CN110932809B - Fiber channel model simulation method, device, electronic equipment and storage medium - Google Patents

Abstract

Embodiments of the present invention provide a method, device, electronic device, and storage medium for simulating a fiber channel model. The method includes: acquiring signal data required for fiber simulation transmission; inputting the signal data and fiber length parameters into a preset A deep neural network model, according to the output result of the deep neural network model, to determine the signal data output by the optical fiber channel of the optical fiber length; wherein, the deep neural network model, according to the determined optical fiber length and the sample signal of the channel output result data, obtained after training. The trained deep neural network network can realize high-speed and high-robust Fibre Channel simulation of signal data. Compared with current methods, it has lower complexity, which greatly reduces the expertise and complexity required to model the fiber channel. As long as sufficient input and output data and distance parameters are obtained, any fiber channel can be modeled , and the time required for modeling and model running is shorter.

Description

Fibre channel model simulation method, apparatus, electronic device and storage medium

technical field

The present invention relates to the technical field of optical communication, and in particular, to a method, device, electronic device and storage medium for simulating a fiber channel model.

Background technique

In communication, channels are often divided into microwave channels, optical fiber channels, cable channels, etc. according to their physical composition. Assuming that the transmission characteristics of the channel are known, the channel can be abstractly described by a model, and the mathematical characteristics of its input/output signals can be used to describe the channel. and the mathematical characteristics of the relationship between the input/output signals.

In the field of optical communication, the traditional calculation method of analog signal transmission through optical fiber is to solve the nonlinear Schrödinger equation using the fractional-step Fourier method, and also consider stimulated Raman scattering, four-wave mixing, self-phase modulation, dispersion and attenuation. and so on factors. The time complexity of the algorithm is relatively high, and the required time increases significantly with the increase of the transmission distance. According to the situation of different fiber channels, such calculation is more complicated and not accurate enough. The current fiber modeling method is complex, and there are still some disadvantages that it is difficult to accurately model some situations.

SUMMARY OF THE INVENTION

In order to solve the above problems, embodiments of the present invention provide a method, apparatus, electronic device, and storage medium for simulating a fiber channel model.

In a first aspect, an embodiment of the present invention provides a method for simulating a fiber channel model, including: acquiring signal data required for fiber simulation transmission; inputting the signal data and fiber length parameters into a preset deep neural network model, according to The output result of the deep neural network model determines the signal data output by the optical fiber channel of the optical fiber length; wherein, the deep neural network model is obtained after training according to the determined optical fiber length and the sample signal data of the channel output result .

Further, before inputting the signal data and the fiber length parameter into the preset deep neural network model, the method further includes: sampling each bit of the signal data according to a preset sampling rate; correspondingly , inputting the signal data and fiber length parameters into a preset deep neural network model, specifically: inputting the signal data and fiber length parameters sampled at a preset sampling rate into a preset deep neural network network model.

Further, the deep neural network model is specifically a bidirectional long short-term memory neural network model.

Further, the inputting the signal data and the fiber length parameter into the preset deep neural network model includes: according to the time series of the signal data, inputting the data per bit of the signal data and the fiber length parameter. Combination, input to the forward LSTM layer and reverse LSTM layer of the bidirectional long short-term memory neural network model through the input layer; the output results of the forward LSTM layer and the reverse LSTM layer are jointly input to the fully connected layer, and from the output The layer obtains the signal transmission result data.

Further, before the inputting the signal data and the fiber length parameter into the preset deep neural network model, the method further includes: acquiring a plurality of signal data samples, and the corresponding signal data samples in the signal of the fiber channel of the determined length. Transmission result data; take the combination of pre-transmission signal data, fiber length and signal transmission result data corresponding to each signal data sample as a training sample, thereby obtaining multiple training samples, and using the multiple training samples for the bidirectional Short-term memory neural network model for training.

Further, using the plurality of training samples to train the bidirectional long short-term memory neural network model includes: inputting the pre-transmission signal and fiber length parameter of any signal data sample into the bidirectional long short-term memory. The neural network model uses a forward propagation algorithm to calculate the signal transmission result data of the signal data sample under the length parameter; based on the back propagation algorithm, the model parameters of the bidirectional long short-term memory neural network model are updated; The error between the output and the input of the bidirectional long short-term memory neural network model is calculated, and the accuracy rate of the bidirectional long short-term memory neural network model is calculated. If the accuracy rate is greater than a preset threshold or the number of training times reaches a preset number, the The training of the long short-term memory neural network model is completed.

Further, the backpropagation algorithm is a gradient descent based backpropagation algorithm.

In a second aspect, an embodiment of the present invention provides an optical fiber channel model simulation device, including: a receiving module for acquiring signal data to be transmitted by optical fiber simulation; a processing module for inputting the signal data and optical fiber length parameters into a preset The deep neural network model, according to the output result of the deep neural network model, to determine the signal data of the optical fiber channel output of the optical fiber length; wherein, the deep neural network model, according to the determined optical fiber length and the channel output result sample data Signal data, obtained after training.

In a third aspect, an embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the program, the fiber channel model simulation of the first aspect of the present invention is implemented. steps of the method.

In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the steps of the method for simulating a fiber channel model in the first aspect of the present invention.

In the fiber channel model simulation method, device, electronic device, and storage medium provided by the embodiments of the present invention, the preset deep neural network model is obtained after training according to the determined fiber length and sample signal data of the channel output result, and the output signal data can be The output result of a fixed-length Fibre Channel can be used as an accurate Fibre Channel model. The trained deep neural network network can realize the Fibre Channel simulation of signal data with high precision and high robustness. Compared with the current method, it has lower complexity, which greatly reduces the professional knowledge and complexity required for the traditional method to model the fiber channel. Using the deep learning technology of deep neural network, as long as sufficient input and output data and Any Fibre Channel can be modeled using the distance parameter in less time.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description These are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative efforts.

1 is a flowchart of a method for simulating a fiber channel model provided by an embodiment of the present invention;

2 is a schematic structural diagram of a fiber channel model provided by an embodiment of the present invention;

3 is a comparison diagram of an output waveform and a label waveform of a fiber channel model simulation method provided by an embodiment of the present invention;

4 is a comparison diagram of the mean square error between the waveform generated by the BiLSTM algorithm, the BP-ANN, and the BiRNN algorithm provided by the embodiment of the present invention and the label waveform;

5 is a comparison diagram of the frequency domain error between the BiLSTM algorithm and the BP-ANN, BiRNN algorithm generated waveform and the label waveform provided by the embodiment of the present invention;

6 is a structural diagram of an apparatus for simulating a fiber channel model provided by an embodiment of the present invention;

FIG. 7 is a schematic diagram of a physical structure of an electronic device according to an embodiment of the present invention.

Detailed ways

In order to make the purposes, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments These are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

FIG. 1 is a flowchart of a method for simulating a fiber channel model provided by an embodiment of the present invention. As shown in FIG. 1 , an embodiment of the present invention provides a method for simulating a fiber channel model, including:

101. Acquire signal data that needs to be simulated and transmitted by the optical fiber.

In the channel transmission in the prior art, after the binary sequence is formed into a pulse waveform, it is converted into an analog signal by digital-to-analog, and then converted into an optical signal by a semiconductor laser, and transmitted through an optical fiber. At the receiving end of the optical fiber, the photodiode is converted into an electrical signal, and then the digital signal transmitted by the optical fiber channel is obtained through analog-to-digital conversion.

In 101, in this embodiment, the above process is mainly simulated. The signal data needs to be transmitted through the simulated fiber channel, and the result after transmission through the fiber channel is obtained. The signal data can be a set of binary sequences, such as OOK modulated signals. In this embodiment, a single-mode fiber channel is used as an example, and signal data to be simulated and transmitted is obtained first. The simulation software used in this embodiment is VPItransmissionMaker 8.6, and a simulation module for transmitting OOK signals through a single-mode optical fiber is built on the simulation software, and the signal data is obtained based on this module.

102. Input the signal data and the optical fiber length parameter into a preset deep neural network model, and determine the signal data output by the optical fiber channel of the optical fiber length according to the output result of the deep neural network model.

Machine learning techniques have greatly facilitated the development of fields such as computer vision, natural language processing, and speech processing. In recent years, machine learning technology has also been applied to the field of optical communication, including optical performance detection, nonlinear damage compensation and channel monitoring compensation, mainly including K-proximity algorithm, SVM support vector machine, etc. It is also gradually applied to the physical channel of optical communication, but because the physical channel is more complicated to process according to different scenarios, the learning ability of traditional machine learning methods is limited. The neural network algorithm of deep learning has extremely powerful learning ability, and different structures can be used for classification, image processing, time series processing and other problems. After the neural network is trained, it is equivalent to a function, which can process and transform the data with low time complexity.

Using the powerful learning capabilities of deep learning neural networks, Fibre Channel modeling can be achieved. Only a certain amount of fiber input and output data is required to complete the modeling. The method is simple and does not require strong professional knowledge, and the running time of the neural network after training is short and does not increase with the increase of transmission distance.

In 102, the preset deep neural network model is obtained after training with sample signal data. The sample signal data is the signal data for which the fiber length of the fiber channel and the actual channel transmission result corresponding to the fiber length are known in advance, and the corresponding known channel output result is used as the label of each sample signal data. After the deep neural network model is established, a large amount of such sample signal data is used for training to obtain a preset deep neural network model. For the signal data that needs to be simulated and transmitted subsequently, the data of each bit of the signal data and the channel to be simulated are calculated. The fiber length is input to the preset deep neural network model, and the output result of the corresponding channel can be quickly and accurately obtained.

By using the deep neural network model, accurate simulation results can be obtained.

Preferably, factors such as attenuation factor, dispersion factor, nonlinear factor, etc. of the fiber channel to be modeled and simulated are determined, and only the transmission distance can be adjusted.

The deep neural network can be set according to the needs, such as including BP-ANN (back-propagation artificial neural network), BiRNN (bidirectional recurrent neural network) and BiRNN (bidirectional long short-term memory neural network), etc.

In the fiber channel model simulation method provided by the embodiment of the present invention, the preset deep neural network model is obtained after training according to the determined fiber length and the sample signal data of the channel output result, and can output the output result of the signal data in the fiber channel of a given length, Thus, it can be used as an accurate Fibre Channel model. The trained deep neural network network can realize the Fibre Channel simulation of signal data with high precision and high robustness. Compared with the current method, it has lower complexity, which greatly reduces the professional knowledge and complexity required for the traditional method to model the fiber channel. Using the deep learning technology of deep neural network, as long as sufficient input and output data and Any Fibre Channel can be modeled using the distance parameter in less time.

Based on the content of the foregoing embodiment, as an optional embodiment, before inputting the signal data and the fiber length parameter into the preset deep neural network model, the method further includes: The data is sampled; correspondingly, the signal data and fiber length parameters are input into the preset deep neural network model, specifically: the signal data and fiber length parameters sampled at the preset sampling rate are input into the preset Deep Neural Network Model.

The preset sampling rate is the number of sampling bits per bit, and the accuracy of the output result can be improved after each bit of data is sampled at the preset sampling rate before being input to the neural network. For example, the signal data is a group of pseudo-random binary sequences with a total of 8192 bits, and each bit of the signal is sampled 8 bits. The sampled signal data and fiber length parameters are used as the input data of the deep neural network model, and the result data of 8-bit sampling value per bit is obtained. Correspondingly, the sample data used in the training of the deep neural network model is also a signal sampled according to the same preset sampling rate.

Based on the content of the foregoing embodiment, as an optional embodiment, the deep neural network model is specifically a bidirectional long short-term memory neural network model (BiLSTM).

BiLSTM is composed of forward LSTM and backward LSTM. LSTM is an excellent variant model of RNN. It inherits the characteristics of most RNN models and solves the "Vanishing Gradient" caused by the gradual reduction of the gradient backpropagation process. "question. LSTM is very suitable for dealing with highly time series related problems, such as machine translation, dialogue generation, encoding/decoding, etc. The preset deep neural network model sampling BiLSTM can obtain high accuracy and ensure the accuracy of the Fibre Channel model.

Based on the content of the above-mentioned embodiment, as an optional embodiment, inputting the signal data and the fiber length parameter into the preset deep neural network model includes: according to the time series of the signal data, each bit of the signal data and the The combination of fiber length parameters is input to the forward LSTM layer and reverse LSTM layer of the BiLSTM model through the input layer; the output results of the forward LSTM layer and the reverse LSTM layer are jointly input to the fully connected layer, and from the output layer. Obtain signal transmission result data.

In this embodiment, BiLSTM is used as the preset deep neural network model. The structure of the BiLSTM model mainly includes: an input layer and a forward LSTM layer composed of n forward LSTM time steps (LF1, LF2, ..., LFn). , a reverse LSTM layer consisting of n reverse LSTM time steps (LB1, LB2, ..., LBn), m fully connected layers (F1, F2, ..., Fm) and an output layer. Among them, each corresponding forward LSTM time step and reverse LSTM time step (such as LFn and LBn) are a pair, sharing one input, the input of each pair of LSTM time steps is 1-bit data and a distance length parameter, n pairs The input of the LSTM time step and the distance parameter corresponding to each time step are the input layer, and the output of n pairs of LSTM time steps is input to the fully connected layer F1, all the neuron nodes of the F1 layer and the next fully connected layer F2 The neuron node of the layer Perform full connection; connect sequentially through m full connection layers, and the last full connection layer Fm is fully connected with the output layer; the output layer outputs the optical fiber signal that has passed the corresponding transmission distance. where n and m are integers greater than 1.

FIG. 2 is a schematic structural diagram of a fiber channel model provided by an embodiment of the present invention. As shown in FIG. 2, the structure mainly includes the following parts: an input layer, three forward LSTM time steps (LF1, LF2, LF3), three Inverse LSTM time steps (LB1, LB2, LB3), two fully connected layers (F1, F2 in the figure), one output layer. Among them, each corresponding forward LSTM time step and reverse LSTM time step is a pair (LF1 and LB1, LF2 and LB2, LF3 and LB3), sharing one input, and the input of each pair of LSTM time steps is 1-bit data ( One bit in the figure) and a fiber length parameter (L in the figure), the input of the three pairs of LSTM time steps and the distance parameter corresponding to each time step are the input layer, and the outputs of the three pairs of LSTM time steps are input to the fully connected layer F1, All neuron nodes of the F1 layer are fully connected to the neuron nodes of the next fully connected layer F2, and F2 is fully connected to the output layer; the output layer outputs the optical fiber signal that has passed the corresponding transmission distance.

Correspondingly, according to the time series of the signal data, the combination of each bit data of the signal data and the fiber length parameter is respectively input to the forward LSTM layer and the reverse LSTM layer of the bidirectional long short-term memory neural network model through the input layer. layer, specifically:

According to the time series of the signal data, in the combination of each bit data of the signal data and the fiber length parameter, every three bits of data and the fiber length parameter are combined together, and are sequentially input to the forward LSTM layer and the reverse LSTM layer. Three time steps, and output the results in the order of input.

Based on the content of the foregoing embodiment, as an optional embodiment, before inputting the signal data and the fiber length parameter into the preset deep neural network model, the method further includes: acquiring a plurality of signal data samples and corresponding signal data samples The signal transmission result data of the fiber channel of the determined length; the combination of the pre-transmission signal data, the fiber length and the signal transmission result data corresponding to each signal data sample is taken as a training sample, so as to obtain multiple training samples, and use multiple training samples. samples to train the BiLSTM model.

Before inputting the signal data to be simulated and transmitted by the optical fiber into the preset deep neural network model, the neural network needs to be trained to obtain the preset neural network model that can be used as the fiber channel model. The specific steps are as follows.

First, obtain a plurality of signal data samples before transmission, and obtain the result data of the transmission of the plurality of signal data samples, and the length parameter of the corresponding fiber channel during transmission.

Secondly, the signal data before transmission of each signal data sample and the fiber length parameter are used as input, and the signal transmission result data is used as the determined output label as a sample, so as to obtain multiple training samples. Input the pre-transmission signal data and the fiber length parameter in each sample into the constructed deep neural network model, and adjust the relevant parameters of the deep neural network model according to the output results and the output labels that have been pre-determined by the sample, so as to realize the adjustment of the deep neural network model. The training process of the above-mentioned preset deep data network model is obtained.

The method for simulating a fiber channel model provided by the embodiment of the present invention obtains multiple training samples by acquiring multiple signal data samples and the signal transmission result data of the corresponding signal data samples in a fiber channel of a certain length, and using multiple training samples The deep neural network model is trained, so that for the signal data input to the deep neural network model for simulation transmission, an output result consistent with the fiber channel used by the sample can be obtained.

Based on the content of the above embodiment, as an optional embodiment, using multiple training samples to train the BiLSTM model includes: inputting the pre-transmission signal and the fiber length parameter of any signal data sample into the bidirectional long short-term memory neural network The network uses the forward propagation algorithm to calculate the signal transmission result data of the signal data sample under the length parameter; based on the back propagation algorithm, the model parameters of the BiLSTM model are updated; according to the error between the output and input of the BiLSTM model, the accuracy of the BiLSTM model is calculated. If the accuracy rate is greater than the preset threshold or the number of training times reaches the preset number of times, the BiLSTM model training is completed.

The sample signal data is input into the BiLSTM model, and the forward propagation algorithm is used to calculate the output of each neuron. And according to the output value, the back-propagation algorithm is used to update the weight value in the long-short-term memory neural network model. In backpropagation, the weights in the model can be iteratively updated through the gradient descent algorithm, and the error value of each neuron output can be calculated at the same time. Among them, the back-propagation of the error term of BiLSTM includes two directions: one is back-propagation along time, and the other is to propagate the error term to the upper layer of neurons. Based on the corresponding error term, the gradient of each weight is computed to update the weight.

Since the training of the long short-term memory neural network model is an iterative process, the trained model needs to be verified to determine the termination condition. After determining the input and output errors of the model according to the input signal data and the output signal data, the accuracy of the BiLSTM model can be calculated. If the accuracy is greater than or equal to the preset threshold, the training process for the model is ended. If the accuracy rate is less than the preset threshold, continue to repeat the training process until the preset number of training times. Find out the post-training model that satisfies the preset error through circular training and verification, then terminate the training of the model. Alternatively, the training process is terminated after reaching a certain number of training times.

Based on the content of the foregoing embodiment, as an optional embodiment, the foregoing backpropagation algorithm is a gradient descent-based backpropagation algorithm. In the embodiment of the present invention, the backpropagation of the BiLSTM model uses the gradient descent method to gradually adjust its own parameters, thereby minimizing the error between the actual output and the label value. FIG. 3 is a comparison diagram of the output waveform and the label waveform of the fiber channel model simulation method provided by the embodiment of the present invention. In this embodiment, the comparison between the output waveform of the BiLSTM model and the label waveform after 2000 times of training and 15000 times of training is shown in FIG. 3 As shown in the figure, the abscissa of each small box in the figure is time (ns), and the ordinate is optical power (μW). It shows the situation when the signal is transmitted at 20km, 50km, and 80km, respectively. It can be seen that after 15000 steps of training, the BiLSTM model has worked well.

In order to prove the advantages of the present invention based on the BiLSTM model, the BiLSTM algorithm is compared with the other two well-known machine learning algorithms, namely BP-ANN and BiRNN. In the comparison, we compared the mean square error (MSE) of the output signal waveform of the BiLSTM module and the label waveform. The formula is as follows:

为标签输出信号数据。为了从频域方面说明BiLSTM模块输出信号与标签输出信号的相似性，还比较了频域误差(FDE)，其定义为BiLSTM模块输出信号与标签输出信号经过离散快速傅里叶变换(FFT)后，对应点在复平面上的误差距离，公式如下：where m is the number of samples, y is the output signal data of the BiLSTM module,

Output signal data for the tag. In order to illustrate the similarity between the BiLSTM module output signal and the label output signal in the frequency domain, the frequency domain error (FDE) is also compared, which is defined as the discrete Fast Fourier Transform (FFT) between the BiLSTM module output signal and the label output signal. , the error distance of the corresponding point on the complex plane, the formula is as follows:

a=fft(output_series,m),

b=fft(label_series,m)

表示对BILSTM模型输出信号做m点快速傅里叶变换。fft(label_series,m) means to perform m-point fast Fourier transform on the label output signal,

Indicates that m-point fast Fourier transform is performed on the output signal of the BILSTM model.

FIG. 4 is a comparison diagram of the mean square error between the waveform generated by the BiLSTM algorithm and the BP-ANN and BiRNN algorithms provided by the embodiment of the present invention and the label waveform, and FIG. 5 is the waveform generated by the BiLSTM algorithm provided by the embodiment of the present invention and the BP-ANN and BiRNN algorithms. Frequency domain error comparison with label waveform. Specifically, as shown in Figure 4 and Figure 5, according to the MSE error and FDE error between each algorithm and the label signal at the corresponding transmission distance, the BiLSTM algorithm has obvious advantages compared with other algorithms.

BP-ANN is the most basic neural network, its ability to extract time series features is weak, it needs a lot of training data to achieve good results, and it is easy to fall into local minimum and overfitting. Compared with the BILSTM algorithm, the BiRNN algorithm lacks the partial transfer ability of the adjacent time step information, and the effect is slightly weaker than the BiLSTM algorithm.

FIG. 6 is a structural diagram of an apparatus for simulating a fiber channel model provided by an embodiment of the present invention. As shown in FIG. 6 , the apparatus for simulating a fiber channel model includes: a receiving module 601 and a processing module 602 . Among them, the receiving module 601 is used to obtain the signal data that needs to be simulated and transmitted by the optical fiber; the processing module 601 inputs the signal data and the optical fiber length parameter into the preset deep neural network model, and determines the optical fiber length according to the output result of the deep neural network model. The signal data output by the fiber channel; wherein, the deep neural network model is obtained after training according to the determined fiber length and the sample signal data of the channel output result.

The apparatus embodiments provided in the embodiments of the present invention are for implementing the foregoing method embodiments. For specific processes and details, please refer to the foregoing method embodiments, which will not be repeated here.

In the fiber channel model simulation device provided by the embodiment of the present invention, the preset deep neural network model is obtained after training according to the determined fiber length and the sample signal data of the channel output result, and can output the output result of the signal data in the fiber channel of a given length, Thus, it can be used as an accurate Fibre Channel model. The trained deep neural network network can realize the Fibre Channel simulation of signal data with high precision and high robustness. Compared with the current method, it has lower complexity, which greatly reduces the professional knowledge and complexity required for the traditional method to model the fiber channel. Using the deep learning technology of deep neural network, as long as sufficient input and output data and Any Fibre Channel can be modeled using the distance parameter in less time.

FIG. 7 is a schematic diagram of the physical structure of an electronic device according to an embodiment of the present invention. As shown in FIG. 7 , the electronic device may include: a processor (processor) 701, a communications interface (Communications Interface) 702, and a memory (memory) 703 and a bus 704, wherein the processor 701, the communication interface 702, and the memory 703 complete the communication with each other through the bus 704. The communication interface 702 may be used for information transmission of the electronic device. The processor 701 can call the logic instructions in the memory 703 to execute the method including the following: obtaining the signal data that needs to be simulated and transmitted by the fiber; inputting the signal data and the fiber length parameter into the preset deep neural network model, according to the deep neural network The output result of the model determines the signal data output by the fiber channel of the fiber length; wherein, the deep neural network model is obtained after training according to the determined fiber length and the sample signal data of the channel output result.

In addition, the above-mentioned logic instructions in the memory 703 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product in essence, or the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the above method embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes .

On the other hand, an embodiment of the present invention further provides a non-transitory computer-readable storage medium on which a computer program is stored, and the computer program is implemented by a processor to execute the transmission method provided by the above embodiments, for example, including : obtain the signal data that needs to be simulated and transmitted by the optical fiber; input the signal data and the optical fiber length parameter into the preset deep neural network model, and determine the signal data output by the optical fiber channel of the optical fiber length according to the output result of the deep neural network model; wherein, The deep neural network model is obtained after training according to the determined fiber length and the sample signal data of the channel output result.

The device embodiments described above are only illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place , or distributed to multiple network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment. Those of ordinary skill in the art can understand and implement it without creative effort.

From the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and certainly can also be implemented by hardware. Based on this understanding, the above-mentioned technical solutions can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products can be stored in computer-readable storage media, such as ROM/RAM, magnetic Disks, optical discs, etc., include instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to perform the methods of various embodiments or portions of embodiments.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or some technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (8)

1. A method for simulating a fiber channel model, comprising: Obtain the signal data that needs to be transmitted by fiber simulation; Inputting the signal data and the optical fiber length into a preset deep neural network model, and determining the signal data output by the optical fiber channel of the optical fiber length according to the output result of the deep neural network model; Wherein, the deep neural network model is specifically a bidirectional long short-term memory neural network model; Before inputting the signal data and the fiber length into the preset deep neural network model, the method further includes: Acquiring multiple signal data samples, and the signal transmission result data of the corresponding signal data samples in the fiber channel of the determined length; The combination of the pre-transmission signal data, the fiber length and the signal transmission result data corresponding to each signal data sample is used as a training sample, so as to obtain multiple training samples, and the two-way long short-term memory neural network is analyzed by using the multiple training samples. The model is trained. 2. The method for simulating a fiber channel model according to claim 1, wherein before the inputting the signal data and the fiber length into a preset deep neural network model, the method further comprises: sampling each bit of the signal data according to a preset sampling rate; Correspondingly, inputting the signal data and the fiber length into the preset deep neural network model is specifically: Input the signal data and fiber length sampled at the preset sampling rate into the preset deep neural network model. 3. The method for simulating a fiber channel model according to claim 1, wherein the inputting the signal data and the fiber length into a preset deep neural network model comprises: According to the time series of the signal data, the combination of each bit data of the signal data and the fiber length is respectively input to the forward LSTM layer and the reverse LSTM layer of the bidirectional long short-term memory neural network model through the input layer; The output results of the forward LSTM layer and the reverse LSTM layer are jointly input to the fully connected layer, and the signal transmission result data is obtained from the output layer. The method for simulating a fiber channel model according to claim 1, wherein the training of the bidirectional long short-term memory neural network model by using the plurality of training samples comprises: Input the signal data and the fiber length of any signal data sample before transmission into the two-way long short-term memory neural network model, and use the forward propagation algorithm to calculate the signal transmission result data of the signal data sample under the fiber length; Based on the back-propagation algorithm, update the model parameters of the bidirectional long short-term memory neural network model; According to the error between the output and the input of the bidirectional long short-term memory neural network model, the accuracy rate of the bidirectional long short-term memory neural network model is calculated. If the accuracy rate is greater than a preset threshold or the number of training times reaches a preset number, the The training of the bidirectional long short-term memory neural network model is completed. 5 . The method for simulating a fiber channel model according to claim 4 , wherein the backpropagation algorithm is a gradient descent-based backpropagation algorithm. 6 . 6. An apparatus for simulating a fiber channel model, comprising: The receiving module is used to obtain the signal data that needs to be transmitted by fiber simulation; The processing module inputs the signal data and the fiber length into a preset deep neural network model, and determines the signal data output by the fiber channel of the fiber length according to the output result of the deep neural network model; wherein, the The deep neural network model is specifically a bidirectional long short-term memory neural network model; Before inputting the signal data and the fiber length into the preset deep neural network model, the method further includes: Obtain multiple signal data samples and the signal transmission result data of the corresponding signal data samples in the fiber channel of a certain length; take the combination of the pre-transmission signal data, the fiber length and the signal transmission result data corresponding to each signal data sample as a training samples, thereby obtaining multiple training samples, and using the multiple training samples to train the bidirectional long short-term memory neural network model. 7. An electronic device, comprising a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor implements any one of claims 1 to 5 when the processor executes the program The steps of the Fibre Channel model simulation method described in item . 8. A non-transitory computer-readable storage medium on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the method for simulating a fiber channel model according to any one of claims 1 to 5 is implemented. step.

Images