Data transmission method, system, device and computer program product

Abstract

The invention provides a data transmission method, system and device and a computer program product. The method comprises the following steps: a sending end device encodes target data based on a first neural network, and sends first encoding result data obtained by encoding to a receiving end device; and the receiving end device decodes the received moral first coding result data based on the second neural network, and determines target data based on first decoding result data obtained by decoding. Wherein the first neural network and the second neural network are obtained by training based on first loss between the sample data and the second decoding result data, the second decoding result data are obtained by decoding the second encoding result data by the second neural network, and the second encoding result data are obtained by encoding the sample data by the first neural network. Based on the characteristics of the neural network, the computing resources of the sending / receiving end are fully utilized to complete data compression transmission, and the neural network supports fine tuning training and can follow up the demand for dynamic upgrading.

Description

Technical Field

[0001] The present application relates to the field of data transmission, and in particular to a data transmission method, system, device and computer program product. Background Art

[0002] In data transmission scenarios, compressing data before sending and decompressing it is an effective way to reduce bandwidth resource overhead. Currently, this type of transmission technology is implemented by professional software (such as h265, vpp, etc. for video data transmission), which is usually not frequently upgraded and cannot keep up with changes in usage needs. In addition, as the configuration of hardware devices (sender devices, receiver devices) becomes higher and higher, this type of software cannot fully utilize the NPU and GPU computing resources brought by high configuration. Summary of the invention

[0003] The purpose of this application is to provide a data transmission method, system, device and computer program product, which can fully utilize the computing resources of the sending / receiving end to complete data compression transmission based on a neural network, and the neural network supports fine-tuning training and can be dynamically upgraded to follow up on demand.

[0004] In order to achieve the above purpose, the embodiment of the present application is implemented as follows: In a first aspect, a data transmission method is provided, comprising: The transmitting end device encodes the target data based on the first neural network, and sends the first encoded result data obtained by encoding to the receiving end device, wherein the data volume of the first encoded result data is smaller than the data volume of the target data; After receiving the first encoding result data, the receiving end device decodes the first encoding result data based on the second neural network, and determines the target data based on the first decoding result data obtained by decoding; The first neural network and the second neural network are trained based on a first loss between sample data and second decoding result data, the second decoding result data is obtained by decoding the second encoding result data by the second neural network, and the second encoding result data is obtained by encoding the sample data by the first neural network.

[0005] In a second aspect, a data transmission system is provided, including a transmitting device and a receiving device. The transmitting end device encodes the target data based on the first neural network, and sends the first encoding result data obtained by encoding to the receiving end device, wherein the data volume of the first encoding result data is smaller than the data volume of the target data; After receiving the first encoding result data, the receiving end device decodes the first encoding result data based on the second neural network, and determines the target data based on the first decoding result data obtained by decoding; The first neural network and the second neural network are trained based on a first loss between sample data and second decoding result data, the second decoding result data is obtained by decoding the second encoding result data by the second neural network, and the second encoding result data is obtained by encoding the sample data by the first neural network.

[0006] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor; and a memory configured to store computer-executable instructions, wherein the computer-executable instructions, when executed, cause the processor to execute the method described in the first aspect.

[0007] According to a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium is used to store computer-executable instructions, and the computer-executable instructions implement the method described in the first aspect when executed by a processor.

[0008] The embodiment of the present application constructs a first neural network for encoding the transmission data, and a second neural network for decoding, and simulates the transmission scenario based on sample data to train the first neural network and the second neural network. After the training is completed, the first neural network is deployed on the sending end device, and the second neural network is deployed on the receiving end device, so that the sending end device encodes the target data based on the first neural network, and sends the first encoding result data obtained by encoding to the receiving end device; correspondingly, after receiving the first encoding result data, the receiving end device decodes the first encoding result data based on the second neural network, and determines the target data based on the first decoding result data obtained by decoding. Since the embodiment of the present application adopts artificial intelligence technology to realize data compression transmission, based on the characteristics of the neural network, it can make full use of the hardware computing resources of the sending end device and the receiving end device to quickly complete the encoding and decoding process. At the same time, the neural network also supports fine-tuning training, which is more suitable for upgrading according to changes in usage requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0010] Figure 1A flowchart of a data transmission method according to an embodiment of the present application.

[0011] Figure 2 This is a schematic diagram of the first neural network and the second neural network in an embodiment of the present application when transmitting data.

[0012] Figure 3 Schematic diagram of the structure of the latent space module in the first neural network of the embodiment of the present application.

[0013] Figure 4 Schematic diagram of the structure of the latent space module in the second neural network of the embodiment of the present application.

[0014] Figure 5 Schematic diagram of training the first neural network and the second neural network based on this embodiment.

[0015] Figure 6 Schematic diagram of the structure of the data transmission device according to an embodiment of the present application.

[0016] Figure 7 A schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0018] An embodiment of the present application provides a data transmission method. Figure 1 : is a flow chart of the data transmission method, comprising the following steps: S102, the transmitting end device encodes the target data based on the first neural network, and sends the first encoding result data obtained by encoding to the receiving end device, and the data volume of the first encoding result data is smaller than the data volume of the target data; S104, after receiving the first encoding result data, the receiving end device decodes the first encoding result data based on the second neural network, and determines the target data based on the decoded first decoding result data.

[0019] In this embodiment, the first neural network and the second neural network are trained based on the first loss between the sample data and the second decoding result data, the second decoding result data is obtained by decoding the second encoding result data by the second neural network, and the second encoding result data is obtained by encoding the sample data by the first neural network.

[0020] The general training principle of the first neural network and the second neural network is: the sample data is input into the first neural network and then encoded, and the second encoded result data obtained by the encoding is decoded by the second neural network to obtain the second decoded result data. Among them, the second decoded result data should be as close to the sample data as possible. For this purpose, the network parameters of the first neural network and the second neural network can be adjusted for the purpose of reducing the first loss. In this way, the network parameters are continuously iterated until the second decoded result data can be used as sample data.

[0021] Correspondingly, after the training is completed, when applied to the transmission of target data, the first decoding result data finally decoded can be regarded as the target data.

[0022] It should be noted that the target data of this embodiment may be image data, text data, audio data, etc., and this document does not specifically limit the data type of the target data.

[0023] In summary, the method of this embodiment constructs a first neural network for encoding transmission data and a second neural network for decoding, and simulates the transmission scenario based on sample data to train the first neural network and the second neural network. After the training is completed, the first neural network is deployed on the sending end device, and the second neural network is deployed on the receiving end device, so that the sending end device encodes the target data based on the first neural network, and sends the first encoding result data obtained by encoding to the receiving end device; correspondingly, after receiving the first encoding result data, the receiving end device decodes the first encoding result data based on the second neural network, and determines the target data based on the first decoding result data obtained by decoding. Since the method of this embodiment uses artificial intelligence technology to realize data compression transmission, based on the characteristics of the neural network, it can make full use of the hardware computing resources of the sending end device and the receiving end device to quickly complete the encoding and decoding process. At the same time, the neural network also supports fine-tuning training, which is more suitable for upgrading according to changes in usage requirements.

[0024] Here we assume that the target data is video YUV data. For further reference Figure 2 , Figure 2 The network structures of the first neural network and the second neural network are exemplified by target data transmission. The first neural network includes a multi-level Bottom-up (from top to bottom) series of downsampling units A and a multi-level Top-down (from bottom to top) series of first upsampling units B. Each downsampling unit includes a downsampling module and a residual module, each first upsampling unit includes a latent space module and an upsampling module, the downsampling unit corresponds to the level of the first upsampling unit one by one, the output data of the downsampling unit is used as the input data of the first upsampling unit corresponding to the same level, and the first encoding result data is a collection of the output data (bit stream) of the latent space module in each first upsampling unit.

[0025] The second neural network includes a second upsampling unit B' connected in series with multiple levels of Top-down. Each first upsampling unit includes a latent space module and an upsampling module. The second upsampling unit corresponds to the level of the first upsampling unit one by one. The output data (bit stream) of the latent space module in each first upsampling unit is used as the input data of the latent space module in the corresponding second upsampling unit at the same level. It should be noted that Figure 2 The constant matrix in is a parameter agreed upon by the transmitting device and the receiving device, and is not specifically limited in this article.

[0026] Figure 3 : is a schematic diagram of the structure of the latent space module of the first upsampling unit, which mainly includes: a first convolution layer, a second convolution layer, a first probability mass function layer, an entropy coding layer, a residual rounding layer and a third convolution layer. Among them, the input end of the first convolution layer is connected to the first input data (reference Figure 1 f1 in ), the input of the second convolutional layer is connected to the second input data (refer to Figure 1 f2), the output ends of the first convolutional layer and the second convolutional layer are connected to the input ends of the first probability mass function layer and the residual rounding layer, the output end of the first probability mass function layer is connected to the input end of the entropy coding layer, and the output ends of the entropy coding layer and the residual rounding layer are connected to the input end of the third convolutional layer; the first input data comes from the corresponding downsampling unit, and the second input data comes from other first upsampling units at the previous level.

[0027] Figure 3 In the example, f2 is calculated by the residual module and convolved with f1 by the first convolutional layer to obtain At the same time, f2 is calculated by the residual module and then by the convolution of the second convolutional layer to obtain . and After rounding the residual, we get . After the convolution calculation of the third convolutional layer, the result of f2 after the residual module calculation is added to the feature map, and then output after the residual module calculation. In addition, The probability mass after the first probability mass function layer After calculation and After entropy coding, the bit data is output.

[0028] The formula for the above residual rounding calculation is .

[0029] The probability mass calculation formula of the first probability mass function layer is: , is an integer.

[0030] The formula for the above entropy coding is .

[0031] Figure 4 It is a structural diagram of the latent space module of the second upsampling unit, and the latent space module of the second upsampling unit includes: a fourth convolutional layer, a second probability mass function layer, an entropy decoding layer and a fifth convolutional layer. Among them, the input end of the fourth convolutional layer is connected to the third input data f3, the output end of the fourth convolutional layer is connected to the input end of the second probability mass function layer, the output end of the second probability mass function layer and the fourth input data f4 are connected to the input end of the entropy decoding layer, and the output end of the entropy decoding layer is connected to the input end of the fifth convolutional layer; the third input data comes from the first upsampling unit corresponding to the same level, and the fourth input data comes from other second upsampling units of the previous level.

[0032] Figure 4 In the example, f4 is calculated by the residual module, the convolution calculation of the fourth convolutional layer, and the probability mass calculation of the second probability mass function layer, and then entropy decoding is performed on the received bit data f3 to obtain , After the convolution calculation of the fifth convolutional layer, the feature map is added to the feature map of f4 calculated by the residual module, and then the output feature map is obtained after calculation by the residual module.

[0033] The probability mass calculation formula of the second probability mass function layer may refer to the first probability mass function layer.

[0034] The formula for the above entropy decoding calculation is: , Represents input.

[0035] In addition, in this embodiment, the first neural network and the second neural network are also trained based on a second loss. The second loss is determined based on a relative entropy between the first input data and the second input data.

[0036] in, Figure 3 The training method of the latent space module shown can be referred to Figure 5As shown in the figure, f2 is calculated by the residual module, concatenated with f1, and then calculated by the convolution layer. , and then use Initialize the posterior uniform distribution In addition, the feature map f2 is obtained after the residual module and convolution layer calculation , and then use Initialize the prior distribution . In the calculation, we get and After that, the KL divergence (relative entropy) is calculated, which is and The loss between them, the training process is to minimize the KL divergence and jointly adjust the network parameters of the first neural network and the second neural network.

[0037] Above The distribution of ; Above The distribution of .

[0038] The calculation formula of the above KL divergence can adopt the standard relative entropy formula.

[0039] In addition, corresponding to Figure 1 In addition to the method shown, another embodiment of this embodiment also provides a data transmission system. Figure 6 It is a schematic diagram of the structure of the data transmission system 600 , including: a sending end device 610 and a receiving end device 620 .

[0040] The transmitting end device 610 encodes the target data based on the first neural network, and sends the first encoding result data obtained by encoding to the receiving end device 620, wherein the data volume of the first encoding result data is smaller than the data volume of the target data; After receiving the first encoding result data, the receiving end device 620 decodes the first encoding result data based on the second neural network, and determines the target data based on the first decoding result data obtained by decoding; The first neural network and the second neural network are trained based on a first loss between sample data and second decoding result data, the second decoding result data is obtained by decoding the second encoding result data by the second neural network, and the second encoding result data is obtained by encoding the sample data by the first neural network.

[0041] The device of this embodiment constructs a first neural network for encoding the transmission data and a second neural network for decoding, and simulates the transmission scenario based on sample data to train the first neural network and the second neural network. After the training is completed, the first neural network is deployed on the sending end device, and the second neural network is deployed on the receiving end device, so that the sending end device encodes the target data based on the first neural network, and sends the first encoding result data obtained by encoding to the receiving end device; correspondingly, after receiving the first encoding result data, the receiving end device decodes the first encoding result data based on the second neural network, and determines the target data based on the first decoding result data obtained by decoding. Since the device of this embodiment adopts artificial intelligence technology to realize data compression transmission, based on the characteristics of the neural network, it can make full use of the hardware computing resources of the sending end device and the receiving end device to quickly complete the encoding and decoding process. At the same time, the neural network also supports fine-tuning training, which is more suitable for upgrading according to changes in usage requirements.

[0042] Optionally, the first neural network includes multiple levels of Bottom-up downsampling units connected in series and multiple levels of Top-down first upsampling units connected in series, each of the downsampling units includes a downsampling module and a residual module, each of the first upsampling units includes a latent space module and an upsampling module, the downsampling units correspond one-to-one to the levels of the first upsampling units, the output data of the downsampling units serves as the input data of the first upsampling units corresponding to the same level, and the first encoding result data is a collection of output data of the latent space module in each of the first upsampling units.

[0043] Optionally, the second neural network includes a second upsampling unit connected in series with multiple levels of Top-down, each of the first upsampling units includes a latent space module and an upsampling module, the second upsampling units correspond one-to-one to the levels of the first upsampling units, and the output data of the latent space module in each of the first upsampling units serves as the input data of the latent space module in the second upsampling unit corresponding to the same level.

[0044] Optionally, the latent space module of the first upsampling unit includes: a first convolutional layer, a second convolutional layer, a first probability mass function layer, an entropy coding layer, a residual rounding layer and a third convolutional layer; wherein the input end of the first convolutional layer is connected to the first input data, the input end of the second convolutional layer is connected to the second input data, the output ends of the first convolutional layer and the second convolutional layer are connected to the input ends of the first probability mass function layer and the residual rounding layer, the output end of the first probability mass function layer is connected to the input end of the entropy coding layer, and the output ends of the entropy coding layer and the residual rounding layer are connected to the input end of the third convolutional layer; the first input data comes from the corresponding downsampling unit, and the second input data comes from other first upsampling units at the previous level.

[0045] Optionally, the first neural network and the second neural network are also trained based on a second loss, and the second loss is determined based on a relative entropy between the first input data and the second input data.

[0046] Optionally, the latent space module of the second upsampling unit includes: a fourth convolutional layer, a second probability mass function layer, an entropy decoding layer and a fifth convolutional layer; wherein, the input end of the fourth convolutional layer is connected to the third input data, the output end of the fourth convolutional layer is connected to the input end of the second probability mass function layer, the output end of the second probability mass function layer and the fourth input data are connected to the input end of the entropy decoding layer, and the output end of the entropy decoding layer is connected to the input end of the fifth convolutional layer; the third input data comes from the first upsampling unit corresponding to the same level, and the fourth input data comes from other second upsampling units of the previous level.

[0047] Optionally, the target data is video data encoded by a YUV algorithm.

[0048] The data transmission system of this embodiment can be used as Figure 1 The execution subject of the method shown can thus realize Figure 1 The steps and functions in the method shown.

[0049] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 7 At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. Of course, the electronic device may also include hardware required for other services.

[0050] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0051] The memory is used to store the computer program. Specifically, the computer program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provides the computer program to the processor. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming the above-mentioned logic layer. Figure 6 The data transmission system shown. Correspondingly, the processor executes the program stored in the memory and is specifically used to perform the following operations: The transmitting end device encodes the target data based on the first neural network, and sends the first encoding result data obtained by encoding to the receiving end device, wherein the data volume of the first encoding result data is smaller than the data volume of the target data; After receiving the first encoding result data, the receiving end device decodes the first encoding result data based on the second neural network, and determines the target data based on the first decoding result data obtained by decoding; The first neural network and the second neural network are trained based on a first loss between sample data and second decoding result data, the second decoding result data is obtained by decoding the second encoding result data by the second neural network, and the second encoding result data is obtained by encoding the sample data by the first neural network.

[0052] The above is as in this manual Figure 1The data transmission method disclosed in the illustrated embodiment can be applied to a processor and implemented by the processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in a decoding processor. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0053] Of course, in addition to software implementation, the electronic device of this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0054] In addition, the present application also provides a computer-readable storage medium, which stores one or more computer programs, wherein the one or more computer programs include instructions. When the instructions are executed by a portable electronic device including multiple application programs, the portable electronic device can execute Figure 1 The steps in the method shown include: Based on a first objective function coupled to a vehicle kinematics model, a vehicle state parameter of a target vehicle is optimized; wherein the vehicle kinematics model includes a mathematical relationship between vehicle driving control parameters and vehicle state parameters.

[0055] Based on a second objective function coupled to the vehicle kinematic model, the vehicle driving control parameters of the target vehicle are optimized with the optimized vehicle state parameters of the target vehicle as constraints.

[0056] Based on the optimized vehicle driving control parameters of the target vehicle, driving control is performed on the target vehicle.

[0057] It should be understood by those skilled in the art that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0058] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0059] The above are only embodiments of this specification and are not intended to limit this specification. For those skilled in the art, this specification may be subject to various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this specification shall be included in the scope of the claims of this specification. In addition, all other embodiments obtained by ordinary technicians in this field without creative work shall fall within the scope of protection of this document.

Claims

1. A data transmission method, characterized in that: include: The transmitting end device encodes the target data based on the first neural network, and sends the first encoding result data obtained by encoding to the receiving end device, wherein the data volume of the first encoding result data is smaller than the data volume of the target data; After receiving the first encoding result data, the receiving end device decodes the first encoding result data based on the second neural network, and determines the target data based on the first decoding result data obtained by decoding; The first neural network and the second neural network are trained based on a first loss between sample data and second decoding result data, the second decoding result data is obtained by decoding the second encoding result data by the second neural network, and the second encoding result data is obtained by encoding the sample data by the first neural network.

2. The method according to claim 1, characterized in that The first neural network includes a multi-level Bottom-up downsampling unit connected in series and a multi-level Top-down first upsampling unit connected in series, each of the downsampling units includes a downsampling module and a residual module, each of the first upsampling units includes a latent space module and an upsampling module, the downsampling units correspond to the levels of the first upsampling units one by one, the output data of the downsampling units serves as the input data of the first upsampling units corresponding to the same level, and the first encoding result data is a collection of the output data of the latent space module in each of the first upsampling units.

3. The method according to claim 2, characterized in that The second neural network includes a second upsampling unit connected in series with multiple levels of Top-down, each of the first upsampling units includes a latent space module and an upsampling module, the second upsampling units correspond one-to-one to the levels of the first upsampling units, and the output data of the latent space module in each of the first upsampling units serves as the input data of the latent space module in the second upsampling unit corresponding to the same level.

4. The method according to claim 2, characterized in that: The latent space module of the first upsampling unit includes: A first convolutional layer, a second convolutional layer, a first probability mass function layer, an entropy coding layer, a residual rounding layer, and a third convolutional layer; Among them, the input end of the first convolution layer is connected to the first input data, the input end of the second convolution layer is connected to the second input data, the output ends of the first convolution layer and the second convolution layer are connected to the input ends of the first probability mass function layer and the residual rounding layer, the output end of the first probability mass function layer is connected to the input end of the entropy coding layer, and the output ends of the entropy coding layer and the residual rounding layer are connected to the input end of the third convolution layer; the first input data comes from the corresponding down-sampling unit, and the second input data comes from other first up-sampling units at the previous level.

5. The method according to claim 4, characterized in that The first neural network and the second neural network are also trained based on a second loss, and the second loss is determined based on a relative entropy between the first input data and the second input data.

6. The method according to claim 3, characterized in that The latent space module of the second upsampling unit includes: a fourth convolutional layer, a second probability mass function layer, an entropy decoding layer and a fifth convolutional layer; Among them, the input end of the fourth convolutional layer is connected to the third input data, the output end of the fourth convolutional layer is connected to the input end of the second probability mass function layer, the output end of the second probability mass function layer and the fourth input data are connected to the input end of the entropy decoding layer, and the output end of the entropy decoding layer is connected to the input end of the fifth convolutional layer; the third input data comes from the first upsampling unit corresponding to the same level, and the fourth input data comes from other second upsampling units of the previous level.

7. The method according to any one of claims 1 to 6, characterized in that The target data is video data encoded by a YUV algorithm.

8. A data transmission system, comprising a transmitting device and a receiving device, characterized in that: The transmitting end device encodes the target data based on the first neural network, and sends the first encoding result data obtained by encoding to the receiving end device, wherein the data volume of the first encoding result data is smaller than the data volume of the target data; After receiving the first encoding result data, the receiving end device decodes the first encoding result data based on the second neural network, and determines the target data based on the first decoding result data obtained by decoding; The first neural network and the second neural network are trained based on a first loss between sample data and second decoding result data, the second decoding result data is obtained by decoding the second encoding result data by the second neural network, and the second encoding result data is obtained by encoding the sample data by the first neural network.

9. An electronic device, comprising: processor; and a memory arranged to store computer executable instructions, wherein the executable instructions, when executed, cause the processor to perform the method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer-readable storage medium storing a computer program, characterized in that: The computer program is operable to cause a computer to perform the method according to any one of claims 1 to 7.

Images