Optical communication nonlinear compensation method, system, medium, and apparatus

Abstract

The application provides an optical communication nonlinear compensation method, system, medium and equipment, comprising: establishing an exponential decay model, obtaining an initial parameter range of the exponential decay model according to a communication channel condition; setting different exponential decay model parameters on the basis of the initial parameter range of the exponential decay model, obtaining learnable DBP models of different structures and corresponding model complexities; initializing parameters of the learnable DBP models of different structures; training the learnable DBP models after the parameter initialization, obtaining corresponding Q factors; comparing the Q factors and the model complexities of different learnable DBP models, and selecting a learnable DBP model with the lowest complexity and the highest Q factor as an optimal result; and using the trained optimal learnable DBP model for optical fiber communication nonlinear compensation. The application adopts an exponential decay and learnable DBP fusion model, and solves the problems of high algorithm complexity and poor precision of an optical fiber communication nonlinear algorithm under long-distance and large baud rate.

Description

Technical Field

[0001] This invention relates to the field of optical communication nonlinearity compensation technology, specifically to an optical communication nonlinearity compensation method, system, medium, and device. More particularly, it relates to an optical communication nonlinearity compensation method based on a fusion model of exponential decay and learnable DBP. Background Technology

[0002] With the increasing popularity of social media, video applications, online games, cloud computing, and other applications, the demand for communication traffic is constantly growing. Statistics show that global communication traffic has increased several hundredfold since 2010. It is projected that the growth rate of communication traffic will continue to grow exponentially in the coming years. Coherent optical communication, with its advantages of higher transmission bandwidth efficiency and longer transmission distance, has been used to solve the aforementioned problems. However, in current optical communication systems, compensating for fiber nonlinear damage remains a difficult problem due to the interaction between the nonlinear Kerr effect, dispersion effect, and amplified spontaneous emission noise.

[0003] Chinese patent application CN2021110835603 discloses a fiber optic nonlinear equalization method based on an improved adaptive backpropagation (MA-DBP) algorithm. The method includes: constructing a long-distance optical transmission system using PDM-16QAM; acquiring experimental data from the system's receiver; selecting an improved nonlinear step size for the step size of the digital backpropagation algorithm in digital signal processing; calculating the cost function of Godard's error based on the improved cascaded multimode algorithm; using the Fibonacci search algorithm to find the optimal fiber nonlinear coefficient corresponding to the cost function; and performing fiber nonlinear compensation using the calculated optimal fiber nonlinear coefficient.

[0004] Patent No. 2021100813789 discloses a method for joint equalization of optical fiber signal impairments based on the MC-DBP algorithm. By using all received signals as input to a single CPU on a multi-core DSP processor, each CPU performs joint equalization for the dispersion and nonlinear effects experienced by one of the signals. The CPUs do not need to exchange data when calculating the XPM effect, achieving completely independent parallel computation. Furthermore, the traditional single step size of the MC-DBP algorithm is replaced with a self-phase modulation step size and a cross-phase modulation step size, using a larger step size when compensating for the XPM effect, thus reducing the computational load of the MC-DBP algorithm.

[0005] However, the above algorithm is too complex to be suitable for practical applications in scenarios with long fiber optic transmission distances and high signal baud rates, and the compensation effect cannot be guaranteed in different scenarios. Summary of the Invention

[0006] In view of the deficiencies in the prior art, the purpose of this invention is to provide a method, system, medium and device for optical communication nonlinear compensation.

[0007] The optical communication nonlinearity compensation method provided by the present invention includes:

[0008] Step S1: Establish an exponential decay model and obtain the initial parameter range of the exponential decay model based on the communication channel conditions;

[0009] Step S2: Based on the initial parameter range of the exponential decay model, set different parameters for the exponential decay model to obtain learnable DBP models with different structures and their corresponding model complexities;

[0010] Step S3: Initialize the parameters of learnable DBP models with different structures;

[0011] Step S4: Train the learnable DBP model after parameter initialization to obtain the corresponding Q factor;

[0012] Step S5: Compare the Q-factor and model complexity of different learnable DBP models, and select the learnable DBP model with the lowest complexity and the highest Q-factor as the optimal result;

[0013] Step S6: Use the trained optimal learnable DBP model to perform nonlinear compensation for optical fiber communication.

[0014] Preferably, step S1 includes:

[0015] The expression for the exponential decay model is:

[0016]

[0017] in, , For model hyperparameters; For learnable DBP Number of taps in the dispersion layer;

[0018] By substituting different Calculate the number of dispersive layer taps under different layers. ;

[0019] Based on the communication channel conditions, the initial parameter range of the exponential decay model is obtained by considering the influence of fiber dispersion on the symbol length. The expression is as follows:

[0020]

[0021] in, For signal bandwidth; The center frequency of the signal; The duration of the signal; This refers to the length of the optical fiber. These are the second-order dispersion coefficients of the optical fiber; The range of values ​​is ,when When, it indicates that the learnable DBP model has not been simplified. The larger the value, the higher the simplification of the learnable DBP model; c represents the speed of light propagation.

[0022] Preferably, step S2 includes:

[0023] The learnable DBP model can be established using the following expression:

[0024] ,

[0025] in, P The fiber attenuation coefficient; These are the second-order dispersion coefficients of the optical fiber; It is a complex number; The frequency of the i-th signal is represented by t; t represents the signal duration at time t.

[0026] No. The linear layer tap representation is as follows:

[0027]

[0028] in, For fiber nonlinearity coefficients, The signal amplitude;

[0029] The number of dispersive layer taps at different levels of the learnable DBP model was calculated using the exponential decay model. Thus, the learnable DBP model structure is determined;

[0030] The expression for determining the complexity of a learnable DBP model is:

[0031]

[0032] in, is the number of fiber optic segments; RM represents the total complexity.

[0033] Preferably, step S4 includes:

[0034] The training process for the learnable DBP model parameters is as follows: The original signal from the transmitting end and the signal from the receiving end, after passing through an optical communication nonlinear compensation algorithm based on an exponential decay and learnable DBP fusion model, are collected, and the bit error rate is calculated. Then through the bit error rate Calculate the loss function Then, backpropagation is performed to update the parameters of the dispersive and nonlinear layers, expressed as:

[0035]

[0036] in, This represents the complementary error function.

[0037] The optical communication nonlinearity compensation system provided by the present invention includes:

[0038] Module M1: Establish the exponential decay model and obtain the initial parameter range of the exponential decay model based on the communication channel conditions;

[0039] Module M2: Based on the initial parameter range of the exponential decay model, set different parameters for the exponential decay model to obtain learnable DBP models with different structures and their corresponding model complexities;

[0040] Module M3: Initializes parameters for learnable DBP models with different structures;

[0041] Module M4: Trains the learnable DBP model after parameter initialization to obtain the corresponding Q factor;

[0042] Module M5: Compare the Q-factor and model complexity of different learnable DBP models, and select the learnable DBP model with the lowest complexity and the highest Q-factor as the optimal result;

[0043] Module M6: Uses the trained optimal learnable DBP model for nonlinear compensation in fiber optic communication.

[0044] Preferably, the module M1 includes:

[0045] The expression for the exponential decay model is:

[0046]

[0047] in, , For model hyperparameters; For learnable DBP Number of taps in the dispersion layer;

[0048] By substituting different Calculate the number of dispersive layer taps under different layers. ;

[0049] Based on the communication channel conditions, the initial parameter range of the exponential decay model is obtained by considering the influence of fiber dispersion on the symbol length. The expression is as follows:

[0050]

[0051] in, For signal bandwidth; The center frequency of the signal; The duration of the signal; This refers to the length of the optical fiber. These are the second-order dispersion coefficients of the optical fiber; The range of values ​​is ,when When, it indicates that the learnable DBP model has not been simplified. The larger the value, the higher the simplification of the learnable DBP model; c represents the speed of light propagation.

[0052] Preferably, the module M2 includes:

[0053] The learnable DBP model can be established using the following expression:

[0054] ,

[0055] in, P The fiber attenuation coefficient; These are the second-order dispersion coefficients of the optical fiber; It is a complex number; The frequency of the i-th signal is represented by t; t represents the signal duration at time t.

[0056] No. The linear layer tap representation is as follows:

[0057]

[0058] in, For fiber nonlinearity coefficients, The signal amplitude;

[0059] The number of dispersive layer taps at different levels of the learnable DBP model was calculated using the exponential decay model. Thus, the learnable DBP model structure is determined;

[0060] The expression for determining the complexity of a learnable DBP model is:

[0061]

[0062] in, is the number of fiber optic segments; RM represents the total complexity.

[0063] Preferably, the module M4 includes:

[0064] The training process for the learnable DBP model parameters is as follows: The original signal from the transmitting end and the signal from the receiving end, after passing through an optical communication nonlinear compensation algorithm based on an exponential decay and learnable DBP fusion model, are collected, and the bit error rate is calculated. Then through the bit error rate Calculate the loss function Then, backpropagation is performed to update the parameters of the dispersive and nonlinear layers, expressed as:

[0065]

[0066] in, This represents the complementary error function.

[0067] According to the present invention, a computer-readable storage medium storing a computer program is provided, wherein when the computer program is executed by a processor, the steps of the optical communication nonlinear compensation method are implemented.

[0068] The electronic device provided by the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the optical communication nonlinear compensation method.

[0069] Compared with the prior art, the present invention has the following beneficial effects:

[0070] By adopting a fusion model of exponential decay and learnable DBP, the problem of high algorithm complexity and poor accuracy of nonlinear algorithms in optical fiber communication under long distance and high baud rate was solved. The complexity was reduced to 1% of the original model while the nonlinear compensation performance remained unchanged. The results were verified under different power, number of channels, transmission distance and baud rate. Attached Figure Description

[0071] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0072] Figure 1 This is a flowchart of the present invention;

[0073] Figure 2 This is a comparison chart of Q factors with different power levels under single-channel conditions according to the present invention; EP-LDBP represents the effect of the present invention.

[0074] Figure 3 This is a comparison chart of Q factors with different power levels under 5-channel conditions, and EP-LDBP represents the effect of this invention;

[0075] Figure 4 This is a comparison chart of Q factors with different power levels under 10-channel conditions, and EP-LDBP represents the effect of this invention;

[0076] Figure 5a and Figure 5b To compare the complexity ratio of TD-LDBP and FD-LDBP at different distances, EP-LDBP represents the effect of this invention;

[0077] Figure 6aand Figure 6b To illustrate the complexity ratio of TD-LDBP and FD-LDBP at different baud rates, EP-LDBP represents the effect of this invention. Detailed Implementation

[0078] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0079] Example 1:

[0080] like Figure 1 This invention provides a nonlinear compensation method for optical communication based on a fusion model of exponential decay and learnable DBP, comprising:

[0081] Step S1: Establish the exponential decay model; based on the communication channel conditions, the initial parameter range of the exponential decay model can be obtained;

[0082] Step S2: Based on the initial parameter range of the exponential decay model in step S1, set different parameters for the exponential decay model to obtain learnable DBP models with different structures and their corresponding model complexities.

[0083] Step S3: Based on the different structures of the learnable DBP model obtained in Step S2, and in conjunction with the DBP model initialization parameter method, initialize the parameters of the learnable DBP model obtained in S2.

[0084] Step S4: Train the learnable DBP model after parameter initialization in step S3 to obtain the corresponding Q factor;

[0085] Step S5: Compare the Q-factor and model complexity of the different learnable DBP models obtained in Step S4, and select the learnable DBP model with the lowest complexity and the highest Q-factor as the optimal result.

[0086] Step S6: Perform nonlinear compensation for optical fiber communication using the learned DBP model selected in step S5.

[0087] Step S1 adopts the following:

[0088] The exponential decay model is established as follows: The exponential decay model is a mathematical model, as shown below:

[0089]

[0090] in, , For model hyperparameters, For learnable DBP Number of taps in the dispersion layer. By substituting different... The number of dispersive layer taps under different layers can be calculated. .

[0091] Estimation of the initial parameter range for the exponential decay model: Based on the communication channel conditions, the initial parameter range of the exponential decay model is obtained by considering the influence of fiber dispersion on the symbol length, as shown below:

[0092]

[0093] in, For signal bandwidth, The center frequency of the signal. For signal duration, The length of the optical fiber. These are the second-order dispersion coefficients of the optical fiber. The exponential decay model can be calculated using this model. . The range of values ​​is ,when When, it indicates that the learnable DBP model has not been simplified. A larger value indicates a higher degree of simplification in the learnable DBP model. c represents the speed of light propagation.

[0094] Step S2 employs the following:

[0095] Learnable DBP model establishment: Part 1 The linear layer tap representation is as follows:

[0096] ,

[0097] in, P The fiber attenuation coefficient, These are the second-order dispersion coefficients of the optical fiber; It is a complex number; The frequency of the i-th signal is represented by t; t represents the signal duration at time t.

[0098] No. The linear layer tap representation is as follows:

[0099]

[0100] in, For fiber nonlinearity coefficients, This represents the signal amplitude.

[0101] Method for determining the structure of a learnable DBP model: Calculate the number of dispersive layer taps at different levels within the learnable DBP using an exponential decay model. This allows us to determine the structure of the learnable DBP model.

[0102] Learnable methods for determining the complexity of DBP models:

[0103]

[0104] in, is the number of fiber optic segments; RM represents the total complexity.

[0105] Step S4 employs the following:

[0106] Learnable DBP model parameter training method: Collect the original signal from the transmitting end and the signal from the receiving end after passing through an optical communication nonlinear compensation algorithm based on an exponential decay and learnable DBP fusion model, and calculate the bit error rate. Then through the bit error rate Calculate the loss function Then, backpropagation is performed to update the parameters of the dispersive and nonlinear layers.

[0107] in, The complementary error function is represented as follows:

[0108] .

[0109] like Figure 2 The figure shows the comparison results between the proposed method and the traditional CDC, DBP, and FD-LDBP methods under different transmit powers in a single channel. As can be seen from the figure, the compensation effect obtained by the proposed method at different transmit powers is close to that of the FD-LDBP method, and its performance is higher than that of CDC and DBP.

[0110] like Figure 3 The figure shows the comparison results of the proposed method with traditional CDC, DBP, and FD-LDBP methods under different transmit powers in a 5-channel WDM system. As can be seen from the figure, the compensation effect obtained by the proposed method at different transmit powers is close to that of the FD-LDBP method, and its performance is superior to CDC and DBP.

[0111] like Figure 4 The figure shows the comparison results of the proposed method with traditional CDC, DBP, and FD-LDBP methods under different transmit powers in a 10-channel WDM system. As can be seen from the figure, the compensation effect obtained by the proposed method at different transmit powers is close to that of the FD-LDBP method, and its performance is superior to CDC and DBP.

[0112] like Figure 5a and Figure 5bThe figure shows a comparison of the complexity of the proposed method and the FD-LDBP method at different transmission distances. As can be seen from the figure, the complexity of EP-LDBP is about 1% of that of FD-LDBP, and the complexity decreases more significantly with increasing distance.

[0113] like Figure 6a and Figure 6b The figure shows a comparison of the complexity of the proposed method and the FD-LDBP method at different baud rates. As can be seen from the figure, the complexity of EP-LDBP is about 1% of that of FD-LDBP, and the complexity decreases more significantly with increasing baud rate.

[0114] This invention transforms the learnable DBP from a time-domain model to a frequency-domain model; it integrates the exponential decay model with the learnable DBP model, reducing complexity by decreasing the number of taps in the linear layer; and it uses a grid search method to optimize the exponential decay model, finding the optimal compensation effect and the least complex structure of the learnable DBP.

[0115] Example 2:

[0116] This invention also provides an optical communication nonlinear compensation system based on a fusion model of exponential decay and learnable DBP. The optical communication nonlinear compensation system based on the fusion model of exponential decay and learnable DBP can be implemented by executing the process steps of the optical communication nonlinear compensation method based on the fusion model of exponential decay and learnable DBP. That is, those skilled in the art can understand the optical communication nonlinear compensation method based on the fusion model of exponential decay and learnable DBP as a preferred embodiment of the optical communication nonlinear compensation system based on the fusion model of exponential decay and learnable DBP.

[0117] The optical communication nonlinear compensation system provided by the present invention includes: module M1: establishing an exponential decay model and obtaining the initial parameter range of the exponential decay model according to the communication channel conditions; module M2: setting different exponential decay model parameters based on the initial parameter range of the exponential decay model, and obtaining learnable DBP models with different structures and their corresponding model complexities; module M3: initializing the parameters of learnable DBP models with different structures; module M4: training the learnable DBP models after parameter initialization to obtain the corresponding Q factor; module M5: comparing the Q factor and model complexity of different learnable DBP models, and selecting the learnable DBP model with the lowest complexity and the highest Q factor as the optimal result; module M6: using the trained optimal learnable DBP model to perform nonlinear compensation for optical fiber communication.

[0118] The module M1 includes:

[0119] The expression for the exponential decay model is:

[0120]

[0121] in, , For model hyperparameters; For learnable DBP Number of taps in the dispersion layer;

[0122] By substituting different Calculate the number of dispersive layer taps under different layers. ;

[0123] Based on the communication channel conditions, the initial parameter range of the exponential decay model is obtained by considering the influence of fiber dispersion on the symbol length. The expression is as follows:

[0124]

[0125] in, For signal bandwidth; The center frequency of the signal; The duration of the signal; This refers to the length of the optical fiber. These are the second-order dispersion coefficients of the optical fiber; The range of values ​​is ,when When, it indicates that the learnable DBP model has not been simplified. The larger the value, the higher the simplification of the learnable DBP model; c represents the speed of light propagation.

[0126] The module M2 includes:

[0127] The learnable DBP model can be established using the following expression:

[0128] ,

[0129] in, P The fiber attenuation coefficient; These are the second-order dispersion coefficients of the optical fiber; It is a complex number; The frequency of the i-th signal is represented by t; t represents the signal duration at time t.

[0130] No. The linear layer tap representation is as follows:

[0131]

[0132] in, For fiber nonlinearity coefficients, The signal amplitude;

[0133] The number of dispersive layer taps at different levels within the learnable DBP was calculated using the exponential decay model. Thus, the learnable DBP model structure is determined;

[0134] The expression for determining the complexity of a learnable DBP model is:

[0135]

[0136] in, is the number of fiber optic segments; RM represents the total complexity.

[0137] The module M4 includes:

[0138] The training process for the learnable DBP model parameters is as follows: The original signal from the transmitting end and the signal from the receiving end, after passing through an optical communication nonlinear compensation algorithm based on an exponential decay and learnable DBP fusion model, are collected, and the bit error rate is calculated. Then through the bit error rate Calculate the loss function Then, backpropagation is performed to update the parameters of the dispersive and nonlinear layers, expressed as:

[0139]

[0140] in, This represents the complementary error function.

[0141] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0142] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for nonlinear compensation in optical communication, characterized in that, include: Step S1: Establish an exponential decay model and obtain the initial parameter range of the exponential decay model based on the communication channel conditions; Step S2: Based on the initial parameter range of the exponential decay model, set different parameters for the exponential decay model to obtain learnable DBP models with different structures and their corresponding model complexities; Step S3: Initialize the parameters of learnable DBP models with different structures; Step S4: Train the learnable DBP model after parameter initialization to obtain the corresponding Q factor; Step S5: Compare the Q-factor and model complexity of different learnable DBP models, and select the learnable DBP model with the lowest complexity and the highest Q-factor as the optimal result; Step S6: Use the trained optimal learnable DBP model to perform nonlinear compensation for optical fiber communication; Step S2 includes: The learnable DBP model can be established using the following expression: ， Where P is the fiber attenuation coefficient; These are the second-order dispersion coefficients of the optical fiber; It is a complex number; The frequency of the i-th signal is represented by t; t represents the signal duration at time t. No. The linear layer tap representation is as follows: in, For fiber nonlinearity coefficients, The signal amplitude; The number of dispersive layer taps at different levels of the learnable DBP model was calculated using the exponential decay model. Thus, the learnable DBP model structure is determined; The expression for determining the complexity of a learnable DBP model is: in, is the number of fiber optic segments; RM represents the total complexity.

2. The optical communication nonlinearity compensation method according to claim 1, characterized in that, Step S1 includes: The expression for the exponential decay model is: in, , For model hyperparameters; For learnable DBP Number of taps in the dispersion layer; By substituting different Calculate the number of dispersive layer taps under different layers. ; Based on the communication channel conditions, the initial parameter range of the exponential decay model is obtained by considering the influence of fiber dispersion on the symbol length. The expression is as follows: in, For signal bandwidth; The center frequency of the signal; The duration of the signal; This refers to the length of the optical fiber. These are the second-order dispersion coefficients of the optical fiber; The range of values ​​is ,when When, it indicates that the learnable DBP model has not been simplified. The larger the value, the higher the simplification of the learnable DBP model; c represents the speed of light propagation.

3. The optical communication nonlinearity compensation method according to claim 1, characterized in that, Step S4 includes: The training process for the learnable DBP model parameters is as follows: The original signal from the transmitting end and the signal from the receiving end, after passing through an optical communication nonlinear compensation algorithm based on an exponential decay and learnable DBP fusion model, are collected, and the bit error rate is calculated. Then through the bit error rate Calculate the loss function Then, backpropagation is performed to update the parameters of the dispersive and nonlinear layers, expressed as: in, This represents the complementary error function.

4. A nonlinear compensation system for optical communication, characterized in that, include: Module M1: Establish the exponential decay model and obtain the initial parameter range of the exponential decay model based on the communication channel conditions; Module M2: Based on the initial parameter range of the exponential decay model, set different parameters for the exponential decay model to obtain learnable DBP models with different structures and their corresponding model complexities; Module M3: Initializes parameters for learnable DBP models with different structures; Module M4: Trains the learnable DBP model after parameter initialization to obtain the corresponding Q factor; Module M5: Compare the Q-factor and model complexity of different learnable DBP models, and select the learnable DBP model with the lowest complexity and the highest Q-factor as the optimal result; Module M6: Performs nonlinear compensation for fiber optic communication using the trained optimal learnable DBP model; The module M2 includes: The learnable DBP model can be established using the following expression: ， in, P The fiber attenuation coefficient; These are the second-order dispersion coefficients of the optical fiber; It is a complex number; The frequency of the i-th signal is represented by t; t represents the signal duration at time t. No. The linear layer tap representation is as follows: in, For fiber nonlinearity coefficients, The signal amplitude; The number of dispersive layer taps at different levels of the learnable DBP model was calculated using the exponential decay model. Thus, the learnable DBP model structure is determined; The expression for determining the complexity of a learnable DBP model is: in, is the number of fiber optic segments; RM represents the total complexity.

5. The optical communication nonlinear compensation system according to claim 4, characterized in that, The module M1 includes: The expression for the exponential decay model is: in, , For model hyperparameters; For learnable DBP Number of taps in the dispersion layer; By substituting different Calculate the number of dispersive layer taps under different layers. ; Based on the communication channel conditions, the initial parameter range of the exponential decay model is obtained by considering the influence of fiber dispersion on the symbol length. The expression is as follows: in, For signal bandwidth; The center frequency of the signal; The duration of the signal; This refers to the length of the optical fiber. These are the second-order dispersion coefficients of the optical fiber; The range of values ​​is ,when When, it indicates that the learnable DBP model has not been simplified. The larger the value, the higher the simplification of the learnable DBP model; c represents the speed of light propagation.

6. The optical communication nonlinearity compensation system according to claim 4, characterized in that, The module M4 includes: The training process for the learnable DBP model parameters is as follows: The original signal from the transmitting end and the signal from the receiving end, after passing through an optical communication nonlinear compensation algorithm based on an exponential decay and learnable DBP fusion model, are collected, and the bit error rate is calculated. Then through the bit error rate Calculate the loss function Then, backpropagation is performed to update the parameters of the dispersive and nonlinear layers, expressed as: in, This represents the complementary error function.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the optical communication nonlinear compensation method according to any one of claims 1 to 3.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of the optical communication nonlinear compensation method according to any one of claims 1 to 3.

Images