; and a first replacement module, configured to replace a target to-be-selected illustration selected by the user terminal with the interface layout picture on the interactive interface at the illustration generation position to generate a target interface illustration.

[0107] Optionally, the generation apparatus of the interactive interface further comprises: a first acquisition module, configured to acquire a plurality of historical interface layout pictures and historical fusion feature vectors to obtain a training set; a second input module, configured to input the historical interface layout pictures and the historical fusion feature vectors in the training set to a layout encoder to output a latent space vector, wherein the layout encoder comprises a multi-layer convolutional neural network; a third input module, configured to input the historical fusion feature vectors and random Gaussian noise to a layout generator to output a to-be-pushed interface layout picture, wherein the layout generator comprises a multi-layer deconvolutional neural network; a first composition module, configured to compose the to-be-pushed interface layout picture and the random Gaussian noise into a first type of binary tuple and compose the historical interface layout picture and the latent space vector into a second type of binary tuple; a first discrimination module, configured to discriminate real pictures in the first type of binary tuple and the second type of binary tuple by using a discriminator to generate a loss function value; a first adjustment module, configured to adjust structures of the deconvolutional neural networks in the layout generator based on the loss function value; and a first determination module, configured to determine that the training of the layout generator is completed in a case where the loss function value is less than a preset function threshold.

[0108] Optionally, the first output module comprises: a first fusion submodule, configured to fuse the text condition vector into the designated semantic segmentation model by using a preset text classification strategy; a first estimation submodule, configured to estimate the added noise corresponding to each step in the K steps by using the designated semantic segmentation model to obtain a noise estimation result; and a first restoration submodule, configured to perform restoration processing on the target random noise based on the noise estimation result and the text condition vector.

[0109] The generation apparatus of the interactive interface described above can further include a processor and a memory, and the parallel unit 51, the input unit 52, the output unit 53, the layout unit 54, and the like are stored in the memory as program units, and the corresponding functions are realized by the processor executing the program units stored in the memory.

[0110] The processor described above includes a kernel, and the corresponding program units are called from the memory by the kernel. The kernel can be one or more, and the interactive interface is generated by personalization of the kernel parameters.

[0111] The memory described above can include a non-permanent memory in a computer readable medium, a random access memory (RAM), and / or a non-volatile memory such as a read-only memory (ROM) or a flash memory (flash RAM), and the memory includes at least one memory chip.

[0112] According to another aspect of the embodiment of the present application, a computer readable storage medium including a stored computer program is also provided, wherein the computer readable storage medium controls the device where the computer readable storage medium is located to execute the generation method of the interactive interface of any one of the embodiments when the computer program runs.

[0113] According to another aspect of the embodiment of the present application, an electronic device is also provided, including one or more processors and a memory, and the memory is used to store one or more programs, wherein the one or more programs make the one or more processors realize the generation method of the interactive interface of any one of the embodiments when the one or more programs are executed by the one or more processors.

[0114] The present application also provides a computer program product adapted to execute the program of the following method steps when executed on a data processing device: performing parallel processing on the user function vector and the attribute feature vector to obtain a combined feature vector, inputting the combined feature vector into a fully connected neural network, performing feature fusion processing and dimension reduction processing on the combined feature vector by using the fully connected neural network to obtain a fusion feature vector, inputting the fusion feature vector and a random Gaussian noise into a layout generator, outputting an interface layout picture, performing layout on an interactive interface to be displayed on a terminal screen of a user terminal based on the interface layout picture, and generating a target interactive interface.

[0115] Figure 6is a hardware structural block diagram of an electronic device (or a mobile device) according to an embodiment of a method for generating an interactive interface of the present application. As shown in Figure 6 the electronic device can include one or more (shown in the figure as 602a, 602b, …, 602n) processors 602 (the processor 602 can include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 604 for storing data. In addition, it can also include a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which can be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply and / or a camera. Those skilled in the art can understand that Figure 6 the structure shown is only schematic, and does not limit the structure of the above-mentioned electronic device. For example, the electronic device can also include more or fewer components than Figure 6 shown, or have a different configuration from Figure 6 shown.

[0116] The above-mentioned serial numbers of the embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0117] In the above-mentioned embodiments of the present application, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0118] In several embodiments provided in the present application, it should be understood that the disclosed technical content can be implemented by other ways. Among them, the above-mentioned device embodiments are only schematic, for example, the division of the units can be a logical function division, and actual implementation can have another division way, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the coupling or direct coupling or communication connection between the units or modules shown or discussed can be indirect coupling or communication connection through some interfaces, units or modules, which can be electrical or other forms.

[0119] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or they can be distributed to multiple units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0120] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be realized in the form of hardware or software functional unit.

[0121] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, essentially or in other words, the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to make a computer device (which can be a personal computer, a server or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0122] The above is only the preferred embodiment of the present application, it should be pointed out that, for those skilled in the art, without departing from the principles of the present application, can make a number of improvements and refinements, these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A method of generating an interactive interface, characterized by, The method comprises the following steps: Parallel processing of a user function vector and an attribute feature vector to obtain a combined feature vector, wherein the user function vector corresponds to the functions required by the financial account selected by the user terminal, and the attribute feature vector corresponds to the account attributes of the associated financial account selected by the user terminal; Inputting the combined feature vector into a fully connected neural network, performing feature fusion processing and dimension reduction processing on the combined feature vector by using the fully connected neural network to obtain a fusion feature vector; Inputting the fusion feature vector and random Gaussian noise into a layout generator to output an interface layout picture; Based on the interface layout picture, the layout of the interactive interface to be displayed on the terminal screen of the user terminal is performed to generate a target interactive interface; Before parallel processing of the user function vector and the attribute feature vector, the method further comprises the following steps:

2. The generation method of claim 1, wherein, Receiving a function requirement set input by the user terminal, wherein the function requirement set at least contains N functions required by the financial account selected by the user terminal, each function required by the financial account corresponds to a function icon, and N is a positive integer greater than or equal to 1; randomly combining the function icons of the N functions required by the financial account; pasting the function icons of the N functions required by the financial account after random combination onto a specified background template to generate an initial function picture; and inputting the initial function picture into a picture classification model to output the user function vector, wherein the picture classification model contains a multi-layer convolutional neural network. Before parallel processing of the user function vector and the attribute feature vector, the method further comprises the following steps: Receiving an attribute selection set input by the user terminal, wherein the attribute selection set at least contains M account attributes of the associated financial account selected by the user terminal, each account attribute corresponds to an attribute embedding word table, and M is a positive integer greater than or equal to 1; Based on the attribute embedding word table corresponding to each account attribute, converting the category of the account attribute selected by the user terminal into an attribute feature; 3. The generation method of claim 1, wherein, Inputting each attribute feature into a multi-layer perceptron to perform fusion processing and encoding processing on all attribute features by using the multi-layer perceptron to obtain the attribute feature vector. The method further comprises the following steps: Obtaining a plurality of historical interface layout pictures and historical fusion feature vectors to obtain a training set; Inputting the historical interface layout pictures and the historical fusion feature vectors in the training set into a layout encoder to output a hidden space vector, wherein the layout encoder comprises a multi-layer convolutional neural network; Inputting the historical fusion feature vector and the random Gaussian noise into the layout generator to output a to-be-pushed interface layout picture, wherein the layout generator comprises a multi-layer deconvolutional neural network; The to-be-pushed interface layout picture and the random Gaussian noise form a first type of binary tuple, and the historical interface layout picture and the hidden space vector form a second type of binary tuple; Using a discriminator to distinguish the real picture in the first type of binary tuple and the second type of binary tuple to generate a loss function value; Based on the loss function value, adjusting the structure of each deconvolutional neural network in the layout generator; In a case where the loss function value is less than a preset function threshold, it is determined that the layout generator is trained completely.

4. The generation method of claim 1, wherein, After the interface layout picture is used to layout the interactive interface to be displayed on the terminal screen of the user terminal, the method further includes: receiving an interface illustration generation request, wherein the interface illustration generation request at least includes an illustration generation position and illustration text description information, the illustration generation position being any click position or any selected region of the user terminal on the interface layout picture; in response to the interface illustration generation request, inputting the illustration text description information into a layout interface illustration generation model to output at least one to-be-selected illustration; receiving an illustration selection operation of the user terminal; based on the illustration selection operation, replacing a target to-be-selected illustration selected by the user terminal with an interface layout picture on the interactive interface at the illustration generation position to generate a target interface illustration.

5. The generation method of claim 4, wherein, The layout interface illustration generation model is trained by the following method: collecting a plurality of historical interactive pictures to generate a picture training set; using a pre-trained picture encoder to encode each historical interactive picture in the picture training set into a hidden space vector, wherein the picture encoder includes a plurality of convolutional neural networks; adding the random Gaussian noise to the hidden space vector in K steps to obtain a target random noise, wherein K is a positive integer greater than or equal to 2; using a specified semantic segmentation model to restore the target random noise to output an illustration corresponding to the historical interactive picture, wherein the specified semantic segmentation model incorporates a text condition vector generated by a text encoder.

6. The generation method of claim 5, wherein, The step of restoring the target random noise using the specified semantic segmentation model includes: using a preset text classification strategy to incorporate the text condition vector into the specified semantic segmentation model; using the specified semantic segmentation model to estimate the noise corresponding to each step in the K steps to obtain a noise estimation result; based on the noise estimation result and the text condition vector, restoring the target random noise.

7. An apparatus for generating an interactive interface, characterized in that, includes: a parallel unit configured to parallel process a user function vector and an attribute feature vector to obtain a combined feature vector, wherein the user function vector corresponds to a function required by a financial account selected by a user terminal, and the attribute feature vector corresponds to an account attribute of an associated financial account selected by the user terminal; an input unit configured to input the combined feature vector into a fully connected neural network, and use the fully connected neural network to perform feature fusion processing and dimension reduction processing on the combined feature vector to obtain a fusion feature vector; an output unit configured to input the fusion feature vector and a random Gaussian noise into a layout generator to output an interface layout picture; a layout unit configured to use the interface layout picture to layout an interactive interface to be displayed on a terminal screen of the user terminal to generate a target interactive interface; The generation apparatus of the interactive interface further comprises: a first receiving module, configured to receive a function requirement set input by the user terminal, wherein the function requirement set at least comprises required functions of N financial accounts selected by the user terminal, each required function of a financial account corresponds to a function icon, and N is a positive integer greater than or equal to 1; a first combining module, configured to randomly combine the function icons of the required functions of the N financial accounts; a first generation module, configured to paste the function icons of the required functions of the N financial accounts after random combination onto a designated background template to generate an initial function picture; and a first input module, configured to input the initial function picture into a picture classification model to output the user function vector, wherein the picture classification model comprises a plurality of layers of convolutional neural networks.

8. A computer-readable storage medium, characterized in that, A computer readable storage medium comprises a stored computer program, wherein the computer program, when executed, controls a device in which the computer readable storage medium is located to perform the method for generating an interactive interface according to any one of claims 1 to 6.

9. An electronic device, comprising: An apparatus comprising one or more processors and memory storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the method for generating an interactive interface according to any one of claims 1 to 6.

Images