Constellation mapping model training method and device, electronic equipment and storage medium

Abstract

The invention provides a constellation mapping model training method and device, electronic equipment and a storage medium, and relates to the technical field of computers, in particular to the technical field of semantic communication. According to the specific implementation scheme, the method comprises the steps of obtaining original data, and converting the original data into semantic features; determining an entropy corresponding to the semantic feature according to the semantic feature; reordering the semantic features according to the entropy corresponding to the semantic features to obtain rearranged semantic features; mapping the rearranged semantic features into a communication constellation diagram to obtain mapped constellation points; and training a to-be-trained model according to the mapped constellation points to obtain a constellation mapping model. According to the method, the semantic features are sorted according to the entropy, then the rearranged semantic features are mapped to the communication constellation diagram, and finally the mapped constellation points are utilized to train the to-be-trained model, so that the obtained constellation mapping model can reflect and process the semantic content of the original data more accurately, and the accuracy of data processing is improved. And thus, more accurate and efficient information transmission is facilitated.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, specifically to the field of semantic communication technology, and in particular to a constellation mapping model training method, device, electronic device and storage medium. Background Art

[0002] In semantic communication systems, the constellation design of the modulation module is a key link, which involves mapping the continuous source encoder output to discrete constellation points. This step is crucial to achieve the combination of semantic communication and digital communication, because it allows semantic communication to operate effectively in traditional digital communication links, thereby improving the efficiency and accuracy of information transmission and promoting the development and innovation of communication technology.

[0003] However, there are some problems in the constellation design of the existing technology, especially when processing the output of the JSCC (Joint Source-Channel Coding) encoder, because its output distribution is usually uneven, while the traditional QAM (Quadrature Amplitude Modulation) constellation points are regularly and evenly distributed, which results in QAM not being the best modulation scheme, unable to fully utilize the semantic information of the source, and difficult to adapt to the dynamic changes of the channel, limiting the performance improvement of the communication system. Therefore, researching and developing a constellation design method that is more suitable for semantic communication is of great significance to improving the overall performance and adaptability of the communication system. Summary of the invention

[0004] The present invention provides a constellation mapping model training method, device, electronic device and storage medium.

[0005] According to one aspect of the present disclosure, a constellation mapping model training method is provided, the method comprising:

[0006] Acquire raw data, and convert the raw data into semantic features;

[0007] Determine, according to the semantic feature, an entropy corresponding to the semantic feature, wherein the entropy corresponding to the semantic feature represents the importance of the semantic feature;

[0008] Rearranging the semantic features according to the entropy corresponding to the semantic features to obtain rearranged semantic features;

[0009] Mapping the rearranged semantic features into a communication constellation diagram to obtain mapped constellation points;

[0010] The model to be trained is trained according to the mapped constellation points to obtain a constellation mapping model.

[0011] According to another aspect of the present disclosure, a constellation mapping model training device is provided, comprising:

[0012] An acquisition module, used for acquiring raw data and converting the raw data into semantic features;

[0013] A determination module, configured to determine an entropy corresponding to the semantic feature according to the semantic feature, wherein the entropy corresponding to the semantic feature indicates the importance of the semantic feature;

[0014] A rearrangement module, used for reordering the semantic features according to the entropy corresponding to the semantic features to obtain rearranged semantic features;

[0015] A mapping module, used for mapping the rearranged semantic features into a communication constellation diagram to obtain a mapped constellation point;

[0016] The training module is used to train the model to be trained according to the mapped constellation points to obtain the constellation mapping model.

[0017] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0018] at least one processor; and

[0019] a memory communicatively connected to the at least one processor; wherein,

[0020] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any method in any of the above technical solutions.

[0021] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute any one of the methods described in the above technical solutions.

[0022] According to a fifth aspect of the present disclosure, a computer program product is provided, including a computer program, wherein the computer program implements any one of the methods described in the above technical solutions when executed by a processor.

[0023] The present disclosure provides a constellation mapping model training method, device, equipment and storage medium. The present disclosure can capture the core semantic information of the original data by acquiring the original data and converting it into semantic features. Since entropy is used as a metric in this scheme to reflect the uncertainty and richness of feature information, the importance of each semantic feature can be quantified by calculating the entropy of these semantic features. After obtaining the entropy of the semantic features, the semantic features are reordered according to their corresponding entropies, which can ensure that the most important semantic features are processed first, which helps to highlight the most critical content in information transmission and processing. Then, the rearranged semantic features are mapped to the communication constellation diagram, which not only realizes the effective combination of semantic information and communication signals, but also optimizes the transmission efficiency and reliability of the signal through the precise representation of constellation points. Finally, these mapped constellation points are used to train the model to be trained, so that a constellation mapping model can be obtained, through which the semantic content contained in the original data can be more accurately reflected and processed, which is conducive to more accurate and efficient information transmission.

[0024] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0026] Figure 1 is a schematic diagram of the steps of the constellation mapping model training method in an embodiment of the present disclosure;

[0027] Figure 2 is a network architecture diagram corresponding to the constellation mapping model training method in one embodiment of the present disclosure;

[0028] Figure 3 is a schematic diagram of mapping semantic features to a constellation diagram in an embodiment of the present disclosure;

[0029] Figure 4 Principle block diagram of the constellation mapping model training device in the embodiment of the present disclosure;

[0030] Figure 5 It is a block diagram of an electronic device used to implement the constellation mapping model training method of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0031] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0032] This disclosure provides a constellation mapping model training method, see Figure 1 As shown, Figure 1 is a schematic diagram of the steps of the constellation mapping model training method in an embodiment of the present disclosure, the method comprising:

[0033] Step S101, obtaining original data and converting the original data into semantic features.

[0034] Specifically, the original data may include, for example, image data, video data, audio data, and text data. Taking image data as an example, when obtaining the original image, an image capture device (such as a digital camera or scanner) can be used to capture the image directly from the real world, or an existing image file can be retrieved and downloaded from a digital resource (such as a website, a database, a local storage device). These image files may exist in common formats, such as JPEG, PNG, BMP, etc. In this way, by obtaining the original image, it is ensured that the image data can be processed by the subsequent semantic encoder and converted into useful semantic features. It should be noted that in the embodiments of the present disclosure, only image data is taken as an example. Since the data processing methods of other types are similar to the data processing methods of the image type, those skilled in the art can adaptively adjust the data processing methods according to different data types according to actual needs, so they will not be elaborated here.

[0035] After obtaining the original image, the original image is converted into semantic features. When converting the original image into semantic features, the original image can be input into a well-trained semantic encoder, and through a series of convolutional layers, pooling layers, and fully connected layers, key information that can represent the image content is extracted, such as edges, textures, shapes, and objects. Finally, a set of feature vectors (also called semantic features) that can reflect the semantic information of the image can be obtained. These vectors capture the high-level semantic concepts of the image and provide rich information for subsequent analysis and processing.

[0036] In this way, expressive information can be extracted from the image by converting the original image into semantic features. These features not only capture the basic visual elements of the image, but also contain deeper semantic content, such as the context of the scene, the category and attributes of the object, etc. This conversion enables computers to better understand and process images, providing strong support for applications such as image recognition, classification, retrieval and analysis, while also improving the efficiency of image data compression, storage and transmission, allowing more semantic information to be transmitted under limited bandwidth and storage conditions.

[0037] Step S102: determining an entropy corresponding to the semantic feature according to the semantic feature, wherein the entropy corresponding to the semantic feature represents the importance of the semantic feature.

[0038] Specifically, after obtaining the semantic features, entropy calculation is performed on the semantic features. When entropy calculation is performed, the probability distribution of the semantic features is analyzed, and the entropy corresponding to the semantic features can be obtained by calculating the probability of occurrence of each feature value. Among them, the so-called entropy is a measure of the uncertainty or confusion of semantic information, which can help understand the amount of information contained in the semantic features. In this way, by determining the entropy corresponding to these features according to the semantic features, the information value and uncertainty of each feature can be quantified, which not only enhances the understanding of the data content, but also optimizes the transmission and processing of information. In this way, the key information in the data can be highlighted through the identification of high entropy features, so that the system can give priority to processing and protecting the most semantically valuable parts, thereby bringing significant performance improvements in data compression, transmission efficiency, and data recovery under noise interference.

[0039] Step S103: reorder the semantic features according to the entropies corresponding to the semantic features to obtain reordered semantic features.

[0040] Specifically, after obtaining the entropy corresponding to the semantic feature, the semantic features can be reordered according to the calculated entropy value. Sorting the semantic features according to the entropy value can help identify which semantic features contain more uncertainty or information, thereby facilitating the selection of important semantic features in subsequent processing.

[0041] Step S104, mapping the rearranged semantic features into the communication constellation diagram to obtain mapped constellation points.

[0042] Specifically, the communication constellation refers to a diagram used in communication principles to represent the amplitude and phase relationship of a signal. After obtaining the rearranged semantic features, these features are mapped to the corresponding positions of the constellation according to a specific modulation rule (such as QAM or PSK), where each constellation point represents a specific set of signal states, including amplitude and phase information. This process is achieved by converting the feature values ​​into coordinate points on the constellation, so that each semantic feature corresponds to a unique constellation point, which can then be used for signal modulation and transmission to ensure the effective transmission of information in the communication system.

[0043] Step S105, training the model to be trained according to the mapped constellation points to obtain a constellation mapping model.

[0044] Specifically, in order to train the training model according to the mapped constellation points and obtain the constellation mapping model, it is first necessary to collect or generate a series of labeled constellation point data, where the label represents the original semantic feature or expected output corresponding to each constellation point; then, these data are input into an appropriate machine learning model, such as a neural network, and the model parameters are adjusted through supervised learning to minimize the difference between the predicted output and the actual label; during the training process, the model learns how to accurately predict or classify according to the input constellation point features, and finally forms a constellation mapping model that can map the constellation points back to their semantic features, which can be used for signal demodulation and semantic information recovery in communication systems.

[0045] The present disclosure provides a constellation mapping model training method, device, equipment and storage medium. The present disclosure can capture the core semantic information of the original data by acquiring the original data and converting it into semantic features. Since entropy is used as a metric in this scheme to reflect the uncertainty and richness of feature information, the importance of each semantic feature can be quantified by calculating the entropy of these semantic features. After obtaining the entropy of the semantic features, the semantic features are reordered according to their corresponding entropies, which can ensure that the most important semantic features are processed first, which helps to highlight the most critical content in information transmission and processing. Then, the rearranged semantic features are mapped to the communication constellation diagram, so that not only the effective combination of semantic information and communication signals is achieved, but also the transmission efficiency and reliability of the signal are optimized through the precise representation of constellation points. Finally, these mapped constellation points are used to train the model to be trained, so that a constellation mapping model can be obtained, through which the semantic content contained in the original data can be more accurately reflected and processed, which is conducive to more accurate and efficient information transmission.

[0046] In some optional embodiments, determining the entropy corresponding to the semantic feature according to the semantic feature includes:

[0047] The semantic features are input into the entropy model, and the entropy corresponding to the semantic features is obtained through the entropy model.

[0048] Specifically, the entropy model can be an algorithm based on machine learning, which calculates the entropy value of each semantic feature by analyzing the distribution and change of the features. For each semantic feature, the entropy model will output an entropy value, which can be regarded as the weight of the feature, indicating their importance in the image semantics. After obtaining the semantic features, the obtained semantic features are input into the entropy model, so that the entropy corresponding to the semantic features can be obtained. After obtaining the entropy corresponding to each semantic feature, this information is used to guide subsequent signal processing steps, such as modulation and transmission.

[0049] In this way, by inputting semantic features into the entropy model and obtaining the entropy corresponding to these features, Tongshang Entropy can quantify the information value and uncertainty of each semantic feature. This process not only enhances the understanding of data content, but also optimizes the transmission and processing of information. In addition, the entropy value can identify key information in the data, allowing the system to prioritize and protect the most semantically valuable parts, thereby bringing significant performance improvements in data compression, transmission efficiency, and data recovery under noise interference.

[0050] In some optional embodiments, the semantic features are reordered according to the entropies corresponding to the semantic features to obtain the reordered semantic features, including:

[0051] Perform dimension reduction processing on the entropy corresponding to the semantic feature to obtain the entropy after dimension reduction;

[0052] Sort the entropy after dimensionality reduction according to the last dimension in the entropy after dimensionality reduction to obtain the entropy index;

[0053] The semantic features are sorted according to the value of the entropy index to obtain the rearranged semantic features.

[0054] Specifically, the process of re-ranking semantic features involves matching features with their entropy values ​​and simplifying the entropy values ​​through dimensionality reduction techniques so that they are easier to handle. Specifically, the entropy of each semantic feature, which is a measure of its information content, is first identified, and then these entropy values ​​are converted into a more manipulable form through dimensionality reduction. Next, the entropy values ​​after dimensionality reduction are sorted according to their values ​​in the last dimension to generate an entropy index that reflects the importance of each feature. Finally, this entropy index is used to re-rank the original semantic features, ensuring that the most important features are ranked first, thereby obtaining a set of semantic features sorted by importance, which helps to more effectively utilize key information in subsequent data processing and analysis.

[0055] To facilitate understanding of the solution of the disclosed embodiments, an example is given below: assuming that the semantic feature is input into the entropy model to obtain the quantized semantic feature and the entropy corresponding to each semantic feature element, the dimensions of the semantic feature and the entropy are both (B, C, H, W), B represents the batch size, C represents the number of channels, H and W represent the height and width of the tensor respectively, and then the dimensions of the semantic feature and the entropy are reshaped to (B, C*H*W), sorted in the last dimension of the entropy to obtain the entropy index, and each element of the semantic feature is rearranged according to the entropy index.

[0056] In this way, by reducing the dimension and sorting the entropy corresponding to the semantic features, the most important semantic information can be effectively identified and highlighted. This process simplifies the data structure through dimensionality reduction, so that high entropy features (i.e., those that carry more uncertainty and information) are given priority, and the priority of feature processing is ensured through entropy index sorting, thereby optimizing the efficiency and accuracy of data transmission.

[0057] In some optional embodiments, mapping the rearranged semantic features to a communication constellation diagram to obtain mapped constellation points includes:

[0058] Reshape the dimensions of the rearranged semantic features to obtain semantic features after the reshape dimensions;

[0059] According to the distance between the semantic features after reshaping the dimension and the communication constellation diagram, the rearranged semantic features are mapped to the communication constellation diagram to obtain the mapped constellation points.

[0060] Specifically, the process of mapping the rearranged semantic features to the communication constellation diagram first involves reshaping the dimensions of these features and converting them into a form suitable for constellation diagram representation. For example, the dimensions of the rearranged semantic features are reshaped to (B, C*H*W / 2, 2), where the last dimension corresponds to the real and imaginary parts of the constellation respectively. Subsequently, by calculating the distance between each reshaped semantic feature and each point in the constellation diagram, the constellation point with the smallest distance is selected as the mapping target, so that the semantic features are accurately mapped to the constellation diagram. This mapping process ensures the efficient representation of semantic information on the constellation diagram, so that each constellation point can retain the key information of the original semantic features as much as possible, providing an optimized signal state for subsequent signal transmission and reception, and improving the performance and transmission efficiency of the communication system. Among them, the distance between the reshaped semantic features and the communication constellation diagram can be the Euclidean distance, the Manhattan distance, or other distances.

[0061] In this way, the semantic features are first reshaped to fit the structure of the constellation, and then mapped according to the principle of minimum distance between the semantic features and the constellation points, ensuring that each semantic feature corresponds to an optimal point on the constellation. This mapping method not only improves the transmission efficiency of the signal, but also enhances the signal's anti-interference ability, because it reduces the bit error rate by minimizing the distance between signals, thereby achieving more reliable data transmission in the communication system. In addition, this method allows the system to optimize the use of the constellation according to the importance of the semantic features, further improving the quality of communication.

[0062] In some optional embodiments, training a to-be-trained model according to the mapped constellation points to obtain a constellation mapping model includes:

[0063] Get the model to be trained;

[0064] The mapped constellation points are input into the model to be trained, and the constellation mapping model is obtained through the model to be trained.

[0065] Specifically, the process of using the mapped constellation points to train the model to be trained involves first obtaining the initial state of the model, and then providing these constellation points as input data to the model so that the model can learn how to accurately map semantic features to the constellation diagram. Through this training, the model gradually adjusts its internal parameters to minimize the mapping error and optimize the performance. Ultimately, the trained model, namely the constellation mapping model, can accurately convert the input semantic features into points on the constellation diagram, thereby providing an effective signal representation method for the communication system, which helps to improve the transmission quality of the signal and the overall performance of the system.

[0066] Furthermore, the process of training the to-be-trained model according to the mapped constellation points to obtain the constellation mapping model refers to first obtaining a model that has not yet been trained, which may be a deep learning network or other types of machine learning models. Then, the constellation points mapped by a specific algorithm are input into the model as training data. In this way, the model can learn how to optimally map the input semantic features to the constellation diagram, thereby improving the accuracy of signal modulation and the overall performance of the communication system. During the training process, the model continuously adjusts its internal parameters to reduce the difference between the predicted output and the actual constellation points, and finally forms a model that can accurately perform constellation mapping, namely the constellation mapping model. This process is crucial to improving the efficiency and reliability of communication systems when transmitting data.

[0067] In this way, the model to be trained is first obtained, and then the constellation points are input into the model to be trained as training data, so that the model to be trained can optimize its parameters and performance by learning these data. This can significantly improve the model's ability to map semantic features to the constellation diagram, thereby achieving more efficient and accurate signal modulation and demodulation in the communication system, and enhancing the signal transmission quality and system reliability. In addition, this training also helps the model adapt to different communication environments and conditions, so that it can perform well in a variety of application scenarios.

[0068] In some optional embodiments, the mapped constellation points are input into a model to be trained, and a constellation mapping model is obtained through the model to be trained, including:

[0069] Reconstruct the data according to the rearranged semantic features to obtain reconstructed data;

[0070] The loss value between the reconstructed data and the original data is calculated, and the constellation parameters of the model to be trained are adjusted according to the loss value until the model converges to obtain the constellation mapping model.

[0071] Specifically, the mapped constellation points are input into the model to be trained, and the constellation mapping model is obtained through this process, which involves using these constellation points to reconstruct the data, that is, converting the semantic features back to their original visual representation through the model. After the reconstructed data is generated, the model calculates the difference between it and the original data, that is, the loss value. This loss value is a key indicator to measure the quality of the reconstructed data, and the model uses it to guide the adjustment of parameters, with the aim of minimizing the difference between the reconstructed data and the original data. By continuously iterating this process, the model to be trained gradually optimizes its constellation parameters until the performance reaches a stable state, that is, the model converges, so that the constellation mapping model can be obtained, and the irregular constellation points corresponding to the original data can be obtained through the constellation mapping model. Taking image data as an example, the image is reconstructed according to the rearranged semantic features to obtain a reconstructed image. Then, the loss value between the reconstructed image and the original image is calculated, and the constellation parameters of the model to be trained are adjusted according to the loss value until the model converges to obtain the constellation mapping model. Finally, the obtained constellation mapping model can accurately map the semantic features back to a high-quality image, which has important application value in the fields of image transmission, compression and reconstruction.

[0072] In this way, the data is reconstructed using the rearranged semantic features, and then the model parameters are fine-tuned by comparing the loss value between the reconstructed data and the original data. This iterative optimization based on the loss value ensures that the model can gradually reduce the reconstruction error until convergence, thereby obtaining an efficient constellation mapping model. This can significantly improve the quality and accuracy of data reconstruction, optimize the modulation process of the signal, and enhance the reliability and efficiency of the communication system during data transmission.

[0073] In some optional embodiments, see Figure 2 , Figure 2 This is a network architecture diagram corresponding to the constellation mapping model training method in one embodiment of the present disclosure. In this network architecture, taking the data type as an image type as an example, the original image is first obtained at the transmitting end, and then the original image is output as a semantic feature through a semantic encoder. Next, the semantic feature is input into the entropy model to obtain the quantized semantic feature and the entropy corresponding to each semantic feature element, and the semantic feature is reordered according to the entropy corresponding to the semantic feature to obtain the rearranged semantic feature, and the rearranged semantic feature is mapped to the communication constellation map to obtain the mapped constellation point, which is then normalized for channel transmission. At this time, the entropy is also transmitted to the receiving end. After receiving the semantic feature, the receiving end restores the order according to the entropy index and sends it to the semantic decoder to reconstruct the image. The irregular constellation points that can be learned by this method are better than the traditional constellation mapping method.

[0074] To understand the process of mapping semantic features to constellation diagrams, see Figure 3 , Figure 3 It is a schematic diagram of mapping semantic features to a constellation diagram in an embodiment of the present disclosure, which shows the process of mapping semantic features to a constellation diagram. First, the original data (such as an image, a video frame or an audio signal) is obtained, and then the original data is sent to a semantic encoder, which is responsible for extracting the key semantic information of the data and converting it into a set of semantic features. These semantic features may include quantized representations of information such as color, texture, shape, and motion. Next, the constellation diagram is prepared. In digital communication, the constellation diagram is a tool used to represent the signal modulation method. Each point (constellation point) in the constellation diagram represents a possible signal state, which is used to encode information during the modulation process. Then, the entropy corresponding to the semantic feature is determined according to the semantic feature, and the semantic feature is reordered according to the entropy corresponding to the semantic feature to obtain the rearranged semantic feature. Next, feature mapping is performed to map the extracted semantic features to the constellation diagram. This process may involve converting a multi-dimensional feature vector into a point on the constellation diagram. The mapping rule can be based on minimizing the error, maximizing the Euclidean distance between signals, or other optimization criteria.

[0075] The following describes an apparatus embodiment of the present application, which can be used to execute the constellation mapping model training method in the above embodiment of the present application. For details not disclosed in the apparatus embodiment of the present application, please refer to the above embodiment of the constellation mapping model training method of the present application.

[0076] The present disclosure also provides a constellation mapping model training device 400, such as Figure 4 As shown, including:

[0077] The acquisition module 401 is used to acquire original data and convert the original data into semantic features;

[0078] A determination module 402 is used to determine an entropy corresponding to the semantic feature according to the semantic feature, wherein the entropy corresponding to the semantic feature indicates the importance of the semantic feature;

[0079] A rearrangement module 403, configured to reorder the semantic features according to the entropies corresponding to the semantic features to obtain rearranged semantic features;

[0080] A mapping module 404 is used to map the rearranged semantic features into a communication constellation diagram to obtain a mapped constellation point;

[0081] The training module 405 is used to train the to-be-trained model according to the mapped constellation points to obtain a constellation mapping model.

[0082] In some optional embodiments, the determining module 402 determines the entropy corresponding to the semantic feature according to the semantic feature, including:

[0083] The semantic features are input into the entropy model, and the entropy corresponding to the semantic features is obtained through the entropy model.

[0084] In some optional embodiments, the rearrangement module 403 reorders the semantic features according to the entropies corresponding to the semantic features to obtain rearranged semantic features, including:

[0085] Perform dimension reduction processing on the entropy corresponding to the semantic feature to obtain the entropy after dimension reduction;

[0086] Sort the entropy after dimensionality reduction according to the last dimension in the entropy after dimensionality reduction to obtain the entropy index;

[0087] The semantic features are sorted according to the value of the entropy index to obtain the rearranged semantic features.

[0088] In some optional embodiments, the mapping module 404 maps the rearranged semantic features to the communication constellation diagram to obtain mapped constellation points, including:

[0089] Reshape the dimensions of the rearranged semantic features to obtain semantic features after the reshape dimensions;

[0090] According to the distance between the semantic features after reshaping the dimension and the communication constellation diagram, the rearranged semantic features are mapped to the communication constellation diagram to obtain the mapped constellation points.

[0091] In some optional embodiments, the training module 405 trains the to-be-trained model according to the mapped constellation points to obtain a constellation mapping model, including:

[0092] Get the model to be trained;

[0093] The mapped constellation points are input into the model to be trained, and the constellation mapping model is obtained through the model to be trained.

[0094] In some optional embodiments, the training module 405 inputs the mapped constellation points into the model to be trained, and obtains the constellation mapping model through the model to be trained, including:

[0095] Reconstruct the data according to the rearranged semantic features to obtain reconstructed data;

[0096] The loss value between the reconstructed data and the original data is calculated, and the constellation parameters of the model to be trained are adjusted according to the loss value until the model converges to obtain the constellation mapping model.

[0097] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0098] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0099] Figure 5 A schematic block diagram of an example electronic device 500 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0100] like Figure 5 As shown, the electronic device 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0101] A number of components in the device 500 are connected to the I / O interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the device 500 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0102] The computing unit 501 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 501 performs the various methods and processes described above, such as the constellation mapping model training method. For example, in some embodiments, the constellation mapping model training method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 500 via ROM 502 and / or communication unit 509. When the computer program is loaded into RAM 503 and executed by the computing unit 501, one or more steps of the applet distribution described above may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to perform the constellation mapping model training method in any other appropriate manner (e.g., by means of firmware).

[0103] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0104] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0105] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0106] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0107] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communications network). Examples of communications networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0108] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0109] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0110] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A constellation mapping model training method, the method comprising: Acquire raw data, and convert the raw data into semantic features; Determine, according to the semantic feature, an entropy corresponding to the semantic feature, wherein the entropy corresponding to the semantic feature represents the importance of the semantic feature; Rearranging the semantic features according to the entropy corresponding to the semantic features to obtain rearranged semantic features; Mapping the rearranged semantic features into a communication constellation diagram to obtain mapped constellation points; The model to be trained is trained according to the mapped constellation points to obtain a constellation mapping model.

2. The method according to claim 1, wherein: The determining, according to the semantic feature, the entropy corresponding to the semantic feature comprises: The semantic feature is input into an entropy model, and the entropy corresponding to the semantic feature is obtained through the entropy model.

3. The method according to claim 1, wherein: The step of reordering the semantic features according to the entropies corresponding to the semantic features to obtain the reordered semantic features includes: Performing dimensionality reduction processing on the entropy corresponding to the semantic feature to obtain entropy after dimensionality reduction; Sorting the entropy after dimensionality reduction according to the last dimension in the entropy after dimensionality reduction to obtain an entropy index; The semantic features are sorted according to the values ​​of the entropy indexes to obtain rearranged semantic features.

4. The method according to claim 1, wherein: Mapping the rearranged semantic features into a communication constellation diagram to obtain mapped constellation points includes: Reshaping the dimensions of the rearranged semantic features to obtain semantic features after the dimensions have been reshaped; The rearranged semantic features are mapped to the communication constellation diagram according to the distance between the reshaped dimensioned semantic features and the communication constellation diagram to obtain mapped constellation points.

5. The method according to any one of claims 1 to 4, wherein: The training of the model to be trained according to the mapped constellation points to obtain the constellation mapping model includes: Get the model to be trained; The mapped constellation points are input into the model to be trained, and a constellation mapping model is obtained through the model to be trained.

6. The method according to claim 5, wherein: The step of inputting the mapped constellation points into a model to be trained, and obtaining a constellation mapping model through the model to be trained, comprises: Reconstructing data according to the rearranged semantic features to obtain reconstructed data; A loss value between the reconstructed data and the original data is calculated, and the constellation parameters of the model to be trained are adjusted according to the loss value until the model converges to obtain the constellation mapping model.

7. A constellation mapping model training device, comprising: An acquisition module, used for acquiring raw data and converting the raw data into semantic features; A determination module, configured to determine an entropy corresponding to the semantic feature according to the semantic feature, wherein the entropy corresponding to the semantic feature indicates the importance of the semantic feature; A rearrangement module, used for reordering the semantic features according to the entropy corresponding to the semantic features to obtain rearranged semantic features; A mapping module, used for mapping the rearranged semantic features into a communication constellation diagram to obtain a mapped constellation point; The training module is used to train the model to be trained according to the mapped constellation points to obtain the constellation mapping model.

8. The device according to claim 7, wherein: The determining module determines the entropy corresponding to the semantic feature according to the semantic feature, including: The semantic feature is input into an entropy model, and the entropy corresponding to the semantic feature is obtained through the entropy model.

9. The device according to claim 7, wherein: The rearrangement module reorders the semantic features according to the entropy corresponding to the semantic features to obtain rearranged semantic features, including: Performing dimensionality reduction processing on the entropy corresponding to the semantic feature to obtain entropy after dimensionality reduction; Sorting the entropy after dimensionality reduction according to the last dimension in the entropy after dimensionality reduction to obtain an entropy index; The semantic features are sorted according to the values ​​of the entropy indexes to obtain rearranged semantic features.

10. The device according to claim 7, wherein: The mapping module maps the rearranged semantic features into a communication constellation diagram to obtain mapped constellation points, including: Reshaping the dimensions of the rearranged semantic features to obtain semantic features after the dimensions have been reshaped; The rearranged semantic features are mapped to the communication constellation diagram according to the distance between the reshaped dimensioned semantic features and the communication constellation diagram to obtain mapped constellation points.

11. The device according to any one of claims 7 to 10, wherein: The training module trains the model to be trained according to the mapped constellation points to obtain a constellation mapping model, including: Get the model to be trained; The mapped constellation points are input into the model to be trained, and a constellation mapping model is obtained through the model to be trained.

12. The device according to claim 11, wherein The training module inputs the mapped constellation points into the model to be trained, and obtains a constellation mapping model through the model to be trained, including: Reconstructing data according to the rearranged semantic features to obtain reconstructed data; A loss value between the reconstructed data and the original data is calculated, and the constellation parameters of the model to be trained are adjusted according to the loss value until the model converges to obtain the constellation mapping model.

13. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 6.

14. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-6.

15. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images