Power splitter design method, power splitter and device

Abstract

The present disclosure provides a power splitter design method, a power splitter and a device, which can be applied to the field of optical fiber communication technology. The method includes: in response to a design request for a target power splitter, obtaining the splitting ratio information of the target power splitter; obtaining a single-frequency response spectrum corresponding to the target power splitter according to the splitting ratio information of the target power splitter; obtaining a broadband target response spectrum according to the Gaussian distribution of the single-frequency response spectrum and first preset noise information, where the target response spectrum represents the output spectrum for designing the target power splitter; using a first neural network model to output a target topology vector corresponding to the target response spectrum according to the target response spectrum; and designing the target power splitter according to the target topology vector.

Description

Technical Field

[0001] The present disclosure relates to the field of optical fiber communication technologies, and particularly to a power splitter design method, a power splitter, and a device. Background Art

[0002] Optical fiber communication technology is relatively important in communication systems. Due to limitations in design methods and process precision, the types and performance of photonic devices are difficult to meet requirements. A power splitter is the most basic photonic device, which can split the input light according to a preset splitting ratio.

[0003] In the process of implementing the inventive concept of the present disclosure, the inventors found that in the related art, simulating and designing the topological structure of a power splitter consumes too much computer computing power and the design efficiency is low. Summary of the Invention

[0004] In view of the above problems, the present disclosure provides a power splitter design method, a power splitter, and a device.

[0005] According to a first aspect of the present disclosure, there is provided a power splitter design method, including: in response to a design request for a target power splitter, obtaining splitting ratio information of the target power splitter; obtaining a single-frequency response spectrum corresponding to the target power splitter according to the splitting ratio information of the target power splitter; obtaining a broadband target response spectrum according to the Gaussian distribution of the single-frequency response spectrum and first preset noise information, where the target response spectrum represents an output spectrum for designing the target power splitter; using a first neural network model to output a target topological structure vector corresponding to the target response spectrum according to the target response spectrum; and designing the target power splitter according to the target topological structure vector.

[0006] According to an embodiment of the present disclosure, using a first neural network model to output a target topological structure vector corresponding to the target response spectrum according to the target response spectrum includes: obtaining spectral features corresponding to the target response spectrum according to the target response spectrum; inputting the spectral features into the first neural network model, and outputting a target topological structure vector corresponding to the spectral features.

[0007] According to an embodiment of the present disclosure, the above power splitter design method further includes:

[0008] obtaining a test response spectrum corresponding to the target topological structure vector through simulation according to the target topological structure vector; and determining the similarity between the test response spectrum and the target response spectrum, where the similarity is used to verify the accuracy of the target topological structure vector.

[0009] According to an embodiment of the present disclosure, the above power splitter design method further includes: in the case where the similarity is less than a preset similarity threshold, adjusting parameters in the first neural network model and re-determining the target topological structure vector.

[0010] According to an embodiment of the present disclosure, inputting a spectral feature into a first neural network model to output a target topological structure vector corresponding to the spectral feature includes: using the first neural network model and a rectified linear unit function to perform feature transformation on a plurality of spectral features to obtain the target topological structure vector, where the rectified linear unit function represents the activation function of the first neural network model.

[0011] According to an embodiment of the present disclosure, the value of each dimension in the target topological structure vector represents the material used in the target power divider, and the material is selected from silicon and silicon dioxide.

[0012] According to an embodiment of the present disclosure, the training process of the first neural network model includes:

[0013] Randomly generating a plurality of topological structure vectors; using the finite-difference time-domain method to obtain a plurality of sample single-frequency response spectra according to the plurality of topological structure vectors; obtaining a plurality of broadband response spectra according to the Gaussian distribution of the plurality of sample single-frequency response spectra and second preset noise information; obtaining a plurality of response spectrum samples according to the plurality of broadband response spectra; and inputting the plurality of response spectrum samples into the neural network model to be trained to obtain the trained first neural network model.

[0014] According to an embodiment of the present disclosure, the above power divider design method further includes: obtaining a plurality of vector samples according to the plurality of topological structure vectors; and inputting the plurality of vector samples into the neural network model to be trained to obtain the trained second neural network model.

[0015] A second aspect of the present disclosure provides a power divider designed by using the above design method.

[0016] A third aspect of the present disclosure provides a power divider design device, including: a first acquisition module, configured to acquire the splitting ratio information of a target power divider in response to a design request for the target power divider; a second acquisition module, configured to obtain a single-frequency response spectrum corresponding to the target power divider according to the splitting ratio information of the target power divider; a third acquisition module, configured to obtain a broadband target response spectrum according to the Gaussian distribution of the single-frequency response spectrum and first preset noise information, where the target response spectrum represents the output spectrum used for designing the target power divider; a first output module, configured to use the first neural network model to output a target topological structure vector corresponding to the target response spectrum according to the target response spectrum; and a design module, configured to design the target power divider according to the target topological structure vector.

[0017] According to the power splitter design method, power splitter, and device provided by the present disclosure, by simulating the broadband response spectrum through the Gaussian distribution and noise of the single-frequency response spectrum, the determination of the topological structure directly using the broadband response spectrum is avoided, saving computing power and improving efficiency. Then, by using a neural network model, the target topological structure is obtained, and the target power splitter is designed according to the target topological structure, further reducing the consumption of computer computing power and improving the design efficiency. Moreover, since the first neural network model is used to determine the target topological structure vector corresponding to the splitting ratio and the target power splitter is designed according to the target topological structure vector, the accuracy of the splitting ratio of the designed target power splitter is improved, thereby reducing the loss of the target power splitter. Description of the Drawings

[0018] Through the following description of the embodiments of the present disclosure with reference to the drawings, the above content and other objects, features, and advantages of the present disclosure will become clearer. In the drawings:

[0019] Figure 1 The application scenario diagram of the power splitter design method according to the embodiment of the present disclosure is schematically shown;

[0020] Figure 2 The flowchart of the power splitter design method according to the embodiment of the present disclosure is schematically shown;

[0021] Figure 3 The schematic diagram of the true single-frequency response according to the embodiment of the present disclosure is schematically shown;

[0022] Figure 4 The schematic diagram of the first neural network model according to the embodiment of the present disclosure is schematically shown;

[0023] Figure 5 The schematic diagram of the forward problem and the inverse problem according to the embodiment of the present disclosure is schematically shown;

[0024] Figure 6 The structural block diagram of the power splitter design device according to the embodiment of the present disclosure is schematically shown; and

[0025] Figure 7 The block diagram of the electronic device suitable for implementing the power splitter design method according to the embodiment of the present disclosure is schematically shown. Detailed Embodiments

[0026] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for the sake of explanation, numerous specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments can also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present disclosure.

[0027] The terms used herein are merely for describing specific embodiments and are not intended to limit the present disclosure. The terms "including", "comprising", etc. used herein indicate the presence of the described features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0028] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0029] In the case of using expressions such as "at least one of A, B, and C, etc.", generally, it should be interpreted according to the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but not be limited to a system having only A, having only B, having only C, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.).

[0030] In the technical solution of the present disclosure, the processing of the collection, storage, use, processing, transmission, provision, disclosure, and application, etc. of the user's personal information all comply with the provisions of relevant laws and regulations, necessary confidentiality measures are taken, and it does not violate public order and good customs.

[0031] In the technical solution of the present disclosure, the processing of the acquisition, collection, storage, use, processing, transmission, provision, disclosure, and application, etc. of data all comply with the provisions of relevant laws and regulations, necessary confidentiality measures are taken, and it does not violate public order and good customs.

[0032] In the related art, a power splitter can be designed by a reverse design method, that is, the topological structure of the power splitter is determined by the response spectrum. By the reverse design method, a photonics device that meets the requirements can be realized, and the photonics device designed by the reverse design method has the advantages of small size, stable performance, easy manufacturing, etc., and can be compatible with the semiconductor manufacturing process in the related art.

[0033] However, since traditional reverse design uses the gradient descent algorithm and requires repeated FDTD (finite-difference time-domain method) verification, it consumes a large amount of computer computing power. Moreover, the Y-branch power divider designed in related technologies has too high losses and poor performance in practical applications.

[0034] Based on this, the inventors found that neural networks can fit complex functional relationships without consuming excessive computer computing power and are suitable for fitting the complex input and output responses of photonic devices.

[0035] The training methods of neural networks generally include training methods such as supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning.

[0036] Supervised learning mainly trains the neural network model by giving labels, so the training effect can usually meet the requirements. Unsupervised learning only trains the neural network model with training data without giving labels. Therefore, the network cannot accurately know which data have which labels and can only analyze the characteristics of the data with its powerful computing ability to obtain certain results, usually some sets of data that are the same or similar in certain characteristics. Semi-supervised learning combines a large amount of unlabeled data and a small amount of labeled data during the training stage. Its basic rule is that the distribution of data is not completely random. Through the local characteristics of some labeled data and the overall distribution of more unlabeled data, a classification result that can meet the requirements can be obtained.

[0037] The essence of reinforcement learning is to solve decision-making problems, that is, to make decisions automatically and can make continuous decisions. It mainly includes four elements: agent, environmental state, action, and reward. The goal of reinforcement learning is to obtain the most cumulative rewards. The specific process of reinforcement learning can include: the agent observes the environmental state and takes actions, which will have a certain impact and change on the environmental state. Then the agent will obtain rewards from the new environment. Repeat the above steps. The goal of the agent's learning is to maximize the expected rewards. Thus, the neural network model can be trained with the preset label information and the data of the response spectrum, and then the trained neural network model is used to output a topological structure that meets the requirements.

[0038] Moreover, in related technologies, usually, the noise in the broadband response spectrum is filtered to the greatest extent, and then a large number of simulation processes are repeated according to the filtered broadband response spectrum to avoid the interference of noise on the simulation results.

[0039] However, in the process of implementing the inventive concept of the present disclosure, the inventors found that by combining the Gaussian distribution of the single-frequency response spectrum and the preset noise information, and then using the trained neural network model, a topological structure vector corresponding to the single-frequency response spectrum can be obtained, avoiding the use of the broadband response spectrum, and thus the computing power can be reduced.

[0040] Based on this, an embodiment of the present disclosure provides a power divider design method, including:

[0041] In response to a design request for a target power divider, obtaining the splitting ratio information of the target power divider;

[0042] According to the splitting ratio information of the target power divider, obtaining a single-frequency response spectrum corresponding to the target power divider;

[0043] According to the Gaussian distribution of the single-frequency response spectrum and the first preset noise information, obtaining a broadband target response spectrum, where the target response spectrum represents the output spectrum for designing the target power divider;

[0044] Using a first neural network model, according to the target response spectrum, outputting a target topological structure vector corresponding to the target response spectrum;

[0045] According to the target topological structure vector, designing the target power divider.

[0046] Figure 1 Schematically shows an application scenario diagram of the power divider design method according to an embodiment of the present disclosure.

[0047] As Figure 1 shown, the application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used to provide a medium for communication links between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0048] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0049] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices with a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0050] The server 105 may be a server providing various services, such as a background management server (only for example) that supports the websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process data such as user requests received, and feedback the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0051] It should be noted that the power splitter design method provided in the embodiments of the present disclosure can generally be executed by the server 105. Correspondingly, the power splitter design device provided in the embodiments of the present disclosure can generally be set in the server 105. The power splitter design method provided in the embodiments of the present disclosure can also be executed by a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105. Correspondingly, the power splitter design device provided in the embodiments of the present disclosure can also be set in a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105.

[0052] It should be understood that Figure 1 the numbers of terminal devices, networks, and servers in

[0053] are merely illustrative. According to actual needs, there can be any number of terminal devices, networks, and servers. Figure 1 The following will be based on Figures 2 to 5 the described scenario to describe the power splitter design method of the disclosed embodiments in detail through

[0054] Figure 2 FIG. schematically shows a flowchart of the power splitter design method according to an embodiment of the present disclosure.

[0055] As Figure 2 shown, the power splitter design method of this embodiment includes operations S210 to S250.

[0056] In operation S210, in response to a design request for a target power splitter, obtain the splitting ratio information of the target power splitter.

[0057] According to an embodiment of the present disclosure, for example, the design request for the target power splitter can be input on an electronic device or from other electronic devices. The design request for the target power splitter can include the splitting ratio information of the target power splitter.

[0058] In operation S220, according to the splitting ratio information of the target power splitter, a single-frequency response spectrum corresponding to the target power splitter is obtained.

[0059] According to an embodiment of the present disclosure, the corresponding single-frequency response spectrum can be determined according to the splitting ratio information. For example, multiple topological structures can be randomly generated first, the splitting ratios corresponding to the multiple topological structures are determined, and then the single-frequency response spectra corresponding to the multiple topological structures are determined through simulation. Furthermore, the corresponding splitting ratio and single-frequency response spectrum can be matched according to the topological structure. In the case where a target power splitter needs to be designed, the single-frequency response spectrum corresponding to the splitting ratio of the target power splitter can be determined according to the splitting ratio of the target power splitter.

[0060] According to an embodiment of the present disclosure, for example, the single-frequency response spectrum corresponding to the target power splitter can be obtained, and then the topological structure corresponding to the target power splitter can be obtained according to the single-frequency response spectrum, so as to design the target power splitter according to the topological structure.

[0061] According to an embodiment of the present disclosure, for example, the splitting ratio information can be 0.5. The topological structure corresponding to the power splitter with a splitting ratio of 0.5 can be obtained according to the splitting ratio information, and then the power splitter with a splitting ratio of 0.5 can be designed according to the topological structure.

[0062] In operation S230, according to the Gaussian distribution of the single-frequency response spectrum and the first preset noise information, a broadband target response spectrum is obtained, where the target response spectrum represents the output spectrum for designing the target power splitter.

[0063] According to an embodiment of the present disclosure, for example, the first preset noise information can include wide-frequency bands. The wide-frequency noise information can be combined with the Gaussian distribution of the single-frequency response spectrum to expand the frequency of the single-frequency response spectrum and obtain a broadband target response spectrum. By the above method, the computing power used in the process of simulating the target response spectrum can be saved and the efficiency can be improved. And since the target response spectrum is obtained from the single-frequency response spectrum and noise, the computing power used in the process of determining the topological structure according to the target response spectrum can be saved and the design efficiency can be improved.

[0064] Figure 3 A schematic diagram of the true single-frequency response according to an embodiment of the present disclosure is schematically shown.

[0065] As Figure 3As shown, the Gaussian distribution of the single-frequency response spectrum, where the abscissa represents the wavelength and the ordinate represents the transmittance of the power splitter. The black circles in the figure represent the intersection points of the dashed line and the spectrum diagram, and these intersection points are the true single-frequency responses of the single-frequency response spectrum. Among them, the transmittance corresponds to the splitting ratio, and the transmittance can be the ratio of the light output at the output end of the target power splitter to the light input at the input end. The target response spectrum can be obtained by combining the true single-frequency response and the first preset noise information to replace the broadband response spectrum of the true response. For example, the wavelength of the true single-frequency response can be 1550 nm.

[0066] In operation S240, using the first neural network model, according to the target response spectrum, output a target topology vector corresponding to the target response spectrum.

[0067] According to an embodiment of the present disclosure, the target topology vector can be used to characterize the distribution corresponding to the refractive index in the design region of the target power splitter.

[0068] According to an embodiment of the present disclosure, the first neural network model can be used to determine the topology of the optical splitter according to the response spectrum. For example, the response spectrum can be input into the first neural network model, the first neural network model extracts the features of the target response spectrum, transforms the extracted features, and outputs a topology vector corresponding to the input response spectrum.

[0069] According to an embodiment of the present disclosure, for example, the first neural network model can include an input layer, a hidden layer, and an output layer. The input layer can be used to input data of the target response spectrum; the hidden layer can be used to perform data fitting on the data of the target response spectrum to obtain the target topology vector; the output layer can be used to output the obtained target topology vector.

[0070] Figure 4 Schematically shows a schematic diagram of the first neural network model according to an embodiment of the present disclosure.

[0071] As Figure 4 shown, where 1, 2, 3, and 4 represent the corresponding data of the input target response spectrum, 7, 8, 9, and 10 represent the neurons of the input layer, 11, 12, and 13 represent the neurons of the hidden layer, 14 and 15 represent the neurons of the output layer, and 5 and 6 represent the output data. The neurons of the input layer and the neurons of the hidden layer can be fully connected, and the neurons of the hidden layer and the neurons of the output layer can be fully connected. It should be noted that the number of neurons and the number of layers in the figure are only schematic and do not limit the actual number of neurons and the number of layers.

[0072] According to an embodiment of the present disclosure, for example, the first neural network model can be a relatively simple feedforward neural network to ensure the simplicity of the model structure and improve the efficiency of training the model. And since the neural network model is used to predict the target topology, the efficiency is improved compared with the traditional simulation method.

[0073] In operation S250, according to the target topology vector, a target power splitter is designed.

[0074] According to an embodiment of the present disclosure, for example, according to the target topology vector, the topology of the target power splitter can be determined, and then the target power splitter corresponding to the splitting ratio information can be designed according to the topology. The target power splitter can have a corresponding splitting ratio and can output a response spectrum consistent with the target response spectrum.

[0075] According to an embodiment of the present disclosure, since the broadband response spectrum is simulated through the Gaussian distribution and noise of the single-frequency response spectrum, directly using the broadband response spectrum to determine the topology is avoided, saving computing power and improving efficiency. Then, using the neural network model, the target topology is obtained, and the target power splitter is designed according to the target topology, further reducing the consumption of computer computing power and improving the design efficiency. And, since the first neural network model is used to determine the target topology vector corresponding to the splitting ratio and the target power splitter is designed according to the target topology vector, the accuracy of the splitting ratio of the designed target power splitter is improved, thereby reducing the loss of the target power splitter.

[0076] According to an embodiment of the present disclosure, the number of hidden layers of the first neural network model can be set according to requirements, so that the first neural network model can fit a complex function model and improve the accuracy of the target topology vector.

[0077] According to an embodiment of the present disclosure, by adjusting the splitting ratio, an optical attenuator can be designed. For example, since the ratio of the splitting ratio of the power splitter is 0.5, if the ratio of the splitting ratio is adjusted to 0.3, the output of the output channel of the designed power splitter can be changed to 60% of the original power splitter to achieve light energy attenuation, and this power splitter can be used as an optical attenuator.

[0078] According to an embodiment of the present disclosure, using the first neural network model, according to the target response spectrum, the target topology vector corresponding to the target response spectrum is output, including:

[0079] According to the target response spectrum, the spectral features corresponding to the target response spectrum are obtained;

[0080] The spectral features are input into the first neural network model, and the target topology vector corresponding to the spectral features is output.

[0081] According to an embodiment of the present disclosure, for example, spectral features extracted from a target response spectrum can be input into a first neural network model, and then the first neural network performs feature transformation on the input spectral features to obtain a topological structure vector corresponding to the spectral features and output it.

[0082] According to an embodiment of the present disclosure, since the first neural network model is used to output a target topological structure vector corresponding to the spectral features, computing power is saved and the efficiency of obtaining the target topological structure vector is improved.

[0083] According to an embodiment of the present disclosure, the above power splitter design method further includes:

[0084] According to the target topological structure vector, a test response spectrum corresponding to the target topological structure vector is obtained through simulation;

[0085] Determine the similarity between the test response spectrum and the target response spectrum, where the similarity is used to verify the accuracy of the target topological structure vector.

[0086] According to an embodiment of the present disclosure, for example, according to the target topological structure vector, a response spectrum corresponding to the target topological structure vector can be determined through a simulation method, and then according to the similarity between the response spectrum and the target response spectrum, it can be determined whether the target topological structure vector meets the requirements. When the target topological structure vector meets the requirements, a power splitter with a splitting ratio meeting the requirements can be designed according to the target topological structure vector.

[0087] According to an embodiment of the present disclosure, the test response spectrum can be a response spectrum corresponding to the target topological structure vector. By determining the similarity between the test response spectrum and the target response spectrum, it can be determined whether the target topological structure vector meets the requirements. For example, when the similarity is greater than a preset similarity threshold, it can be determined that the target topological structure vector meets the requirements.

[0088] According to an embodiment of the present disclosure, since the similarity between the test response spectrum and the target response spectrum is determined, the accuracy of the target topological structure vector can be determined according to the similarity, and thus the accuracy of the target topological structure vector can meet the requirements.

[0089] According to an embodiment of the present disclosure, for example, the splitting ratio corresponding to the test response spectrum can be determined. By determining the difference between the splitting ratios corresponding to the test response spectrum and the target response spectrum, the accuracy of the target topological structure vector can be determined. When the difference between the splitting ratios corresponding to the test response spectrum and the target response spectrum is less than a preset value, it can be determined that the target topological structure vector meets the requirements.

[0090] According to an embodiment of the present disclosure, the above power splitter design method further includes:

[0091] When the similarity is less than a preset similarity threshold, the parameters in the first neural network model are adjusted to redetermine the target topological structure vector.

[0092] According to the embodiments of the present disclosure, for example, different similarity thresholds can be set according to requirements, and the similarity threshold can be 90%. When the similarity is greater than or equal to 90%, it can be determined that the target topological structure vector meets the requirements; when the similarity is less than 90%, the parameters in the first neural network model can be adjusted to redetermine the target topological structure vector.

[0093] According to an embodiment of the present disclosure, when the similarity is less than a preset similarity threshold, it can be determined that the target topology structure vector does not meet the requirements. The target topology structure vector can be re-determined by adjusting the parameters in the first neural network, and then the similarity with the target response spectrum can be determined based on the test response spectrum corresponding to the re-determined target topology structure vector. When the similarity is greater than the preset similarity threshold, it can be determined that the target topology structure vector meets the requirements, and the target power divider can be designed based on the target topology structure vector that meets the requirements.

[0094] According to an embodiment of the present disclosure, when the similarity is less than a preset similarity threshold, the parameters in the first neural network model are adjusted and the target topology structure vector is re-determined, thereby improving the accuracy of the target topology structure vector and further improving the accuracy of the designed target power divider.

[0095] According to an embodiment of the present disclosure, for example, when the difference between the splitting ratio corresponding to the test response spectrum and the splitting ratio corresponding to the target response spectrum is greater than a preset value, the parameters in the first neural network model can be adjusted and the target topological structure vector can be re-determined so that the above difference is less than the preset value, and then it can be determined that the target topological structure vector meets the requirements.

[0096] According to an embodiment of the present disclosure, for example, the first neural network model may further include an activation function and a parameter matrix. The activation function is used to transform the data output by the hidden layer; the parameter matrix is ​​used to characterize the transformation relationship between the output data of each hidden layer, and the parameter matrix may be initialized by randomly taking values ​​between 0 and 1.

[0097] According to an embodiment of the present disclosure, a spectral feature is input into a first neural network model, and a target topological structure vector corresponding to the spectral feature is output, including:

[0098] The first neural network model and the linear rectification function are used to perform feature transformation on multiple spectral features to obtain a target topological structure vector, wherein the linear rectification function represents the activation function of the first neural network model.

[0099] According to an embodiment of the present disclosure, for example, activation functions play an important role in enabling a neural network model to learn and understand complex non-linear functions. Non-linear characteristics can be introduced into the neural network through activation functions. In a neural network model, the data input by the input layer can be adjusted by an activation function after being processed by the hidden layer. Introducing an activation function can increase the non-linear characteristics of the neural network model. For example, hidden layers without activation functions are equivalent to matrix multiplications, i.e., linear operations.

[0100] Based on this, the rectified linear unit (RELU) can be used as the activation function of the first neural network model to control the output of the hidden layer in the first neural network model. Compared with the traditional sigmoid activation function, the RELU activation function has fewer problems and is less affected by issues such as vanishing gradients, making it more suitable for neural networks with a deeper depth. Therefore, by using the RELU activation function, the number of hidden layers of the first neural network can be increased as required.

[0101] According to an embodiment of the present disclosure, in the target topology structure vector, the value of each dimension represents the material used for the target power divider, and the material is selected from silicon and silicon dioxide.

[0102] According to an embodiment of the present disclosure, for example, the target topology structure vector can be represented in binary form. The silicon-on-insulator (SOI) process can be used to represent the transmittance of the power divider designed from the target topology structure using silicon and silicon dioxide. For example, the target topology structure vector can be a 20*20 matrix, that is, the matrix can contain 400 elements. The element values in the matrix can be 0 or 1. 0 can represent that the material corresponding to the element is silicon dioxide, and 1 can represent that the material corresponding to the element is silicon.

[0103] According to an embodiment of the present disclosure, since the materials used for the target power divider are selected from silicon and silicon dioxide, complex materials and processes are avoided, which can improve the quality of the target power divider.

[0104] According to an embodiment of the present disclosure, for example, the corresponding splitting ratio information can be determined through the transmittance or the refractive index of the power divider, and then the corresponding single-frequency response spectrum can be obtained based on the splitting ratio information.

[0105] According to an embodiment of the present disclosure, for example, a neural network model to be trained can be trained by constructing a forward problem and a reverse problem. The forward problem can be used to train the neural network model to be trained using the topology structure to obtain a second neural network model; the reverse problem can be used to train the neural network model to be trained using the response spectrum to obtain a first neural network model.

[0106] Figure 5A schematic diagram showing the forward problem and the inverse problem according to an embodiment of the present disclosure is schematically illustrated.

[0107] As Figure 5 shown, the upper half of the figure represents the forward problem, that is, taking the topological structure as the input data of the model and the response spectrum as the label for model training; the lower half of the figure represents the inverse problem, that is, taking the response spectrum as the input data of the model and the topological structure as the label for model training. It should be noted that the spectral response in the figure does not represent the real spectral response, and the number of sampling points on the spectral response does not represent the real number of sampling points either.

[0108] According to an embodiment of the present disclosure, the training process of the first neural network model includes:

[0109] Randomly generate a plurality of topological structure vectors;

[0110] Using the finite-difference time-domain method, according to a plurality of topological structure vectors, obtain a plurality of sample single-frequency response spectra;

[0111] According to the Gaussian distribution of a plurality of sample single-frequency response spectra and the second preset noise information, obtain a plurality of broadband response spectra;

[0112] According to a plurality of broadband response spectra, obtain a plurality of response spectrum samples;

[0113] Input a plurality of response spectrum samples into the neural network model to be trained to obtain the trained first neural network model.

[0114] According to an embodiment of the present disclosure, for example, the finite-difference time-domain method can be used to simulate the corresponding single-frequency response spectrum through the randomly generated topological structure, and then the single-frequency response spectrum and the second preset noise information are combined to obtain a broadband response spectrum. Response spectrum samples can be obtained according to the broadband response spectrum, and then the neural network model can be trained using the response spectrum samples. For example, a plurality of topological structure vectors can be randomly generated first, and then the FDTD module of Lumerical software can be used to simulate the above-mentioned plurality of sample single-frequency response spectra corresponding to the plurality of topological vectors by using the finite-difference time-domain method. Since the broadband response spectrum combined with the sample single-frequency response spectrum and noise is used to train the neural network model, computer resources can be saved and the training efficiency can be improved. For example, the training set and the test set can be divided according to the response spectrum samples and the corresponding topological structure vectors. Each training set includes a plurality of response spectrum samples and the topological structure vectors corresponding to the plurality of response spectrum samples, and each test set includes a plurality of response spectrum samples and the topological structure vectors corresponding to the plurality of response spectrum samples. The training set can be used to train the neural network model, and the test set can be used to test the accuracy of the trained neural network model.

[0115] For example, after training a neural network model using a training set, the trained neural network model can be tested using a test set. If the test results do not meet the requirements, the neural network model can be adjusted and retrained; if the test results meet the requirements, the neural network model can be used as the first neural network model.

[0116] Also for example, the topological structure vector output by the neural network model during the training process can be used as the test topological structure vector, and the similarity between the test topological structure vector and the topological structure vector corresponding to the response spectrum sample can be determined; if the similarity meets the preset value, the neural network model can be determined as the first neural network model; if the similarity does not meet the preset value, the parameters of the neural network model can be adjusted and the neural network model can be retrained. For example, the preset value can be 90%. According to an embodiment of the present disclosure, for example, 5000 matrices of size 63×1 can be used as the response spectrum samples, and the values in the matrices represent the spectral responses at a certain wavelength. In 63*1, 63 can correspond to the transmittance of the spectral response, and 1 can correspond to the wavelength.

[0117] According to an embodiment of the present disclosure, for example, the first neural network model can be a 12-layer neural network model. The input layer can include 400 neurons, and these 400 neurons can correspond to a 20×20 matrix of the target topological structure vector. The number of hidden layers can be 10 layers, and the number of neurons in each layer can decrease layer by layer in the propagation direction. The output layer can include 63 neurons, which can correspond to a matrix of 63×1 response spectrum samples.

[0118] According to an embodiment of the present disclosure, for example, the GD (gradient descent) algorithm can be used to train the neural network model to minimize the difference between the final output data and the true output data of the neural network model. The adam gradient descent algorithm in the Keras module of Python software can be used to update the parameter matrix.

[0119] According to an embodiment of the present disclosure, the above power splitter design method further includes:

[0120] Obtaining a plurality of vector samples according to a plurality of topological structure vectors;

[0121] Inputting the plurality of vector samples into a neural network model to be trained to obtain a trained second neural network model.

[0122] According to an embodiment of the present disclosure, the second neural network model can be used to output a corresponding response spectrum according to the input topological structure vector. According to an embodiment of the present disclosure, for example, a training set and a test set can be divided based on vector samples and the corresponding broadband response spectra. Each training set includes a plurality of vector samples and the corresponding broadband response spectra of the plurality of vector samples, and each test set includes a plurality of vector samples and the corresponding broadband response spectra of the plurality of vector samples. The training set can be used to train the neural network model, and the test set can be used to test the accuracy of the trained neural network model.

[0123] For example, after training the neural network model using the training set, the trained neural network model can be tested using the test set. If the test results do not meet the requirements, the neural network model can be adjusted and retrained; if the test results meet the requirements, the neural network model can be used as the second neural network model.

[0124] Also for example, the response spectrum output by the neural network model during the training process can be used as the test spectrum, and the similarity between the test spectrum and the corresponding broadband response spectrum of the vector sample can be determined; if the similarity meets the preset value, the neural network model can be determined as the second neural network model; if the similarity does not meet the preset value, the parameters of the neural network model can be adjusted and the neural network model can be retrained. For example, the preset value can be 90%.

[0125] The embodiments of the present disclosure also provide a power divider designed by using the design method of the embodiments of the present disclosure.

[0126] According to an embodiment of the present disclosure, by simulating the broadband response spectrum through the Gaussian distribution and noise of the single-frequency response spectrum, the determination of the topological structure directly using the broadband response spectrum is avoided, saving computing power and improving efficiency. Then, using the neural network model, the target topological structure is obtained, and the target power divider is designed according to the target topological structure, further reducing the consumption of computer computing power and improving the design efficiency. And because the first neural network model is used to determine the target topological structure vector corresponding to the splitting ratio and the target power divider is designed according to the target topological structure vector, the accuracy of the splitting ratio of the designed target power divider is improved, and thus the loss of the target power divider is reduced.

[0127] Based on the above power divider design method, the present disclosure also provides a power divider design device. The following will be combined with Figure 6 to describe the device in detail.

[0128] Figure 6 The structural block diagram of the power divider design device according to an embodiment of the present disclosure is schematically shown.

[0129] AsFigure 6 As shown in Figure 6 , the power divider design device 600 of this embodiment includes a first acquisition module 610, a second acquisition module 620, a third acquisition module 630, an output module 640, and a design module 650.

[0130] The first acquisition module 610 is configured to acquire the splitting ratio information of the target power divider in response to a design request for the target power divider. In one embodiment, the first acquisition module 610 may be used to perform the operation S210 described above, which will not be elaborated here.

[0131] The second acquisition module 620 is configured to obtain a single-frequency response spectrum corresponding to the target power divider according to the splitting ratio information of the target power divider. In one embodiment, the acquisition module 620 may be used to perform the operation S220 described above, which will not be elaborated here.

[0132] The third acquisition module 630 is configured to obtain a broadband target response spectrum according to the Gaussian distribution of the single-frequency response spectrum and the first preset noise information, where the target response spectrum represents the output spectrum for designing the target power divider. In one embodiment, the second acquisition module 630 may be used to perform the operation S230 described above, which will not be elaborated here.

[0133] The output module 640 is configured to use the first neural network model to output a target topology structure vector corresponding to the target response spectrum according to the target response spectrum. In one embodiment, the output module 640 may be used to perform the operation S240 described above, which will not be elaborated here.

[0134] The design module 650 is configured to design the target power divider according to the target topology structure vector. In one embodiment, the design module 650 may be used to perform the operation S250 described above, which will not be elaborated here.

[0135] According to an embodiment of the present disclosure, the output module 640 includes a first acquisition sub-module and an output sub-module. Among them, the first acquisition sub-module is configured to obtain a spectrum feature corresponding to the target response spectrum according to the target response spectrum; the output sub-module is configured to input the spectrum feature into the first neural network model and output a target topology structure vector corresponding to the spectrum feature.

[0136] According to an embodiment of the present disclosure, the above power divider design device further includes a third acquisition module and a determination module. Among them, the third acquisition module is configured to obtain a test response spectrum corresponding to the target topology structure vector according to the target topology structure vector; the determination module is configured to determine the similarity between the test response spectrum and the target response spectrum, where the similarity is used to verify the accuracy of the target topology structure vector.

[0137] According to an embodiment of the present disclosure, the power splitter design device further includes an adjustment module, which is configured to adjust the parameters in the first neural network model and re-determine the target topology structure vector when the similarity is less than a preset similarity threshold.

[0138] According to an embodiment of the present disclosure, the output sub-module includes an acquisition unit, which is configured to perform feature transformation on the spectral features by using the first neural network model and the rectified linear unit function to obtain the target topology structure vector, where the rectified linear unit function represents the activation function of the first neural network model.

[0139] According to an embodiment of the present disclosure, the design module 650 may further be configured to characterize the materials used in the target power splitter by the values of each dimension in the target topology structure vector, and the materials are selected from silicon and silicon dioxide.

[0140] According to an embodiment of the present disclosure, the output module 640 further includes a generation sub-module, a second acquisition sub-module, a third acquisition sub-module, a fourth acquisition sub-module, a fifth acquisition sub-module, and a first input sub-module. Among them, the generation sub-module is configured to randomly generate a plurality of topology structure vectors; the second acquisition sub-module is configured to use the finite-difference time-domain method to obtain a plurality of sample single-frequency response spectra according to the plurality of topology structure vectors; the third acquisition sub-module is configured to obtain a plurality of broadband response spectra according to the Gaussian distribution of the plurality of sample single-frequency response spectra and the second preset noise information; the fourth acquisition sub-module is configured to obtain a plurality of response spectrum samples according to the plurality of broadband response spectra; the fifth acquisition sub-module is configured to obtain the topology structure vectors corresponding to the plurality of response spectrum samples according to the plurality of response spectrum samples; the first input sub-module is configured to input the plurality of response spectrum samples into the neural network model to be trained to obtain the trained first neural network model and a plurality of topology structure vectors.

[0141] According to an embodiment of the present disclosure, the output module 640 further includes a sixth acquisition sub-module and a second input sub-module. Among them, the sixth acquisition sub-module is configured to obtain a plurality of vector samples according to the plurality of topology structure vectors; the second input sub-module is configured to input the plurality of vector samples into the neural network model to be trained to obtain the trained second neural network model.

[0142] According to an embodiment of the present disclosure, any multiple of the first acquisition module 610, the second acquisition module 620, the third acquisition module 630, the output module 640, and the design module 650 may be combined and implemented in one module, or any one of them may be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules may be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the first acquisition module 610, the second acquisition module 620, the third acquisition module 630, the output module 640, and the design module 650 may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on substrate, a system on package, an application specific integrated circuit (ASIC), or any other reasonable manner that can integrate or package circuits, etc., implemented by hardware or firmware, or implemented in any one of the three implementation manners of software, hardware, and firmware, or in an appropriate combination of any several of them. Alternatively, at least one of the first acquisition module 610, the second acquisition module 620, the third acquisition module 630, the output module 640, and the design module 650 may be at least partially implemented as a computer program module, and when the computer program module is run, it can execute corresponding functions.

[0143] Figure 7 A block diagram of an electronic device suitable for implementing a power divider design method according to an embodiment of the present disclosure is schematically shown.

[0144] As Figure 7 shown, the electronic device 700 according to an embodiment of the present disclosure includes a processor 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage section 708 into a random access memory (RAM) 703. The processor 701 may include, for example, a general microprocessor (such as a CPU), an instruction set processor, and / or a related chipset, and / or a dedicated microprocessor (such as an application specific integrated circuit (ASIC)), etc. The processor 701 may also include on-board memory for caching purposes. The processor 701 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0145] In the RAM 703, various programs and data required for the operation of the electronic device 700 are stored. The processor 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. The processor 701 performs various operations of the method flow according to the embodiments of the present disclosure by executing the programs in the ROM 702 and / or the RAM 703. It should be noted that the programs may also be stored in one or more memories other than the ROM 702 and the RAM 703. The processor 701 may also perform various operations of the method flow according to the embodiments of the present disclosure by executing the programs stored in the one or more memories.

[0146] According to an embodiment of the present disclosure, the electronic device 700 may further include an input / output (I / O) interface 705, and the input / output (I / O) interface 705 is also connected to the bus 704. The electronic device 700 may further include one or more of the following components connected to the I / O interface 705: an input portion 706 including a keyboard, a mouse, etc.; an output portion 707 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage portion 708 including a hard disk, etc.; and a communication portion 709 including a network interface card such as a LAN card, a modem, etc. The communication portion 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the I / O interface 705 as needed. A removable medium 711, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is mounted on the drive 710 as needed so that a computer program read from it can be installed into the storage portion 708 as needed.

[0147] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist separately without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the one or more programs are executed, the method according to the embodiments of the present disclosure is implemented.

[0148] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, it may include but is not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, device, or apparatus. For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the above-described ROM 702 and / or RAM 703 and / or one or more memories other than ROM 702 and RAM 703.

[0149] An embodiment of the present disclosure further includes a computer program product, which includes a computer program that contains program code for executing the method shown in the flowchart. When the computer program product runs in a computer system, the program code is used to enable the computer system to implement the power splitter design method provided by the embodiment of the present disclosure.

[0150] When the computer program is executed by the processor 701, it executes the above functions defined in the system / apparatus of the embodiment of the present disclosure. According to an embodiment of the present disclosure, the above-described systems, apparatuses, modules, units, etc. can be implemented by computer program modules.

[0151] In one embodiment, the computer program can rely on tangible storage media such as optical storage devices and magnetic storage devices. In another embodiment, the computer program can also be transmitted and distributed in the form of a signal on a network medium, and is downloaded and installed through the communication part 709, and / or installed from the removable medium 711. The program code contained in the computer program can be transmitted by any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0152] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 709, and / or installed from the removable medium 711. When the computer program is executed by the processor 701, it executes the above functions defined in the system of the embodiment of the present disclosure. According to an embodiment of the present disclosure, the above-described systems, devices, apparatuses, modules, units, etc. can be implemented by computer program modules.

[0153] According to embodiments of the present disclosure, program code for executing the computer programs provided by the embodiments of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. The programming languages include, but are not limited to, programming languages such as Java, C++, Python, the "C" language, or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device can be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., by connecting through the Internet using an Internet service provider).

[0154] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram can represent a module, a program segment, or a part of code that contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks can occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks shown can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and combinations of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0155] Those skilled in the art can understand that the features recited in the various embodiments and / or claims of the present disclosure can be combined or / and combined in various ways, even if such combinations or combinations are not explicitly recited in the present disclosure. In particular, without departing from the spirit and teachings of the present disclosure, the features recited in the various embodiments and / or claims of the present disclosure can be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of the present disclosure.

[0156] The embodiments of the present disclosure have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although the embodiments have been described separately above, this does not mean that the measures in each embodiment cannot be used advantageously in combination. The scope of the present disclosure is defined by the appended claims and their equivalents. Without departing from the scope of the present disclosure, those skilled in the art can make various substitutions and modifications, and all such substitutions and modifications should fall within the scope of the present disclosure.

Claims

1. A power splitter design method, including: In response to a design request for a target power splitter, obtaining the splitting ratio information of the target power splitter; According to the splitting ratio information of the target power splitter, obtaining a single-frequency response spectrum corresponding to the target power splitter; According to the Gaussian distribution of the single-frequency response spectrum and first preset noise information, obtaining a broadband target response spectrum, where the target response spectrum characterizes the output spectrum for designing the target power splitter; Using a first neural network model, according to the target response spectrum, outputting a target topology structure vector corresponding to the target response spectrum; According to the target topology structure vector, designing the target power splitter.

2. The method according to claim 1, wherein, The step of using the first neural network model to output a target topology structure vector corresponding to the target response spectrum according to the target response spectrum includes: According to the target response spectrum, obtaining a spectrum feature corresponding to the target response spectrum; Inputting the spectrum feature into the first neural network model and outputting the target topology structure vector corresponding to the spectrum feature.

3. The method according to claim 1, further including: According to the target topology structure vector, obtaining a test response spectrum corresponding to the target topology structure vector through simulation; Determining the similarity between the test response spectrum and the target response spectrum, where the similarity is used to verify the accuracy of the target topology structure vector.

4. The method according to claim 3, further including: In the case where the similarity is less than a preset similarity threshold, adjusting the parameters in the first neural network model and re-determining the target topology structure vector.

5. The method according to claim 2, wherein, The step of inputting the spectrum feature into the first neural network model and outputting the target topology structure vector corresponding to the spectrum feature includes: Using the first neural network model and a rectified linear unit function to perform feature transformation on the spectrum feature to obtain the target topology structure vector, where the rectified linear unit function characterizes the activation function of the first neural network model.

6. The method according to claim 1, wherein, The value of each dimension in the target topology structure vector characterizes the material used in the target power splitter, and the material is selected from silicon and silicon dioxide.

7. The method according to claim 1, wherein, The training process of the first neural network model includes: Randomly generating a plurality of topology structure vectors; Using the finite-difference time-domain method, obtaining a plurality of sample single-frequency response spectra according to the plurality of topology structure vectors; According to the Gaussian distribution of the plurality of sample single-frequency response spectra and second preset noise information, obtaining a plurality of broadband response spectra; According to the plurality of broadband response spectra, obtaining a plurality of response spectrum samples; inputting the plurality of response spectrum samples into the neural network model to be trained to obtain the trained first neural network model.

8. The method according to claim 7, further including: According to the plurality of topology structure vectors, obtaining a plurality of vector samples; Input the multiple vector samples into the neural network model to be trained, and obtain a trained second neural network model.

9. A power splitter design device comprising: a first acquisition module, configured to acquire the splitting ratio information of the target power splitter in response to a design request for the target power splitter; a second acquisition module, configured to obtain a single-frequency response spectrum corresponding to the target power splitter according to the splitting ratio information of the target power splitter; a third acquisition module, configured to obtain a broadband target response spectrum according to the Gaussian distribution of the single-frequency response spectrum and first preset noise information, where the target response spectrum represents an output spectrum for designing the target power splitter; an output module, configured to use a first neural network model to output a target topology vector corresponding to the target response spectrum according to the target response spectrum; a design module, configured to design the target power splitter according to the target topology vector.

Images