Method and system for calculating mutual information amount of optical fiber communication system

By enhancing the mutual information neural estimation network and combining the joint distribution and edge distribution of signal data for training, the error and instability problems of mutual information calculation in fiber communication systems are solved, and the accurate mutual information estimation of fiber channels is achieved.

CN118971956AInactive Publication Date: 2024-11-15SHANGHAI JIAOTONG UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410959424.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art has errors and instability in the calculation of mutual information amount in fiber optic communication systems, and is especially unable to adapt to the nonlinear effects of fiber optic channels and the gradient backtransmission process of neural network training.

Method used

The enhanced mutual information neural estimation network is adopted to construct the training data set and calculate the loss function, and the neural network is trained using gradient backpropagation and gradient rise algorithms, and the mutual information amount is calculated by combining the joint distribution and edge distribution of signal data.

Benefits of technology

It realizes accurate and stable mutual information estimation of the fiber optic communication system, improves the accuracy and stability of mutual information estimation, and is suitable for fiber optic channel transmission scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118971956A_ABST
    Figure CN118971956A_ABST
Patent Text Reader

Abstract

The invention provides an optical fiber communication system mutual information calculation method and system, and the method comprises the following steps: S1, carrying out the data transmission based on an optical fiber communication transmission system, and constructing a training data set; s2, inputting the training data set into a neural network, and calculating a loss function; s3, training the enhanced mutual information neural estimation network based on the calculated loss function; and S4, calculating the mutual information amount of the optical fiber communication transmission system by using the trained enhanced mutual information neural estimation network. According to the method, front and back symbol joint data of signal data is input into the neural network, data joint distribution and edge distribution are used for jointly calculating a loss function, and the enhanced mutual information neural estimation network is trained, so that accurate and stable mutual information amount estimation of the optical fiber communication system is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of communication systems, and in particular to a method and system for calculating mutual information of an optical fiber communication system, and in particular to a method for calculating mutual information of an optical fiber communication system based on an enhanced mutual information neural estimation network. Background Art

[0002] In the field of optical fiber communication transmission systems, the calculation of mutual information is of great significance to system evaluation, design and tuning. Traditional methods for calculating mutual information include Monte-Carlo approximation and Gauss-Hermite approximation. These traditional calculation methods are based on the assumption of the channel Gaussian noise model and calculate the mutual information through the mathematical expression of Gaussian distribution. However, due to the influence of the nonlinear effect of optical fiber, there are differences between the optical fiber channel and the Gaussian noise model channel in the real optical fiber communication transmission system, and the mathematical expression of the optical fiber channel cannot be directly written. Therefore, when using traditional calculation methods to calculate the mutual information of the transceiver end of the optical fiber communication transmission system, only the approximate calculation of the mutual information can be completed. There is an error between the calculation result and the real mutual information. At the same time, the traditional calculation method cannot be applied to the gradient back propagation process of neural network training.

[0003] A method and system for geometric and probabilistic joint constellation shaping based on mutual information estimation is disclosed in a patent document with publication number CN114499755A, which relates to the technical field of communication systems, including: step S1: constructing and training a mutual information estimator neural network according to the input and output of the communication channel, pre-training a geometric shaping encoder neural network and a probability generator respectively, so as to maximize the mutual information calculated based on the mutual information estimator; step S2: iteratively training the mutual information estimator, the geometric shaping encoder neural network and the probability generator neural network until convergence, so as to maximize the mutual information calculated based on the mutual information estimator, and obtain a geometric shaping constellation diagram and the probability value of each constellation point, that is, the geometric and probabilistic joint shaping under the channel condition. The present invention can calculate the mutual information of the current shaping through mutual information estimation, and then iteratively train the encoders of geometric shaping and probabilistic shaping through gradient back propagation and gradient descent, so as to explicitly improve the mutual information.

[0004] This patent document proposes a method for realizing geometric and probabilistic joint constellation shaping based on a mutual information estimator in a communication system. However, this method is implemented under an additive white Gaussian noise channel, is not adapted to the fiber channel transmission scenario, and does not optimize the training of mutual information estimation, lacking the accuracy and stability of mutual information estimation.

[0005] Therefore, it is necessary to propose a new technical solution to improve the above technical problems. Summary of the invention

[0006] In view of the defects in the prior art, an object of the present invention is to provide a method and system for calculating mutual information in an optical fiber communication system.

[0007] According to a method for calculating mutual information of an optical fiber communication system provided by the present invention, the method comprises the following steps:

[0008] Step S1: performing data transmission based on an optical fiber communication transmission system and constructing a training data set;

[0009] Step S2: Input the training data set into the neural network and calculate the loss function;

[0010] Step S3: Based on the calculated loss function, the enhanced mutual information neural estimation network is trained;

[0011] Step S4: using the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system.

[0012] Preferably, step S1 comprises the following steps:

[0013] Step S1.1: Inputting the transmitting end signal into the optical fiber channel to obtain the receiving end signal;

[0014] Step S1.2: Collect the transmitting end signal and the corresponding receiving end signal as raw data;

[0015] Step S1.3: Preprocess the acquired raw data to construct a training data set, wherein the training data set is composed of signal sampling data, and the training data set is:

[0016]

[0017] where x i =(x i-k ,…,x i ,…,x i+k ) represents the signal at the transmitting end; y i =(y i-k ,…,y i ,…,y i+k ) represents the signal at the receiving end; subscript x i Represents the i-th group of data; superscript x i represents the signal collected at the i-th moment; x i It consists of the signal collected at the i-th moment and the k symbols before and after it; N is the size of the training data set.

[0018] Preferably, step S2 comprises the following steps:

[0019] Step S2.1: construct an enhanced mutual information neural estimation network;

[0020] Step S2.2: Obtain the joint distribution of signal sampling data, where the joint distribution of signal sampling data is (x i ,y i ), obtained from the diagonal elements of the training dataset;

[0021] Step S2.3: Obtain the marginal distribution of the signal sampling data, where the marginal distribution of the signal sampling data is (x i ,y j ), obtained from the off-diagonal elements of the training dataset;

[0022] Step S2.4: Based on the joint distribution and marginal distribution of the signal sampling data, the neural network is input and divided into two parts to jointly calculate the loss function of the enhanced mutual information neural estimation network. The loss function is:

[0023]

[0024] Among them, T θ represents the enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data, which comes from all diagonal elements of the training data set; (x i ,y j ) represents the marginal distribution of the signal sampling data, which comes from all the non-diagonal elements in the matrix of length M of the training data set; N and M represent the joint distribution and marginal distribution batch data size used to calculate the loss function in the training data set of the two parts in each round of training, respectively.

[0025] Preferably, in step S3:

[0026] The loss value is calculated based on the loss function, and the enhanced mutual information neural estimation network is trained using the gradient back propagation and gradient ascent algorithms;

[0027] The training is repeated iteratively, and when the number of training times reaches a preset number, the training ends, and a trained enhanced mutual information neural estimation network is obtained.

[0028] Preferably, step S4 comprises the following steps:

[0029] Step S4.1: re-collecting the transmitting end signal and the receiving end signal of the optical fiber channel, and constructing an input data set of the enhanced mutual information neural estimation network as test data of the enhanced mutual information neural estimation network;

[0030] Step S4.2: Input the test data into the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system. In the process of calculating the mutual information, the Donsker-Varadhan expression is used as the mutual information calculation formula:

[0031]

[0032] Among them, T opt Represents the trained enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data; (x i ,y j ) represents the marginal distribution of the signal sampling data; L represents the size of the test input data; dividing by log2 represents the calculation under the unified unit to bit unit.

[0033] The present invention also provides a system for calculating mutual information of an optical fiber communication system, the system comprising the following modules:

[0034] Module M1: Data transmission based on optical fiber communication transmission system and construction of training data set;

[0035] Module M2: Input the training data set into the neural network and calculate the loss function;

[0036] Module M3: Training the enhanced mutual information neural estimation network based on the calculated loss function;

[0037] Module M4: Use the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system.

[0038] Preferably, the module M1 includes the following modules:

[0039] Module M1.1: input the transmitter signal into the optical fiber channel to obtain the receiver signal;

[0040] Module M1.2: collects the transmitter signal and the corresponding receiver signal as raw data;

[0041] Module M1.3: Preprocess the acquired raw data to construct a training data set, wherein the training data set is composed of signal sampling data, and the training data set is:

[0042]

[0043] where x i =(x i-k ,…,x i ,…,x i+k ) represents the signal at the transmitting end; y i =(y i-k ,…,y i ,…,y i+k ) represents the signal at the receiving end; subscript x i Represents the i-th group of data; superscript xi represents the signal collected at the i-th moment; x i It consists of the signal collected at the i-th moment and the k symbols before and after it; N is the size of the training data set.

[0044] Preferably, the module M2 includes the following modules:

[0045] Module M2.1: Constructing enhanced mutual information neural estimation network;

[0046] Module M2.2: Obtain the joint distribution of signal sampling data, where the joint distribution of signal sampling data is (x i ,y i ), obtained from the diagonal elements of the training dataset;

[0047] Module M2.3: Obtain the marginal distribution of signal sampling data, where the marginal distribution of signal sampling data is (x i ,y j ), obtained from the off-diagonal elements of the training dataset;

[0048] Module M2.4: Based on the joint distribution and marginal distribution of signal sampling data, the neural network is input and divided into two parts to jointly calculate the loss function of the enhanced mutual information neural estimation network. The loss function is:

[0049]

[0050] Among them, T θ represents the enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data, which comes from all diagonal elements of the training data set; (x i ,y j ) represents the marginal distribution of the signal sampling data, which comes from all the non-diagonal elements in the matrix of length M of the training data set; N and M represent the joint distribution and marginal distribution batch data size used to calculate the loss function in the training data set of the two parts in each round of training, respectively.

[0051] Preferably, in the module M3:

[0052] The loss value is calculated based on the loss function, and the enhanced mutual information neural estimation network is trained using the gradient back propagation and gradient ascent algorithms;

[0053] The iterative training is called repeatedly, and when the number of training times reaches a preset number, the training ends, and a trained enhanced mutual information neural estimation network is obtained.

[0054] Preferably, the module M4 includes the following modules:

[0055] Module M4.1: Re-collect the transmitting and receiving signals of the optical fiber channel, and construct the input data set of the enhanced mutual information neural estimation network as the test data of the enhanced mutual information neural estimation network;

[0056] Module M4.2: Input the test data into the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system. In the process of calculating the mutual information, the Donsker-Varadhan expression is used as the mutual information calculation formula:

[0057]

[0058] Among them, T opt Represents the trained enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data; (x i ,y j ) represents the marginal distribution of the signal sampling data; L represents the size of the test input data; dividing by log2 represents the calculation under the unified unit to bit unit.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] 1. The present invention adopts the method of inputting the joint data of the front and back symbols of the signal data into the neural network, using the joint distribution of the data and the marginal distribution to jointly calculate the loss function and train the enhanced mutual information neural estimation network, thereby realizing accurate and stable mutual information estimation of the optical fiber communication system;

[0061] 2. In the training process of using a mutual information neural estimation network to calculate the mutual information of a fiber-optic communication system, the present invention simultaneously inputs several data before and after the signal data into the neural network to capture channel characteristics with before-and-after correlation; at the same time, when updating the loss function, the joint distribution and marginal distribution of the signal data are divided into two parts for calculation, thereby improving the accuracy and stability of mutual information estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0063] Figure 1 It is the overall flow chart of the present invention;

[0064] Figure 2 is a training flow chart of the present invention;

[0065] Figure 3 This is a mutual information calculation flow chart of the present invention. DETAILED DESCRIPTION

[0066] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0067] Embodiment 1:

[0068] Reference Figure 1 and Figure 2 According to a method for calculating mutual information of an optical fiber communication system provided by the present invention, the method comprises the following steps:

[0069] Step S1: performing data transmission based on an optical fiber communication transmission system and constructing a training data set;

[0070] Step S1.1: Inputting the transmitting end signal into the optical fiber channel to obtain the receiving end signal;

[0071] Step S1.2: Collect the transmitting end signal and the corresponding receiving end signal as raw data;

[0072] Step S1.3: Preprocess the acquired raw data to construct a training data set, wherein the training data set is composed of signal sampling data, and the training data set is:

[0073]

[0074] where x i =(x i-k ,…,x i ,…,x i+k ) represents the signal at the transmitting end; y i =(y i-k ,…,y i ,…,y i+k ) represents the signal at the receiving end; subscript x i Represents the i-th group of data; superscript x i represents the signal collected at the i-th moment; x i It consists of the signal collected at the i-th moment and the k symbols before and after it; N is the size of the training data set.

[0075] Step S2: Input the training data set into the neural network and calculate the loss function;

[0076] Step S2.1: construct an enhanced mutual information neural estimation network;

[0077] Step S2.2: Obtain the joint distribution of signal sampling data, where the joint distribution of signal sampling data is (x i ,yi ), obtained from the diagonal elements of the training dataset;

[0078] Step S2.3: Obtain the marginal distribution of the signal sampling data, where the marginal distribution of the signal sampling data is (x i ,y j ), obtained from the off-diagonal elements of the training dataset;

[0079] Step S2.4: Based on the joint distribution and marginal distribution of the signal sampling data, the neural network is input and divided into two parts to jointly calculate the loss function of the enhanced mutual information neural estimation network. The loss function is:

[0080]

[0081] Among them, T θ represents the enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data, which comes from all diagonal elements of the training data set; (x i ,y j ) represents the marginal distribution of the signal sampling data, which comes from all the non-diagonal elements in the matrix of length M of the training data set; N and M represent the joint distribution and marginal distribution batch data size used to calculate the loss function in the training data set of the two parts in each round of training, respectively.

[0082] Step S3: Based on the calculated loss function, the enhanced mutual information neural estimation network is trained;

[0083] The loss value is calculated based on the loss function, and the enhanced mutual information neural estimation network is trained using the gradient back propagation and gradient ascent algorithms;

[0084] The training is repeated iteratively, and when the number of training times reaches a preset number, the training ends, and a trained enhanced mutual information neural estimation network is obtained.

[0085] Step S4: using the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system;

[0086] Step S4.1: re-collecting the transmitting end signal and the receiving end signal of the optical fiber channel, and constructing an input data set of the enhanced mutual information neural estimation network as test data of the enhanced mutual information neural estimation network;

[0087] Step S4.2: Input the test data into the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system. In the process of calculating the mutual information, the Donsker-Varadhan expression is used as the mutual information calculation formula:

[0088]

[0089] Among them, T opt Represents the trained enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data; (x i ,y j ) represents the marginal distribution of the signal sampling data; L represents the size of the test input data; dividing by log2 represents the calculation under the unified unit to bit unit.

[0090] The present invention also provides a system for calculating the mutual information of an optical fiber communication system. The system for calculating the mutual information of an optical fiber communication system can be implemented by executing the process steps of the method for calculating the mutual information of an optical fiber communication system. That is, those skilled in the art can understand the method for calculating the mutual information of an optical fiber communication system as a preferred implementation of the system for calculating the mutual information of an optical fiber communication system.

[0091] Embodiment 2:

[0092] The present invention also provides a system for calculating mutual information of an optical fiber communication system, the system comprising the following modules:

[0093] Module M1: Data transmission based on optical fiber communication transmission system and construction of training data set;

[0094] Module M1.1: input the transmitter signal into the optical fiber channel to obtain the receiver signal;

[0095] Module M1.2: collects the transmitter signal and the corresponding receiver signal as raw data;

[0096] Module M1.3: Preprocess the acquired raw data to construct a training data set, wherein the training data set is composed of signal sampling data, and the training data set is:

[0097]

[0098] where x i =(x i-k ,…,x i ,…,x i+k ) represents the signal at the transmitting end; y i =(y i-k ,…,y i ,…,y i+k ) represents the signal at the receiving end; subscript x i Represents the i-th group of data; superscript x i represents the signal collected at the i-th moment; x iIt consists of the signal collected at the i-th moment and the k symbols before and after it; N is the size of the training data set.

[0099] Module M2: Input the training data set into the neural network and calculate the loss function;

[0100] Module M2.1: Constructing enhanced mutual information neural estimation network;

[0101] Module M2.2: Obtain the joint distribution of signal sampling data, where the joint distribution of signal sampling data is (x i ,y i ), obtained from the diagonal elements of the training dataset;

[0102] Module M2.3: Obtain the marginal distribution of signal sampling data, where the marginal distribution of signal sampling data is (x i ,y j ), obtained from the off-diagonal elements of the training dataset;

[0103] Module M2.4: Based on the joint distribution and marginal distribution of signal sampling data, the neural network is input and divided into two parts to jointly calculate the loss function of the enhanced mutual information neural estimation network. The loss function is:

[0104]

[0105] Among them, T θ represents the enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data, which comes from all diagonal elements of the training data set; (x i ,y j ) represents the marginal distribution of the signal sampling data, which comes from all the non-diagonal elements in the matrix of length M of the training data set; N and M represent the joint distribution and marginal distribution batch data size used to calculate the loss function in the training data set of the two parts in each round of training, respectively.

[0106] Module M3: Training the enhanced mutual information neural estimation network based on the calculated loss function;

[0107] The loss value is calculated based on the loss function, and the enhanced mutual information neural estimation network is trained using the gradient back propagation and gradient ascent algorithms;

[0108] The iterative training is called repeatedly, and when the number of training times reaches a preset number, the training ends, and a trained enhanced mutual information neural estimation network is obtained.

[0109] Module M4: Calculate the mutual information of the optical fiber communication transmission system using the trained enhanced mutual information neural estimation network;

[0110] Module M4.1: Re-collect the transmitting and receiving signals of the optical fiber channel, and construct the input data set of the enhanced mutual information neural estimation network as the test data of the enhanced mutual information neural estimation network;

[0111] Module M4.2: Input the test data into the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system. In the process of calculating the mutual information, the Donsker-Varadhan expression is used as the mutual information calculation formula:

[0112]

[0113] Among them, T opt Represents the trained enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data; (x i ,y j ) represents the marginal distribution of the signal sampling data; L represents the size of the test input data; dividing by log2 represents the calculation under the unified unit to bit unit.

[0114] Embodiment 3:

[0115] The present invention provides a method for calculating mutual information of an optical fiber communication system based on an enhanced mutual information neural estimation network, comprising the following steps:

[0116] Step 1: Perform data transmission based on the optical fiber communication transmission system and build a training data set;

[0117] Step 2: Input the training data set into the neural network and calculate the loss function in two parts;

[0118] Step 3: Based on the calculated loss function, train the enhanced mutual information neural estimation network;

[0119] Step 4: Use the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system.

[0120] The step 1 comprises the following steps:

[0121] Step 1.1: Input the transmitting end signal into the optical fiber channel to obtain the receiving end signal;

[0122] Step 1.2: Collect the transmitter signal and the corresponding receiver signal as raw data;

[0123] Step 1.3: Preprocess the acquired raw data to construct a training data set, wherein the training data set is composed of signal sampling data, and the training data set is:

[0124]

[0125] where x i =(x i-k ,…,x i ,…,x i+k ) represents the signal at the transmitting end; y i =(y i-k ,…,y i ,…,y i+k ) represents the signal at the receiving end; subscript x i Represents the i-th group of data; superscript x i represents the signal collected at the i-th moment; x i It consists of the signal collected at the i-th moment and the k symbols before and after it; N is the size of the training data set.

[0126] The step 2 comprises the following steps:

[0127] Step 2.1: Construct an enhanced mutual information neural estimation network;

[0128] Step 2.2: Obtain the joint distribution of signal sampling data, where the joint distribution of signal sampling data is (x i ,y i ), obtained from the diagonal elements of the training dataset;

[0129] Step 2.3: Obtain the marginal distribution of the signal sampling data, where the marginal distribution of the signal sampling data is (x i ,y j ), obtained from the off-diagonal elements of the training dataset;

[0130] Step 2.4: Based on the joint distribution and marginal distribution of the signal sampling data, the neural network is input and divided into two parts to jointly calculate the loss function of the enhanced mutual information neural estimation network. The loss function is:

[0131]

[0132] Among them, T θ represents the enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data, which comes from all diagonal elements of the training data set; (x i ,y j) represents the marginal distribution of the signal sampling data, which comes from all the non-diagonal elements in the matrix of length M of the training data set; N and M represent the joint distribution and marginal distribution batch data size used to calculate the loss function in the training data set of the two parts in each round of training, respectively.

[0133] The step 3 comprises the following steps:

[0134] Step 3.1: Calculate the loss value based on the loss function, and train the enhanced mutual information neural estimation network using gradient back propagation and gradient ascent algorithms;

[0135] Step 3.2: Repeat the iterative training. When the number of training times reaches the preset number, the training ends and a trained enhanced mutual information neural estimation network is obtained.

[0136] The step 4 comprises the following steps:

[0137] Step 4.1: re-collect the transmitting end signal and the receiving end signal of the optical fiber channel, and construct an input data set of the enhanced mutual information neural estimation network as the test data of the enhanced mutual information neural estimation network;

[0138] Step 4.2: Input the test data into the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system. In the process of calculating the mutual information, the Donsker-Varadhan expression is used as the mutual information calculation formula:

[0139]

[0140] Among them, T opt Represents the trained enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data; (x i ,y j ) represents the marginal distribution of the signal sampling data; L represents the size of the test input data; dividing by log2 represents the calculation under the unified unit to bit unit.

[0141] Those skilled in the art may understand this embodiment as a more specific description of Embodiment 1 and Embodiment 2.

[0142] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0143] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for calculating mutual information in an optical fiber communication system, characterized in that: The method comprises the following steps: Step S1: Perform data transmission based on an optical fiber communication transmission system and construct a training data set; Step S2: Input the training data set into the neural network and calculate the loss function; Step S3: Based on the calculated loss function, the enhanced mutual information neural estimation network is trained; Step S4: using the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system.

2. The method for calculating mutual information of an optical fiber communication system according to claim 1, characterized in that: The step S1 comprises the following steps: Step S1.1: Input the transmitting end signal into the optical fiber channel to obtain the receiving end signal; Step S1.2: Collect the transmitting end signal and the corresponding receiving end signal as raw data; Step S1.3: Preprocess the acquired raw data to construct a training data set, wherein the training data set is composed of signal sampling data, and the training data set is: where x i =(x i-k ,…,x i ,…,x i+k ) represents the signal at the transmitting end; y i =(y i-k ,…,y i ,…,y i+k ) represents the signal at the receiving end; subscript x i Represents the i-th group of data; superscript x i represents the signal collected at the i-th moment; x i It consists of the signal collected at the i-th moment and the k symbols before and after it; N is the size of the training data set.

3. The method for calculating mutual information of an optical fiber communication system according to claim 1, characterized in that: The step S2 comprises the following steps: Step S2.1: construct an enhanced mutual information neural estimation network; Step S2.2: Obtain the joint distribution of signal sampling data, where the joint distribution of signal sampling data is (x i ,y i ), obtained from the diagonal elements of the training dataset; Step S2.3: Obtain the marginal distribution of the signal sampling data, where the marginal distribution of the signal sampling data is (x i ,y j ), obtained from the off-diagonal elements of the training dataset; Step S2.4: Based on the joint distribution and marginal distribution of the signal sampling data, the neural network is input and divided into two parts to jointly calculate the loss function of the enhanced mutual information neural estimation network. The loss function is: Among them, T θ represents the enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data, which comes from all diagonal elements of the training data set; (x i ,y j ) represents the marginal distribution of the signal sampling data, which comes from all the non-diagonal elements in the matrix of length M of the training data set; N and M represent the joint distribution and marginal distribution batch data size used to calculate the loss function in the training data set of the two parts in each round of training, respectively.

4. The method for calculating mutual information of an optical fiber communication system according to claim 1, characterized in that: In step S3: The loss value is calculated based on the loss function, and the enhanced mutual information neural estimation network is trained using the gradient back propagation and gradient ascent algorithms; The training is repeated iteratively, and when the number of training times reaches a preset number, the training ends, and a trained enhanced mutual information neural estimation network is obtained.

5. The method for calculating mutual information of an optical fiber communication system according to claim 1, characterized in that: The step S4 comprises the following steps: Step S4.1: re-collecting the transmitting end signal and the receiving end signal of the optical fiber channel, and constructing an input data set of the enhanced mutual information neural estimation network as test data of the enhanced mutual information neural estimation network; Step S4.2: Input the test data into the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system. In the process of calculating the mutual information, the Donsker-Varadhan expression is used as the mutual information calculation formula: Among them, T opt Represents the trained enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data; (x i ,y j ) represents the marginal distribution of the signal sampling data; L represents the size of the test input data; dividing by log2 represents the calculation under the unified unit to bit unit.

6. A mutual information calculation system for an optical fiber communication system, characterized in that: The system includes the following modules: Module M1: Data transmission based on optical fiber communication transmission system and construction of training data set; Module M2: Input the training data set into the neural network and calculate the loss function; Module M3: Training the enhanced mutual information neural estimation network based on the calculated loss function; Module M4: Use the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system.

7. The optical fiber communication system mutual information calculation system according to claim 6, characterized in that: The module M1 includes the following modules: Module M1.1: input the transmitter signal into the optical fiber channel to obtain the receiver signal; Module M1.2: collects the transmitter signal and the corresponding receiver signal as raw data; Module M1.3: Preprocess the acquired raw data to construct a training data set, wherein the training data set is composed of signal sampling data, and the training data set is: where x i =(x i-k ,…,x i ,…,x i+k ) represents the signal at the transmitting end; y i =(y i-k ,…,y i ,…,y i+k ) represents the signal at the receiving end; subscript x i Represents the i-th group of data; superscript x i represents the signal collected at the i-th moment; x i It consists of the signal collected at the i-th moment and the k symbols before and after it; N is the size of the training data set.

8. The optical fiber communication system mutual information calculation system according to claim 6, characterized in that: The module M2 includes the following modules: Module M2.1: Constructing enhanced mutual information neural estimation network; Module M2.2: Obtain the joint distribution of signal sampling data, where the joint distribution of signal sampling data is (x i ,y i ), obtained from the diagonal elements of the training dataset; Module M2.3: Obtain the marginal distribution of signal sampling data, where the marginal distribution of signal sampling data is (x i ,y j ), obtained from the off-diagonal elements of the training dataset; Module M2.4: Based on the joint distribution and marginal distribution of signal sampling data, the neural network is input and divided into two parts to jointly calculate the loss function of the enhanced mutual information neural estimation network. The loss function is: Among them, T θ represents the enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data, which comes from all diagonal elements of the training data set; (x i ,y j ) represents the marginal distribution of the signal sampling data, which comes from all the non-diagonal elements in the matrix of length M of the training data set; N and M represent the joint distribution and marginal distribution batch data size used to calculate the loss function in the training data set of the two parts in each round of training, respectively.

9. The optical fiber communication system mutual information calculation system according to claim 6, characterized in that: In the module M3: The loss value is calculated based on the loss function, and the enhanced mutual information neural estimation network is trained using the gradient back propagation and gradient ascent algorithms; The iterative training is called repeatedly, and when the number of training times reaches a preset number, the training ends, and a trained enhanced mutual information neural estimation network is obtained.

10. The optical fiber communication system mutual information calculation system according to claim 6, characterized in that: The module M4 includes the following modules: Module M4.1: Re-collect the transmitting and receiving signals of the optical fiber channel, and construct the input data set of the enhanced mutual information neural estimation network as the test data of the enhanced mutual information neural estimation network; Module M4.2: Input the test data into the trained enhanced mutual information neural estimation network to calculate the mutual information of the optical fiber communication transmission system. In the process of calculating the mutual information, the Donsker-Varadhan expression is used as the mutual information calculation formula: Among them, T opt Represents the trained enhanced mutual information neural estimation network function; (x i ,y i ) represents the joint distribution of signal sampling data; (x i ,y j ) represents the marginal distribution of the signal sampling data; L represents the size of the test input data; dividing by log2 represents the calculation under the unified unit to bit unit.

Citation Information

Patent Citations

  • Construction method and application of optical signal-to-noise ratio monitoring model

    CN113938188A

  • Method and system for calculating mutual information amount of optical fiber communication transmission system based on deep learning

    CN114978306A

  • Balancing method for high-speed optical communication system

    CN116488970A

  • Network traffic anomaly detection method and system

    CN116668083A