Semantic communication method and system based on LSTM network automatic training

Abstract

The invention discloses a semantic communication method and system based on LSTM network automatic training, and the method comprises the steps: constructing an LSTM network as a main architecture of a meta optimizer, and enabling the input of the meta optimizer to pass through two different weight matrixes, and then carrying out the activation through a Relu function; the processed input respectively enters a forgetting gate, an input gate and an output gate of the LSTM network to obtain the output of the LSTM network; processing the obtained output through a weight matrix and a limiting factor to obtain a predicted learning rate value; and the semantic communication main task optimizes the parameters of the semantic codec by using the learning rate predicted by the meta optimizer to complete semantic communication. According to the method, the semantic codec can be automatically trained to meet the convergence requirement, the technical problem that universal training and actual large-scale deployment cannot be achieved due to heterogeneous factors such as semantic communication tasks, data types, data modes and codec networks is solved, and the generalization ability of the semantic codec to different semantic communication tasks is improved.

Description

Background Art

[0002] As a new communication paradigm, semantic communication mainly uses artificial intelligence technology to perform semantic extraction, transmission and reconstruction of transmitted information. It can not only restore the original information semantics when the signal-to-noise ratio is relatively low, but also greatly compress the data volume to solve the problem of future 6G data expansion.

[0003] According to the specific process of semantic communication, the sender first performs semantic coding and channel coding on the information. Among them, semantic coding is to extract the semantics contained in the information, and channel coding is to compress the data volume and reduce the impact of poor channel conditions on the transmission of information. The specific expression is as follows: (1) in, Indicates that the parameter is The channel encoder, Indicates that the parameter is The semantic encoder It represents the information to be transmitted. The encoded information contains sufficient semantics and has a certain robustness to noise.

[0004] After that, the encoded information will be propagated through the wireless channel, which is represented as follows: (2) in, is the channel coefficient, It is additive white Gaussian noise AWGN, that is, the receiving end receives information contaminated by noise.

[0005] Finally, the receiving end extracts and semantically reconstructs the received information through the channel decoder and semantic decoder, which is specifically expressed as follows: (3) in, Indicates that the parameter is The semantic decoder of Indicates that the parameter is The channel decoder, Indicates the decoded information.

[0006] In summary, a pair of encoders and decoders for a semantic communication system consists of four sets of parameters: These parameters are obtained through training based on the knowledge background shared by both senders and receivers. The traditional training process not only requires setting a specific network structure for the codec according to the main task of semantic communication, but also requires comprehensive consideration of data types (such as text, images, audio and video, etc.), data features (such as text form, image resolution, video clarity, etc.), main tasks of semantic communication (such as image transmission, image classification, etc.), codec network structure and other factors to set a reasonable optimizer and learning rate strategy for training.

[0007] However, when the number of participants in semantic communication increases, each pair of users will have differences in data characteristics of semantic information, data modalities of semantic information, main tasks of semantic communication, codec architecture, etc. The semantic communication service provider or maintainer will greatly consume training resources and manpower costs if they set up optimizers and learning rate strategies for each pair of users, making it difficult to achieve large-scale deployment of semantic communication systems.

[0008] At the same time, existing research often focuses on the design of end-to-end semantic communication architecture, which is not conducive to the large-scale application of semantic communication devices, and cannot meet the future vision of "Internet of Everything" and "Intelligent Internet of Everything". Summary of the invention

[0009] The technical problem to be solved by the present invention is to provide a semantic communication method and system based on LSTM network automatic training in response to the deficiencies in the above-mentioned prior art, so as to solve the technical problems that the semantic communication system currently consumes huge training resources and manpower costs, as well as the inability to perform universal training and actual large-scale deployment due to heterogeneous factors such as semantic communication tasks, data types, data modalities, codec networks, etc.

[0010] The present invention adopts the following technical solutions: A semantic communication method based on LSTM network automatic training includes the following steps: S1. Build an LSTM network as the main architecture of the meta-optimizer. The input of the meta-optimizer is activated by the ReLU function after passing through two different weight matrices. S2, the input processed in step S1 enters the forget gate, input gate, and output gate of the LSTM network respectively to obtain the output of the LSTM network; S3, the output obtained in step S2 is processed by the weight matrix and the restriction factor to obtain the predicted learning rate value; S4, the main task of semantic communication, optimizes the parameters of the semantic codec using the learning rate predicted by the meta-optimizer in step S3 to complete the semantic communication.

[0011] Preferably, in step S1, the input of the meta-optimizer is the training loss value of the semantic communication task and the historical information of the previous stage. The training process of the semantic communication task is as follows: S101, initialize the encoder-decoder parameters of the semantic communication system, the weight matrix parameters of the meta-optimizer, the initial learning rate, the cell parameters of the meta-optimizer main LSTM, and set the number of meta-learning rounds T; S102, randomly dividing the training knowledge background of semantic communication into a main task knowledge background and a meta-learning knowledge background according to a set ratio; S103, the input of the semantic communication task is randomly sampled from the main task knowledge background and data preprocessing is performed; S104, taking the data pre-processed in step S103 as input, sequentially entering into the semantic encoder, channel encoder, wireless channel, channel decoder and semantic decoder, and obtaining the output of the main task of semantic communication; S105: After the training loss is calculated using the input and output, a gradient descent is performed according to the initial learning rate to optimize the codec parameters ; S106. After each T rounds of training, the training loss value of the current round is input into the meta-optimizer. The meta-optimizer uses the meta-learning knowledge background to perform the calculation of requirement 1, predicts and records the learning rate that should be taken for the current semantic communication main task, and updates the parameters of the meta-optimizer to ; S107, update the learning rate of the main task to the learning rate predicted by the meta-optimizer, and repeat steps S102-S107 until the training loss in step S105 no longer decreases or reaches the total number of training rounds.

[0012] Preferably, update the parameters of the meta-optimizer for:

[0013] in, Represents the Adam optimizer; LR for Adam optimizer; It is the predicted value of the semantic communication codec parameters at the next moment after calculation using the learning rate estimated by the meta-optimizer that has not yet been updated.

[0014] Preferably, the predicted value of the semantic communication codec parameter at the next moment after calculating the learning rate estimated by the meta-optimizer that has not yet been updated for:

[0015] in, is the input-output relationship of the meta-optimizer, is the Loss value of the input meta-optimizer, is the cell state of the meta-optimizer (including historical training information), is the parameter information of the meta-optimizer, is the gradient descent information of the Loss value about the parameter w on the semantic communication main task network.

[0016] Preferably, the codec parameters For:

[0017] in, is the input-output relationship of the meta-optimizer, is the Loss value of the input meta-optimizer, is the cell state of the meta-optimizer (including historical training information), is the parameter information of the meta-optimizer at the next moment, is the gradient descent information of the Loss value about the parameter w on the semantic communication main task network.

[0018] Preferably, in step S3, the predicted learning rate value is specifically: The two inputs pass through a fully connected layer with a ReLU activation function Do the preprocessing, then import into LSTM and get the output ;use The fully connected layer with activation function right Perform linear transformation to obtain the predicted value ; Finally, multiply the predicted value by the scaling factor Ensure that the final predicted learning rate value is within interval.

[0019] Preferably, in step S4, an online approximation technique is used to jointly update the meta-optimizer parameters and semantic communication codec parameters , will be updated in Several steps ( ) after update Once; and divide the overall knowledge background of semantic communication into the main task knowledge background of semantic communication and the meta-learning knowledge background, and The training of parameters is performed on their respective knowledge backgrounds.

[0020] Preferably, in step S4, in the semantic communication main task and meta-learning task stages, the channel adaptive training method is specifically as follows: Assume that the semantic communication task acts in an actual scenario with a signal-to-noise ratio range of mdB to ndB, and n is greater than or equal to m; set the channel signal-to-noise ratio of the main task of semantic communication to mdB, and adopt the main task knowledge background for training; in the meta-learning stage, adjust the channel signal-to-noise ratio to ndB and adopt the meta-learning knowledge background; the main task knowledge background will be much larger than the meta-learning knowledge background, and the meta-learning stage will only be carried out after multiple rounds of training.

[0021] Preferably, the parameters of the semantic codec for:

[0022]

[0023] in, is the updated codec parameter for the next moment. is the Loss value in the context of verification knowledge, is the input-output relationship of the meta-optimizer, is the input-output relationship of the gradient descent function, is the total number of training rounds.

[0024] In a second aspect, an embodiment of the present invention provides a semantic communication system based on LSTM network automatic training, including: Build the module and construct the LSTM network as the main architecture of the meta-optimizer. The input of the meta-optimizer is activated by the ReLU function after passing through two different weight matrices. Network module, the input processed by the building module enters the forget gate, input gate, and output gate of the LSTM network respectively to obtain the output of the LSTM network; The prediction module processes the output through the weight matrix and the restriction factor to obtain the predicted learning rate value; Optimization module, the main task of semantic communication will use the learning rate predicted by the meta-optimizer to optimize the parameters of the semantic encoder and decoder to complete the semantic communication.

[0025] In a third aspect, a computer device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned semantic communication method based on automatic training of LSTM network when executing the computer program.

[0026] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, comprising a computer program, which, when executed by a processor, implements the steps of the above-mentioned semantic communication method based on automatic training of an LSTM network.

[0027] Compared with the prior art, the present invention has at least the following beneficial effects: A semantic communication method based on automatic training of LSTM network makes full use of the unidirectional sequence memory ability of LSTM network to calculate and weight the historical training information and the current training information inside it to cope with the dynamic changes of the training process. In addition, as a meta-optimizer component, LSTM network is lightweight and plug-and-play, isolated from the main task of semantic communication, meeting the design concept of high cohesion and low coupling.

[0028] Furthermore, the knowledge background of semantic communication is divided into training knowledge background and verification knowledge background. The specific division ratio is related to the channel quality conditions, the size of the knowledge background and the semantic communication task. This method adopts a division ratio of 9:1. In addition, this method adopts a meta-learning scheme to update the meta-optimizer parameters every T rounds. The setting of the T parameter is related to the training cost, training performance, the size of the knowledge background, etc. This method sets T to the number of iterations required to train a complete round of knowledge background to balance the training cost and training performance.

[0029] Furthermore, the fully connected layer W1 can concatenate and transform the Loss value and historical information to meet the input dimension requirements of the LSTM network; the LSTM network makes full use of its unidirectional sequence information processing capabilities to combine the current training information and historical training information in a weighted manner to generate an output cache value; the fully connected layer W3 transforms the output cache dimension to meet the output format requirements of a single learning rate floating point value; the y factor is selectively set. If there is prior knowledge of training, it can be set to ensure that the final predicted learning rate value is within .

[0030] Furthermore, the online approximation technology can reduce the communication delay between the semantic communication system process and the meta-optimizer training process. The loss value of the semantic communication system process will be directly transmitted to the meta-optimizer training process through the shared memory. Compared with the traditional offline technology, this method reduces the time and space cost of reading and writing the cache. In addition, setting the number of meta-learning rounds and the division strategy of the knowledge background helps to reduce the consumption of training resources and meet the adaptive signal-to-noise ratio requirements described in claim 8.

[0031] Furthermore, the study has found that training in high SNR areas is easier to converge, while training in low SNR areas is more difficult to converge due to greater noise. Therefore, this method sets the channel SNR of the main task of semantic communication to mdB to meet the robustness of the semantic communication system in low SNR areas, and adopts the main task knowledge background for training; in the meta-learning stage, the channel SNR is adjusted to ndB and the meta-learning knowledge background is adopted, so that the semantic communication main task will use the learning rate predicted by ASCNet to make the codec parameters more biased towards the high SNR area. This method enables the codec to have higher performance in the entire SNR range rather than tending to a certain SNR value; therefore, this also explains the division strategy of the main task knowledge background and the meta-learning knowledge background (that is, the former is much greater than the latter), and the meta-learning stage is only conducted after multiple rounds of training, which will make full use of training resources, reduce training costs and meet the essential concept of meta-learning.

[0032] It can be understood that the beneficial effects of the second aspect mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0033] In summary, the present invention optimizes the training of the semantic communication codec so that it can automatically train the semantic codec and meet the convergence requirements without manual parameter adjustment. At the same time, the meta-optimizer uses the knowledge background to update parameters, which can solve the technical problems of inability to universalize training and actual large-scale deployment due to heterogeneous factors such as semantic communication tasks, data types, data modalities, and codec networks, and improve its generalization ability for different semantic communication tasks.

[0034] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings to be used in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 Automated training scenarios for semantic communication that power meta-learning optimizers; Figure 2 This is a simulation diagram of the change of Loss with epoch in the present invention; Figure 3 This is a performance simulation diagram of the present invention at 0dB signal-to-noise ratio; Figure 4 This is a performance simulation diagram of the present invention at a 6dB signal-to-noise ratio; Figure 5 This is a simulation diagram of the change of LR with epoch predicted by the present invention; Figure 6 A schematic diagram of a computer device provided by an embodiment of the present invention; Figure 7 The present invention is a block diagram of an electronic device provided according to an embodiment. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] In the description of the present invention, it should be understood that the terms “include” and “comprises” indicate the presence of described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0039] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0040] It should be further understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects are in an "or" relationship.

[0041] It should be understood that, although the terms first, second, third, etc. may be used to describe preset ranges, etc. in the embodiments of the present invention, these preset ranges should not be limited to these terms. These terms are only used to distinguish preset ranges from each other. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0042] The word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.

[0043] Various structural schematic diagrams of the embodiments disclosed in the present invention are shown in the accompanying drawings. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clear expression. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may additionally design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0044] The present invention provides a semantic communication method based on LSTM network automatic training, aiming to optimize the training aspects of the semantic communication codec so that it can ignore various differences, automatically train and optimize and meet the accuracy requirements, while improving its generalization ability for heterogeneous tasks, so as to ultimately meet the automatic training of multiple semantic communication devices.

[0045] Technical issues and comparison algorithms: The problem of training the semantic communication codec is formulated as the following non-convex optimization problem: (4) in, Represents the training loss function Loss in the context of shared knowledge; Indicates the knowledge background of participants in training; Indicates the parameters of the codec to be optimized, such as Specific parameters in Represents the sampling of data in each round of training.

[0046] Commonly used optimizers for semantic communication include stochastic gradient descent SGD and its variants, including Momentu, Adagrad, Adadelta, RMSprop, Adam, etc. In general, these algorithms are expressed as: (5) in, For updated codec parameters, express exist The gradient at Represents historical gradient information, Is the optimizer in the current iteration Hyperparameters such as learning rate.

[0047] To simplify the description, the present invention will be expressed in the form of vanilla SGD, namely: (6) The following are the learning rate update strategies commonly used in current semantic communication codec training: Fixed strategy: ; Multi-step descent strategy: , Given; Exponential Decline Strategy: ; SGDR Strategy: .

[0048] in, represents the initial learning rate, represents the learning rate at the tth iteration, Specify the learning rate setting range for the SGDR strategy. Indicates how many rounds have been executed. Indicated in After the round, SGDR restarted. They represent the decay factors of the multi-step decline strategy and the exponential decline strategy respectively.

[0049] The present invention discloses a semantic communication method based on LSTM network automatic training, introduces an LSTM neural network assisted design meta-learning method, parameterizes the learning rate strategy through the explicit mapping effect of LSTM and names it ASCNet (Auto Semantic Communication Net); in the subsequent description, ASCNet will be used to represent the meta-optimizer, LR will be used to represent the learning rate, and Loss will be used to represent the loss function value; the method comprises the following steps: First, the functional form of ASCNet is expressed as ; The process of learning LR using ASCNet is as follows: (8) in, , , The output of is the LR calculated in the tth iteration, are the parameters of ASCNet (such as weight matrix and bias), is the Loss of the sample at the tth iteration, Represent the output and state of the LSTM cell at the tth iteration, respectively. Represents the dimension of the state vector.

[0050] In each stochastic gradient descent iteration, A clear Loss-LR dependency can be learned, so that the network can , and historical training information stored in the network , adaptive prediction LR.

[0051] For each iteration step, the entire calculation process is expressed as follows:

[0052]

[0053]

[0054]

[0055] (9) in, , , They are the input gate, forget gate and output gate of LSTM respectively; , relu, tanh are three activation functions; the input of ASCNet is historical information and the current training loss ;Parameters of ASCNet , .

[0056] The input and output calculation process of ASCNet is understood as: S1, two inputs pass through a fully connected layer with a ReLU activation function Do the preprocessing, then import into LSTM and get the output ; S2. Utilization The fully connected layer with activation function right Perform linear transformation to obtain the predicted value ; Finally, multiply the predicted value by the scaling factor Ensure that the final predicted LR is located at interval.

[0057] ASCNet with LSTM as the backbone network can handle long-term information dependency problems well, so it can learn LR strategies that are adapted to the current knowledge background and the main task of semantic communication.

[0058] On the one hand, unlike directly learning the LR strategy itself, this technique uses parameters ASCNet is used to learn the LR strategy. This parameterized meta-learner helps to extract the underlying method, that is, how to design an appropriate LR strategy for the training problem of the semantic communication codec, rather than just a fixed mathematical formula. On the other hand, ASCNet is very lightweight. It only uses Loss to learn the LR strategy and is trained once after multiple rounds of training on the main task of semantic communication. It does not occupy too many training resources and is relatively easier to train and better controlled.

[0059] ASCNet step-by-step meta-learning algorithm and channel adaptation algorithm design: This study performs meta-training on ASCNet to improve its generalization performance. The optimization problem can be expressed as finding the semantic communication codec parameters that minimize the Loss on the meta-learning knowledge background Val. ,in It is predicted by ASCNet and is shown as follows: (10)

[0060] in, Represents the codec parameters updated at the next moment Depends on current codec parameters , historical information , current Loss And ASCNet parameters .

[0061] The present invention uses online approximation technology to jointly update ASCNet parameters and semantic communication codec parameters That is, the LR strategy is learned and updated synchronously during the model training process. There are two main problems in the training of the two: Q1. Since semantic communication knowledge background often has large data volume and complex codec structure, Performing gradual optimization will consume a lot of computing power and increase the latency of the semantic communication system; For Q1, the present invention will be updated in Several steps ( ) and then update In addition, the overall knowledge background of semantic communication is divided into the knowledge background of the main task of semantic communication and the knowledge background of meta-learning. and The parameters are trained separately on their own knowledge background to isolate each other and improve performance.

[0062] Q2. As shown in formula (2), the channel state of semantic communication Closely related to codec training, small differences in channel states often lead to large differences in codec parameters. Codecs trained in low signal-to-noise ratio intervals have low performance in high signal-to-noise ratio intervals, and vice versa. To solve this problem, traditional semantic communication systems often add a lot of network redundancy or randomly select signal-to-noise ratios for optimization in each round, which increases network load and randomness and is not suitable for the actual deployment of semantic communication devices.

[0063] For Q2, the present invention proposes a channel adaptive training method based on the two-stage training of the semantic communication main task and the meta-learning task. Specifically: 1. Assume that the semantic communication task is applied in a real scenario with a signal-to-noise ratio ranging from mdB to ndB, and n is greater than or equal to m.

[0064] 2. The channel signal-to-noise ratio of the main task of semantic communication is set to mdB to meet the robustness of the semantic communication system in the low signal-to-noise ratio area, and the main task knowledge background is used for training.

[0065] 3. In the meta-learning stage, the channel SNR is adjusted to ndB and the meta-learning knowledge background is adopted so that when the semantic communication main task adopts the learning rate predicted by ASCNet, the codec parameters are more biased towards the high SNR area, so that the codec has higher performance in the entire SNR range instead of tending to a certain SNR value.

[0066] 4. Since training in high signal-to-noise ratio areas is easier to converge, while training in low signal-to-noise ratio areas is affected by greater noise and is more difficult to converge, the knowledge background of the main task will be much larger than the meta-learning knowledge background, and the meta-learning stage will only be carried out after multiple rounds of training. This will make full use of training resources, reduce training costs and meet the essential concept of meta-learning.

[0067] ASCNet's step-by-step meta-learning method is expressed as: S101, initialize the encoder-decoder parameters of the semantic communication system, the weight matrix parameters of the meta-optimizer, the initial learning rate, the cell parameters of the meta-optimizer main LSTM, and set the number of meta-learning rounds T; S102, randomly dividing the training knowledge background of semantic communication into the main task knowledge background and the meta-learning knowledge background according to a certain ratio; S103, the input of the semantic communication task is randomly sampled from the main task knowledge background and data preprocessing is performed; S104, the pre-processed data is sequentially input into the semantic encoder, the channel encoder, the wireless channel (low signal-to-noise ratio threshold), the channel decoder and the semantic decoder as input to obtain the output of the main task of semantic communication; S105, after obtaining the training loss by using the input and output calculation, the gradient descent is performed according to the initial learning rate to optimize the codec parameters to be ; S106. After each T rounds of training, the training loss value of the current round is input into the meta-optimizer ASCNet. The meta-optimizer will use the meta-learning knowledge background (high signal-to-noise ratio threshold) to calculate and predict the learning rate that should be taken for the current semantic communication main task and record it. Finally, the parameters of ASCNet are updated as follows: ; S107, the learning rate of the main task will be updated to the learning rate predicted by the meta-optimizer, and steps S102 to S107 will be repeated until the training loss in step S105 no longer decreases or the total number of training rounds is reached.

[0068] It will be updated if the requirements are met, otherwise it will remain unchanged. The update will use the semantic codec parameters of the previous iteration And the current ASCNet parameters The loss of meta-learning knowledge background is minimized by formula (10). The specific process can be expressed as follows: (11) in, Represents the Adam optimizer, whose input is the Loss relative to the ASCNet parameters on the meta-learning knowledge background The gradient of LR for Adam optimizer; It is the predicted value of the semantic communication codec parameters at the next moment after calculating the learning rate estimated by the ASCNet that has not been updated yet. The calculation process is as follows: (12) Finally, the updated ASCNet parameters Will guide the semantic communication codec parameters The update process is as follows: (13) ASCNet can be gradually optimized during the learning process and dynamically adjust LR based on the training dynamics of the semantic codec. Moreover, the channel adaptation strategy enables the semantic communication codec to perform well in the entire signal-to-noise ratio range.

[0069] It will be appreciated by those skilled in the art that various aspects of the present invention may be implemented as systems, methods or program products. Therefore, various aspects of the present invention may be specifically implemented in the following forms, namely: complete hardware implementation, complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which may be collectively referred to herein as "circuits", "modules" or "platforms".

[0070] In another embodiment of the present invention, a semantic communication system based on automatic training of LSTM network is provided, which can be used to implement the above-mentioned semantic communication method based on automatic training of LSTM network. Specifically, the semantic communication system based on automatic training of LSTM network includes a construction module, a network module, a prediction module and an optimization module.

[0071] Among them, the construction module builds an LSTM network as the main architecture of the meta-optimizer. The input of the meta-optimizer is activated by the Relu function after passing through two different weight matrices; Network module, the input processed by the building module enters the forget gate, input gate, and output gate of the LSTM network respectively to obtain the output of the LSTM network; The prediction module processes the output through the weight matrix and the restriction factor to obtain the predicted learning rate value; Optimization module, the main task of semantic communication will use the learning rate predicted by the meta-optimizer to optimize the parameters of the semantic encoder and decoder to complete the semantic communication.

[0072] In another embodiment of the present invention, a terminal device is provided, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding functions; the processor described in the embodiment of the present invention can be used for the operation of the semantic communication method based on the automatic training of the LSTM network, including: An LSTM network is constructed as the main architecture of the meta-optimizer. The input of the meta-optimizer passes through two different weight matrices and is activated by the Relu function. The processed input enters the forget gate, input gate, and output gate of the LSTM network respectively to obtain the output of the LSTM network. The output is processed by the weight matrix and the restriction factor to obtain the predicted learning rate value. The main task of semantic communication will use the learning rate predicted by the meta-optimizer to optimize the parameters of the semantic codec to complete semantic communication.

[0073] In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It is understandable that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device, and can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that more specific examples (non-exhaustive list) of the computer-readable storage medium here include: an electrical connection with one or more wires, a portable 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 above.

[0074] Computer readable storage media also include data signals propagated in baseband or as part of a carrier wave, which carry readable program codes. Such propagated data signals can take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable storage medium can also be any readable medium other than a readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program code contained on the readable storage medium can be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0075] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0076] The processor may load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the semantic communication method based on LSTM network automatic training in the above embodiment; the processor may load and execute the following steps: An LSTM network is constructed as the main architecture of the meta-optimizer. The input of the meta-optimizer passes through two different weight matrices and is activated by the Relu function. The processed input enters the forget gate, input gate, and output gate of the LSTM network respectively to obtain the output of the LSTM network. The output is processed by the weight matrix and the restriction factor to obtain the predicted learning rate value. The main task of semantic communication will use the learning rate predicted by the meta-optimizer to optimize the parameters of the semantic codec to complete semantic communication.

[0077] See also Figure 6 , the terminal device is a computer device, and the computer device 60 of this embodiment includes: a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When the computer program 63 is executed by the processor 61, the semantic communication method based on the automatic training of the LSTM network in the embodiment is implemented. To avoid repetition, it is not described one by one here. Alternatively, when the computer program 63 is executed by the processor 61, the functions of each model / unit in the semantic communication system based on the automatic training of the LSTM network in the embodiment are implemented. To avoid repetition, it is not described one by one here.

[0078] The computer device 60 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will appreciate that Figure 6This is only an example of the computer device 60 and does not constitute a limitation of the computer device 60. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device may also include input and output devices, network access devices, buses, etc.

[0079] The processor 61 may be a central processing unit (CPU), or other general-purpose processors, central processing units, graphics processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, data processing logic based on quantum computing, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0080] The memory 62 may be an internal storage unit of the computer device 60, such as a hard disk or memory of the computer device 60. The memory 62 may also be an external storage device of the computer device 60, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc., equipped on the computer device 60.

[0081] Furthermore, the memory 62 may include both an internal storage unit of the computer device 60 and an external storage device. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 may also be used to temporarily store data that has been output or is to be output.

[0082] Any reference to a memory, database or other medium used in the embodiments provided in the present application may include at least one of a non-volatile and a volatile memory. Non-volatile memory may include a read-only memory (ROM), a magnetic tape, a floppy disk, a flash memory, an optical memory, a high-density embedded non-volatile memory, a resistive random access memory (ReRAM), a magnetic random access memory (MRAM), a ferroelectric random access memory (FRAM), a phase change memory (PCM), a graphene memory, etc. Volatile memory may include a random access memory (RAM) or an external cache memory, etc. As an illustration and not limitation, RAM may be in various forms, such as a static random access memory (SRAM) or a dynamic random access memory (DRAM), etc.

[0083] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0084] See also Figure 7 The terminal device 600 is an electronic device, and the electronic device is in the form of a general computing device. The components of the electronic device may include but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.

[0085] The storage unit stores program codes, which can be executed by the processing unit 610, so that the processing unit 610 performs the steps according to various exemplary embodiments of the present invention described in the above method section of this specification. For example, the processing unit 610 can perform the following steps: Figure 1 Follow the steps shown in .

[0086] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 6201 and / or a cache memory unit 6202 , and may further include a read-only memory unit (ROM) 6203 .

[0087] The storage unit 620 may also include a program / utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0088] Bus 630 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0089] The electronic device 600 may also communicate with one or more external devices 700 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 600, and / or communicate with any device that enables the electronic device 600 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 650. Furthermore, the electronic device 600 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 660. The network adapter 660 may communicate with other modules of the electronic device 600 via a bus 630. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms, etc.

[0090] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention described and shown in the drawings here can usually be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0091] Taking the main task of semantic communication as transmitting image semantics and performing image classification through image semantics at the receiving end as an example, the semantic communication signal-to-noise ratio interval is set to (0dB, 6dB), the knowledge background adopts the CIFAR-10 data set, the knowledge background division adopts a ratio of 10:1, the semantic communication codec adopts the Resnet architecture, the wireless channel adopts the AWGN channel, and the comparison method adopts the existing scheme described in the background technology of this invention.

[0092] Figure 2 The figure shows the trend of the loss function (Loss) changing with the number of training rounds (Epoch) when using different optimization algorithms during the training process. From the figure, we can see that when using the ASCNet meta-learning method for training, the training loss shows a gradual downward trend, which shows that ASCNet can effectively guide the update of model parameters, making the model gradually approach the optimal state.

[0093] In particular, when compared with the best performing multi-step stochastic gradient descent method (Mutistep SGD), ASCNet showed very similar performance. In the early stages of training, the training losses of both methods dropped rapidly, indicating that they were able to quickly capture the basic features of the data; as the training progressed, ASCNet continued to maintain a low training loss, and in the later stages of training, its downward trend was almost the same as that of Mutistep SGD, showing very stable performance.

[0094] The reason why the ASCNet meta-learning method can achieve training accuracy similar to that of Multistep SGD is mainly due to its advanced adaptive learning rate adjustment mechanism. By real-time monitoring of dynamic changes during training, ASCNet can intelligently adjust the learning rate of each parameter, thereby achieving fine-tuning of model details while ensuring training speed. In addition, ASCNet's meta-optimization framework also takes into account the improvement of the model's generalization ability. By dynamically adjusting the learning rate and model structure, the model can also show high accuracy when facing unknown data.

[0095] In conclusion, Figure 2 The excellent performance of the ASCNet meta-learning method in the training process is intuitively demonstrated, and its performance is comparable to the current best optimization algorithm. This discovery provides a new idea, that is, by introducing a meta-learning method to optimize the training process of the semantic communication codec, it is expected to achieve higher communication efficiency and more reliable communication services.

[0096] In today's rapidly developing digital communication field, semantic communication, as an emerging communication mode, is gradually becoming a hot topic of research. It aims to achieve more efficient and accurate information transmission by capturing and transmitting semantic information in signals rather than simple symbols or data. In order to further improve the performance of semantic communication systems, researchers have proposed a variety of codecs and optimization techniques. This paper will focus on a meta-optimizer called ASCNet and a new channel adaptation algorithm, and demonstrate their performance under different signal-to-noise ratio conditions through experimental data.

[0097] First, through Figure 3 and Figure 4 The performance comparison between the semantic communication codec trained with ASCNet and the existing comparison algorithm is intuitively shown under the conditions of 0dB and 6dB signal-to-noise ratio. From the figure, we can see that under both signal-to-noise ratio conditions, the codec trained with ASCNet shows better performance than the comparison algorithm. Especially in the low signal-to-noise ratio environment, ASCNet helps the codec better adapt to the harsh channel conditions, thereby reducing information loss and noise interference, and improving the accuracy of downstream classification.

[0098] Furthermore, an innovative channel adaptation method is proposed, which can achieve the optimal performance of the semantic communication codec in the entire signal-to-noise ratio range through a single training. The core idea of ​​this method is to dynamically adjust the parameters of the codec during the training process so that it can be automatically optimized according to different channel conditions, thereby maintaining high communication quality in various signal-to-noise ratio environments.

[0099] In summary, the ASCNet meta-optimizer and the channel adaptation algorithm work together to significantly improve the performance of the semantic communication codec. This improvement is especially evident in harsh channel environments. Through these innovative technologies, it can be foreseen that future semantic communication systems will provide more stable and reliable communication services while maintaining high efficiency.

[0100] Figure 5 The curve of the learning rate (LR) predicted by ASCNet as a function of the number of training iterations (Epoch) is shown. It can be clearly observed from the figure that the learning rate shows an obvious real-time dynamic downward trend throughout the training process. This phenomenon shows that ASCNet can intelligently adjust the learning rate according to the progress of training to achieve more effective learning results.

[0101] Specifically, when training just starts, ASCNet assigns a higher learning rate, which makes the model parameters update quickly and speeds up the model's learning of data features. As training progresses, the learning rate gradually decreases, which helps the model to make fine adjustments in the parameter space, thereby continuously approaching the optimal solution. In addition, this way of dynamically adjusting the learning rate helps avoid overfitting and ensures that the model has better generalization capabilities.

[0102] It is worth noting that the LR decline curve predicted by ASCNet is not monotonically decreasing, but shows a certain volatility. This fluctuation reflects the sensitivity of ASCNet to the complexity of training at different stages. It can moderately increase the learning rate when necessary to break through the training bottleneck, and then reduce the learning rate for detailed adjustment. This strategy makes ASCNet show higher flexibility and adaptability during the training process.

[0103] therefore, Figure 5 It clearly reveals how ASCNet optimizes the training process by intelligently predicting the learning rate. This method not only speeds up the convergence of the model, but also makes the final model have better performance and stability. By deeply understanding this mechanism of ASCNet, the design of semantic communication codec can be further optimized and its performance in practical applications can be pushed to a higher level.

[0104] In summary, the present invention provides a semantic communication method and system based on LSTM network automatic training. Taking the main task of image classification semantic communication as an example, the meta-optimizer proposed in this invention shows the best convergence performance in terms of training loss Loss compared with the traditional optimizer; in terms of classification accuracy of transmission, it has an improvement of 3% to 5% compared with the traditional optimizer in 0dB signal-to-noise ratio environment and 6dB signal-to-noise ratio environment.

[0105] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0106] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0107] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0108] In the embodiments provided by the present invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0109] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0110] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0111] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0112] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0113] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0115] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A semantic communication method based on LSTM network automatic training, characterized in that: The following steps are involved: S1. Build an LSTM network as the main architecture of the meta-optimizer. The input of the meta-optimizer is activated by the ReLU function after passing through two different weight matrices. S2, the input processed in step S1 enters the forget gate, input gate, and output gate of the LSTM network respectively to obtain the output of the LSTM network; S3, the output obtained in step S2 is processed by the weight matrix and the restriction factor to obtain the predicted learning rate value; S4, the main task of semantic communication, optimizes the parameters of the semantic codec using the learning rate predicted by the meta-optimizer in step S3 to complete the semantic communication.

2. The semantic communication method based on LSTM network automatic training according to claim 1, characterized in that: In step S1, the input of the meta-optimizer is the training loss value of the semantic communication task and the historical information of the previous stage. The training process of the semantic communication task is as follows: S101, initialize the encoder-decoder parameters of the semantic communication system, the weight matrix parameters of the meta-optimizer, the initial learning rate, the cell parameters of the meta-optimizer main LSTM, and set the number of meta-learning rounds T; S102, randomly dividing the training knowledge background of semantic communication into a main task knowledge background and a meta-learning knowledge background according to a set ratio; S103, the input of the semantic communication task is randomly sampled from the main task knowledge background and data preprocessing is performed; S104, taking the data pre-processed in step S103 as input, sequentially entering into the semantic encoder, channel encoder, wireless channel, channel decoder and semantic decoder, and obtaining the output of the main task of semantic communication; S105: After the training loss is calculated using the input and output, a gradient descent is performed according to the initial learning rate to optimize the codec parameters ; S106. After each T rounds of training, the training loss value of the current round is input into the meta-optimizer. The meta-optimizer uses the meta-learning knowledge background to perform the calculation of requirement 1, predicts and records the learning rate that should be taken for the current semantic communication main task, and updates the parameters of the meta-optimizer to ; S107, update the learning rate of the main task to the learning rate predicted by the meta-optimizer, and repeat steps S102-S107 until the training loss in step S105 no longer decreases or reaches the total number of training rounds.

3. The semantic communication method based on LSTM network automatic training according to claim 2 is characterized in that: Update meta-optimizer parameters for: in, Represents the Adam optimizer; LR for Adam optimizer; It is the predicted value of the semantic communication codec parameters at the next moment after calculation using the learning rate estimated by the meta-optimizer that has not yet been updated.

4. The semantic communication method based on LSTM network automatic training according to claim 3 is characterized in that: The predicted values ​​of the semantic communication codec parameters at the next moment calculated using the learning rate estimated by the meta-optimizer that has not yet been updated for: in, is the input-output relationship of the meta-optimizer, is the Loss value of the input meta-optimizer, is the cell state of the meta-optimizer (including historical training information), is the parameter information of the meta-optimizer, is the gradient descent information of the Loss value about the parameter w on the semantic communication main task network.

5. The semantic communication method based on LSTM network automatic training according to claim 2 is characterized in that: Codec parameters for: in, is the input-output relationship of the meta-optimizer, is the Loss value of the input meta-optimizer, is the cell state of the meta-optimizer (including historical training information), is the parameter information of the meta-optimizer at the next moment, is the gradient descent information of the Loss value about the parameter w on the semantic communication main task network.

6. The semantic communication method based on LSTM network automatic training according to claim 1, characterized in that: In step S3, the predicted learning rate value is specifically: The two inputs pass through a fully connected layer with a ReLU activation function Do the preprocessing, then import into LSTM and get the output ;use The fully connected layer with activation function right Perform linear transformation to obtain the predicted value ; Finally, multiply the predicted value by the scaling factor Ensure that the final predicted learning rate value is within interval.

7. The semantic communication method based on LSTM network automatic training according to claim 1, characterized in that: In step S4, the meta-optimizer parameters are jointly updated using online approximation techniques and semantic communication codec parameters , will be updated in Several steps ( ) after update once; The overall knowledge background of semantic communication is divided into the knowledge background of the main task of semantic communication and the knowledge background of meta-learning. and The training of parameters is performed on their respective knowledge backgrounds.

8. The semantic communication method based on LSTM network automatic training according to claim 1, characterized in that: In step S4, in the semantic communication main task and meta-learning task stage, the channel adaptive training method is as follows: Assume that the semantic communication task acts on a real scene with a signal-to-noise ratio range of mdB to ndB, and n is greater than or equal to m; set the channel signal-to-noise ratio of the main task of semantic communication to mdB, and adopt the main task knowledge background for training; In the meta-learning stage, the channel signal-to-noise ratio is adjusted to ndB and the meta-learning knowledge background is adopted; the main task knowledge background will be much larger than the meta-learning knowledge background, and the meta-learning stage is only performed after multiple rounds of training.

9. The semantic communication method based on LSTM network automatic training according to claim 7 or 8, characterized in that: Parameters of semantic codec for: in, is the updated codec parameter for the next moment. is the Loss value in the context of verification knowledge, is the input-output relationship of the meta-optimizer, is the input-output relationship of the gradient descent function, is the total number of training rounds.

10. A semantic communication system based on LSTM network automatic training, characterized in that: include: Build the module and construct the LSTM network as the main architecture of the meta-optimizer. The input of the meta-optimizer is activated by the ReLU function after passing through two different weight matrices. Network module, the input processed by the building module enters the forget gate, input gate, and output gate of the LSTM network respectively to obtain the output of the LSTM network; The prediction module processes the output through the weight matrix and the restriction factor to obtain the predicted learning rate value; Optimization module, the main task of semantic communication will use the learning rate predicted by the meta-optimizer to optimize the parameters of the semantic encoder and decoder to complete the semantic communication.

Images