Integrated photon prediction method, device, equipment, medium and product

Abstract

The invention discloses an integrated photon prediction method, device and equipment, a medium and a product, and relates to the technical field of integrated photons. The method comprises the steps that after configuration data used for conducting simulation effect prediction on a target integrated photon object is obtained, sample data are collected from all historical simulation sample data which are abandoned and matched with the target object through a data collection algorithm, and a collection result is added into a sample data set; and then performing calibration verification modeling on the machine learning model based on the artificial neural network by applying the data set to obtain a simulation effect prediction model or a reverse structure prediction model, and finally applying the prediction model to obtain design parameters of the target integrated photon object and outputting the design parameters. In this way, all invalid simulation data generated in the design process can be fully utilized, simulation design time is saved for a user to achieve different functions under the same integrated photon structure, and practical application and popularization are facilitated.

Description

Technical Field

[0001] The present invention belongs to the field of integrated photonics technology, and particularly relates to an integrated photon prediction method, device, equipment, medium and product. Background Art

[0002] In the post-Moore era (i.e., the era after Moore's law fails), as integrated circuits approach the physical and theoretical limits, integrated photonics technology (Photonic Integrated Circuit, PIC, which is a technology system that integrates an optical system onto a single chip, similar to electronic integrated circuit technology; different from electronic integrated circuits that integrate electronic devices such as transistors, capacitors, and / or resistors, integrated photonics technology integrates various different optical devices or optoelectronic devices, such as lasers, electro-optic modulators, photodetectors, optical attenuators, optical multiplexers / demultiplexers, and optical amplifiers, etc.) has received increasing attention as an alternative solution. However, whether it is the design of integrated photon devices or the design of integrated optoelectronic system links, it requires a large amount of simulation time; at the same time, during the design process, the final design is not achieved in one step, and a large number of iterative simulations are required to design the final design structure that meets the requirements.

[0003] Currently, a large amount of invalid simulation data generated from the initial concept to the design result is discarded, resulting in problems of data waste and long design simulation time in existing integrated photon design solutions. Therefore, how to utilize this invalid simulation data to help designers save simulation time is an urgent research topic for those skilled in the art. Summary of the Invention

[0004] The purpose of the present invention is to provide an integrated photon prediction method, device, computer equipment, computer-readable storage medium, and computer program product to solve the problems of data waste and long design simulation time existing in existing integrated photon design solutions.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions:

[0006] In the first aspect, an integrated photon prediction method is provided, including:

[0007] Obtain configuration data for predicting the simulation effect or reverse structure prediction of a target integrated photon object, where the target integrated photon object includes an integrated photon device or an integrated optoelectronic system link to be designed, the configuration data contains design parameter configuration information of the target integrated photon object, and the design parameter configuration information contains multi-dimensional design parameters to be optimized;

[0008] Adopt a data acquisition algorithm to collect the historical simulation sample data from all the historical simulation sample data that has been discarded and matches the target integrated photon object, and add the acquisition result to the sample data set, where the historical simulation sample data includes the parameter values for historical simulation and the historical simulation effect index values of the multi-dimensional design parameters that are used as the model input items and model output items for each other;

[0009] Apply the sample data set to calibrate and verify the modeling of a machine learning model based on an artificial neural network, and obtain a simulation effect prediction model for outputting the corresponding simulation effect index value after inputting the parameter values of the multi-dimensional design parameters or a reverse structure prediction model for outputting the parameter values of the corresponding multi-dimensional design parameters after inputting the simulation effect index value;

[0010] Apply the simulation effect prediction model or the reverse structure prediction model to obtain and output the design parameters of the target integrated photon object.

[0011] Based on the above invention content, a new solution for integrated photon design based on historical invalid simulation data is provided, that is, after obtaining the configuration data for predicting the simulation effect of the target integrated photon object, first adopt a data acquisition algorithm to collect sample data from all the historical simulation sample data that has been discarded and matches the target object, and add the acquisition result to the sample data set, then apply the data set to calibrate and verify the modeling of a machine learning model based on an artificial neural network to obtain a simulation effect prediction model or a reverse structure prediction model, and finally apply the prediction model to obtain and output the design parameters of the target integrated photon object, so that all the invalid simulation data generated in the design process can be fully utilized, saving simulation design time for users to achieve different functions under the same integrated photon structure, and facilitating practical application and promotion.

[0012] In a possible design, when the design parameter configuration information further includes an objective function related to the simulation effect index of the target integrated photon object, applying the simulation effect prediction model to obtain the design parameters of the target integrated photon object includes:

[0013] Apply the simulation effect prediction model to optimize the multi-dimensional design parameters based on a swarm optimization algorithm to obtain the optimization search result of the multi-dimensional design parameters for optimizing the objective function;

[0014] Take the optimization search result of the multi-dimensional design parameters as the optimal design parameters of the target integrated photon object and import them into the integrated optoelectronic link simulation software for simulation to obtain the simulation effect index value corresponding to the optimal design parameters;

[0015] Determine whether the simulation effect index value corresponding to the optimal design parameter meets the preset expected simulation effect. If so, output the optimal design parameter; otherwise, use the optimal design parameter and the simulation effect index value corresponding to the optimal design parameter as a discarded historical simulation sample data.

[0016] In a possible design, use a data acquisition algorithm to collect the historical simulation sample data from all the discarded historical simulation sample data that match the target integrated photon object, and add the collection result to the sample data set, including:

[0017] Process and analyze the current sample data set to determine the current sample data gap direction. Among them, the sample data set contains several simulation sample data, and the simulation sample data contains the simulation parameter values and simulation effect index values of the multi-dimensional design parameters that are used as the model input items and model output items for each other;

[0018] Configure a data acquisition algorithm according to the sample data gap direction, and use the data acquisition algorithm to collect the historical simulation sample data from all the discarded historical simulation sample data that match the target integrated photon object. Among them, the historical simulation sample data contains the historical simulation parameter values and historical simulation effect index values of the multi-dimensional design parameters that are used as the model input items and model output items for each other;

[0019] Judge whether the collected historical simulation sample data matches the sample data gap direction. If so, add the collected historical simulation sample data to the sample data set; otherwise, do not add the collected historical simulation sample data to the sample data set.

[0020] In a possible design, configure a data acquisition algorithm according to the sample data gap direction, and use the data acquisition algorithm to collect the historical simulation sample data from all the discarded historical simulation sample data that match the target integrated photon object, including the following steps S221 - S229:

[0021] S221. According to the sample data gap direction, initialize the initial position corresponding to the initial parameter value of the multi-dimensional design parameter at the peripheral position of the data gap, and then execute step S222. Among them, the initial position and the peripheral position of the data gap are respectively located in the multi-dimensional design parameter space constructed based on the multi-dimensional design parameter, and the peripheral position of the data gap is determined according to the sample data gap direction;

[0022] S222. According to the initial parameter values of the multi-dimensional design parameters, use the direct binary search algorithm to search for the first historical simulation sample data whose model input item matches the initial parameter value from all the historical simulation sample data that has been discarded and matches the target integrated photon object, and then execute step S223, where the historical simulation sample data includes the historical simulation parameter values and historical simulation effect index values of the multi-dimensional design parameters used as the model input item and the model output item respectively;

[0023] S223. Initialize the parameter serial number variable i to 1, and then execute step S224;

[0024] S224. Based on the current parameter values of the multi-dimensional design parameters, change the parameter value of the i-th design parameter in the multi-dimensional design parameters to obtain the new parameter values of the multi-dimensional design parameters, and then execute step S225;

[0025] S225. According to the new parameter values of the multi-dimensional design parameters, use the direct binary search algorithm to search for the second historical simulation sample data whose model input item matches the new parameter value from all the historical simulation sample data, and then execute step S226;

[0026] S226. Compare the historical simulation effect index values used as the model output item in the second historical simulation sample data and the first historical simulation sample data. If the historical simulation effect index value in the second historical simulation sample data is better than the historical simulation effect index value in the first historical simulation sample data, execute step S227, otherwise execute step S228;

[0027] S227. Update the first historical simulation sample data to the second historical simulation sample data, and if the parameter serial number variable i is less than the total number of dimensions of the multi-dimensional design parameters, increment the parameter serial number variable i by 1, and then return to execute step S224, otherwise execute step S229;

[0028] S228. Restore the parameter value of the i-th design parameter to the value before the change, and if the parameter serial number variable i is less than the total number of dimensions of the multi-dimensional design parameters, increment the parameter serial number variable i by 1, and then return to execute step S224, otherwise execute step S229;

[0029] S229. Use the first historical simulation sample data as the acquisition result and end this acquisition.

[0030] In a possible design, process and analyze the current sample data set to determine the current sample data gap direction, including:

[0031] According to the domain of definition of the simulation effect index value of the target integrated photon object, perform statistical analysis on the current sample data set to determine the distribution interval of the simulation effect index value in the sample data set within the domain of definition;

[0032] Exclude the distribution interval in the domain of definition to obtain a distribution missing interval in the domain of definition;

[0033] Determine the current sample data gap direction according to the distribution missing interval.

[0034] In a possible design, the artificial neural network adopts a residual network, where the number of layers of the residual network is 6 to 10 layers.

[0035] In a second aspect, an integrated photon prediction device is provided, including a configuration data acquisition unit, a sample data collection unit, a calibration and verification modeling unit, and a design parameter output unit;

[0036] The configuration data acquisition unit is used to acquire configuration data for predicting the simulation effect or reverse structure prediction of the target integrated photon object. Among them, the target integrated photon object includes an integrated photon device or an integrated optoelectronic system link to be designed, the configuration data includes the design parameter configuration information of the target integrated photon object, and the design parameter configuration information includes multi-dimensional design parameters to be optimized;

[0037] The sample data collection unit is communicatively connected to the configuration data acquisition unit, and is used to collect the historical simulation sample data from all the discarded historical simulation sample data that match the target integrated photon object by using a data collection algorithm, and add the collection result to the sample data set. Among them, the historical simulation sample data includes the historical simulation parameter values and historical simulation effect index values of the multi-dimensional design parameters used as the model input items and model output items;

[0038] The calibration and verification modeling unit is communicatively connected to the sample data collection unit, and is used to apply the sample data set to perform calibration and verification modeling on a machine learning model based on an artificial neural network, and obtain a simulation effect prediction model for outputting the corresponding simulation effect index value after inputting the parameter values of the multi-dimensional design parameters, or a reverse structure prediction model for outputting the parameter values of the corresponding multi-dimensional design parameters after inputting the simulation effect index value;

[0039] The design parameter output unit is communicatively connected to the calibration and verification modeling unit, and is used to apply the simulation effect prediction model or the reverse structure prediction model to obtain the design parameters of the target integrated photon object and output them.

[0040] In a third aspect, the present invention provides a computer device, comprising a memory, a processor and a transceiver that are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer programs and execute the integrated photon prediction method as described in the first aspect or any possible design in the first aspect.

[0041] In a fourth aspect, the present invention provides a computer-readable storage medium, on which instructions are stored. When the instructions run on a computer, they execute the integrated photon prediction method as described in the first aspect or any possible design in the first aspect.

[0042] In a fifth aspect, the present invention provides a computer program product, comprising a computer program or instructions, and when the computer program or the instructions are executed by a computer, they implement the integrated photon prediction method as described in the first aspect or any possible design in the first aspect.

[0043] Beneficial effects of the above solutions:

[0044] (1) The present invention creatively provides a new solution for integrated photon design based on historical invalid simulation data. That is, after obtaining the configuration data for predicting the simulation effect of a target integrated photon object, first use a data acquisition algorithm to collect sample data from all historical simulation sample data that have been discarded and match the target object, and add the collection result to the sample data set. Then, use the data set to perform calibration and verification modeling on a machine learning model based on an artificial neural network to obtain a simulation effect prediction model or a reverse structure prediction model. Finally, apply the prediction model to obtain the design parameters of the target integrated photon object and output them. In this way, all the invalid simulation data generated during the design process can be fully utilized, saving simulation design time for users to achieve different functions under the same integrated photon structure;

[0045] (2) By utilizing the invalid simulation data, it can also help users further explore the functions and application scenarios of devices under the same integrated design structure;

[0046] (3) With the growth of the data volume and the standardization of the data set, the boundaries between single devices (or systems) can also be gradually smoothed by the carefully designed data set, thereby realizing the simulation prediction of more devices and more systems, laying a foundation for supporting the simulation prediction from a single model to multiple topological structures, and facilitating practical application and promotion. Description of the Drawings

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0048] Figure 1 It is a schematic flowchart of the integrated photon prediction method provided by the embodiments of this application.

[0049] Figure 2 It is a schematic structural diagram of the integrated photon prediction device provided by the embodiments of this application.

[0050] Figure 3 It is a schematic structural diagram of the computer device provided by the embodiments of this application. Detailed implementation manners

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the present invention in combination with the accompanying drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other embodiments can be obtained based on these embodiments. It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation to the present invention.

[0052] It should be understood that although terms such as first and second may be used herein to describe various objects, these objects should not be limited by these terms. These terms are only used to distinguish one object from another. For example, the first object can be called the second object, and similarly, the second object can be called the first object, without departing from the scope of the exemplary embodiments of the present invention.

[0053] It should be understood that for the term "and / or" that may appear in this article, it is only a description of the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, B exists alone, or A and B exist simultaneously, etc.; another example, A, B, and / or C can represent any one of A, B, and C or any combination of them; for the term " / and" that may appear in this article, it is a description of another association object relationship, indicating that two relationships can exist. For example, A / and B can represent: A exists alone or A and B exist simultaneously, etc.; in addition, for the character " / " that may appear in this article, generally, it means that the associated objects before and after are an "or" relationship.

[0054] Embodiment

[0055] As Figure 1 shown, the integrated photon prediction method provided in the first aspect of this embodiment can be, but is not limited to, executed by a computer device with certain computing resources, such as a cloud server, a personal computer (PC, referring to a multi-purpose computer suitable for personal use in terms of size, price, and performance; desktop computers, laptops, small laptops, tablet computers, and ultrabooks, etc. all belong to personal computers), a smart phone, a personal digital assistant (PDA), or a wearable device, etc. As Figure 1 shown, the integrated photon prediction method can be, but is not limited to, including the following steps S1 to S4.

[0056] S1. Obtain configuration data for predicting the simulation effect or reverse structure of a target integrated photon object. Among them, the target integrated photon object includes, but is not limited to, an integrated photon device to be designed, an integrated optoelectronic system link, etc. The configuration data includes, but is not limited to, the design parameter configuration information of the target integrated photon object, and the design parameter configuration information includes, but is not limited to, multi-dimensional design parameters to be optimized.

[0057] In the step S1, the integrated photonic device may be, for example but not limited to, a dual-channel wavelength division multiplexing device of a photonic crystal-like structure, and the integrated optoelectronic system link may be, for example but not limited to, a four-channel wavelength division multiplexing link based on a microring. The configuration data may be a default value or a specific value input manually, so it can be obtained conventionally. The simulation effect prediction refers to predicting the simulation effect based on the designed structure, while the reverse structure prediction refers to inversely predicting the designed structure based on the simulation effect, which is completely opposite to the simulation effect prediction. In addition, the configuration data may further include, but is not limited to, an objective function related to the simulation effect index of the target integrated photonic object, algorithm parameter configuration information such as population size, mutation rate, number of iterations, classification operator, selection operator, crossover operator, and / or mutation operator, and option configuration information such as recovery setting options (for example, if recovery is selected, a recovery Json file needs to be selected) and / or recording setting options (for example, recording the recovery file, whether to configure it is optional; recording the recovery duration, with a default of 10 seconds; recording parameters, other parameters are recorded together with the optimization parameters; recording files, whether to configure it is optional); the design parameter configuration information may further include, but is not limited to, design parameter dimensions, parameter upper and lower bounds, parameter precision, equality constraints, and / or inequality constraints; and by way of example, if the target integrated photonic object is specifically a programmable photonic digital link composed of multiple programmable units, each programmable unit can operate electrically in the following three states: a through state (i.e., light input from the upper input port and output from the corresponding upper output port), a cross state (also a full coupling state, i.e., light input from the upper input port and output from the lower output port), and a partial coupling state (between the aforementioned two states, i.e., the proportion of light output from the output port changes with the coupling efficiency), then the multi-dimensional design parameters can be the state values / sum of coupling efficiencies of the multiple programmable units.

[0058] S2. Use a data acquisition algorithm to collect the historical simulation sample data from all the historical simulation sample data that have been discarded and match the target integrated photonic object, and add the acquisition result to the sample dataset, where the historical simulation sample data includes, but is not limited to, the historical simulation parameter values and historical simulation effect index values of the multi-dimensional design parameters used as mutual model input items and model output items.

[0059] In the step S2, all the historical simulation sample data that have been discarded and match the target integrated photon object refer to the following: If the target integrated photon object is a dual-channel wavelength division multiplexing device with a photonic crystal-like structure, the historical simulation sample data must be the sample data obtained from the historical design and simulation of this dual-channel wavelength division multiplexing device, rather than the sample data obtained from the historical design and simulation of other objects such as a four-channel wavelength division multiplexing link or a programmable photonic digital link. Considering different prediction directions, the historical simulation sample data will also be different as follows: If it is necessary to predict the simulation effect of the target integrated photon object, the historical simulation sample data includes but is not limited to the historical simulation parameter values of the multi-dimensional design parameters used as model input items and the historical simulation effect index values used as model output items; if it is necessary to perform reverse structure prediction on the target integrated photon object, the historical simulation sample data includes but is not limited to the historical simulation effect index values used as model input items and the historical simulation parameter values of the multi-dimensional design parameters used as model output items. Generally speaking, the data acquisition algorithm can adopt the direct binary search method (DBS Method, also known as the binary search algorithm, which is a method for efficiently searching for a specific element in an ordered array; its core idea is to locate the position of the target element by continuously narrowing the search range), but it is also considered that the data distribution for the training of a single model neural network at the current stage must meet certain distribution conditions, that is, the distribution for a certain result in all sample data must be relatively uniform, otherwise the trained model cannot achieve the basic prediction function. Therefore, preferably, the data acquisition algorithm is used to collect the historical simulation sample data from all the historical simulation sample data that have been discarded and match the target integrated photon object, and the acquisition result is added to the sample data set, including but not limited to the following steps S21 to S23.

[0060] S21. Process and analyze the current sample data set to determine the current sample data gap direction, where the sample data set includes but is not limited to several simulation sample data, and the simulation sample data includes but is not limited to the simulation parameter values and simulation effect index values of the multi-dimensional design parameters used as each other's model input items and model output items.

[0061] In the step S21, the current sample data set can be a data set imported manually or a data collection result obtained by using the data collection algorithm historically. The simulation effect index value is the simulation result corresponding to the simulation parameter value of the multi-dimensional design parameter, and examples but not limited to include transmission efficiency or specific parameter values of scattering parameters (which specifically include the following four parameters: S11 - reflection coefficient of port 1, indicating the reflection of the input signal at port 1 at port 1; S21 - forward transmission coefficient, indicating the signal gain from port 1 to port 2; S12 - reverse transmission coefficient, indicating the reverse transmission from port 2 to port 1; S22 - reflection coefficient of port 2, indicating the reflection of the input signal at port 2 at port 2). Specifically, processing and analyzing the current sample data set to determine the current sample data gap direction includes, but is not limited to, the following steps S211 to S213.

[0062] S211. According to the domain of definition of the simulation effect index value of the target integrated photon object, perform statistical analysis on the current sample data set to determine the distribution interval of the simulation effect index value in the sample data set within the domain of definition.

[0063] In the step S211, for example, if the simulation effect index value is a scattering parameter value, the domain of definition can be, for example, (0, 1), and then through conventional statistical analysis means, it can be determined, for example, that the distribution interval of the simulation effect index value in the domain of definition is (0, 0.5) and (0.6, 1).

[0064] S212. Exclude the distribution interval from the domain of definition to obtain the distribution missing interval in the domain of definition.

[0065] In the step S212, based on the specific example of the above step S211, the distribution missing interval can be obtained as (0.5, 0.6).

[0066] S213. Determine the current sample data gap direction according to the distribution missing interval.

[0067] S22. Configure the data collection algorithm according to the sample data gap direction, and use the data collection algorithm to collect the historical simulation sample data from all the historical simulation sample data that have been discarded and match the target integrated photon object, where the historical simulation sample data includes, but is not limited to, the historical simulation parameter values and historical simulation effect index values of the multi-dimensional design parameter used as the model input item and model output item for each other.

[0068] In step S22, specifically, configure a data acquisition algorithm according to the sample data gap direction, and use the data acquisition algorithm to collect the historical simulation sample data from all the historical simulation sample data that have been discarded and match the target integrated photon object, including the following steps S221 to S229.

[0069] S221. According to the sample data gap direction, initialize the initial position corresponding to the initial parameter value of the multi-dimensional design parameter at a position around the data gap, and then execute step S222. Wherein, the initial position and the position around the data gap are respectively located in the multi-dimensional design parameter space constructed based on the multi-dimensional design parameter, and the position around the data gap is determined according to the sample data gap direction.

[0070] In step S221, assume that the multi-dimensional design parameter space is a three-dimensional space, and the sample data gap direction is a direction pointing to a certain coordinate position in this three-dimensional space. Then, the multi-dimensional design parameter value closest to this certain coordinate position in the current sample data set can be used as the position around the data gap and the initial position.

[0071] S222. According to the initial parameter value of the multi-dimensional design parameter, use the direct binary search algorithm to search for the first historical simulation sample data whose model input item matches this initial parameter value from all the historical simulation sample data that have been discarded and match the target integrated photon object, and then execute step S223. Wherein, the historical simulation sample data includes the historical simulation parameter values and historical simulation effect index values of the multi-dimensional design parameter used as the model input item and the model output item for each other.

[0072] S223. Initialize the parameter sequence variable i to 1, and then execute step S224.

[0073] S224. Based on the current parameter value of the multi-dimensional design parameter, change the parameter value of the i-th design parameter in the multi-dimensional design parameter to obtain a new parameter value of the multi-dimensional design parameter, and then execute step S225.

[0074] S225. According to the new parameter value of the multi-dimensional design parameter, use the direct binary search algorithm to search for the second historical simulation sample data whose model input item matches this new parameter value from all the historical simulation sample data, and then execute step S226.

[0075] S226. Compare the historical simulation effect index values that are in the second historical simulation sample data and the first historical simulation sample data and are used as model output items. If the historical simulation effect index value in the second historical simulation sample data is better than the historical simulation effect index value in the first historical simulation sample data, then execute step S227; otherwise, execute step S228.

[0076] S227. Update the first historical simulation sample data to the second historical simulation sample data. And if the parameter sequence number variable i is less than the total number of dimensions of the multi-dimensional design parameters, then increment the parameter sequence number variable i by 1, and then return to execute step S224; otherwise, execute step S229.

[0077] S228. Restore the parameter value of the i-th design parameter to the value before the change. And if the parameter sequence number variable i is less than the total number of dimensions of the multi-dimensional design parameters, then increment the parameter sequence number variable i by 1, and then return to execute step S224; otherwise, execute step S229.

[0078] S229. Use the first historical simulation sample data as the acquisition result, and end this acquisition.

[0079] S23. Determine whether the collected historical simulation sample data matches the sample data gap direction. If so, add the collected historical simulation sample data to the sample data set; otherwise, do not add the collected historical simulation sample data to the sample data set.

[0080] In step S23, determining whether the collected historical simulation sample data matches the sample data gap direction may include, but is not limited to: according to the domain of definition of the simulation effect index value of the target integrated photon object, performing statistical analysis on the current sample data set and the collected historical simulation sample data again to determine the new distribution interval of the simulation effect index value in the domain of definition. Then, if the new distribution interval expands (which means the distribution missing interval shrinks at this time), it can be determined that the collected historical simulation sample data matches the sample data gap direction; otherwise, it is determined that the collected historical simulation sample data does not match the sample data gap direction.

[0081] S3. Apply the sample data set to calibrate and verify the modeling of a machine learning model based on an artificial neural network to obtain a simulation effect prediction model for outputting the corresponding simulation effect index value after inputting the parameter values of the multi-dimensional design parameters or a reverse structure prediction model for outputting the corresponding parameter values of the multi-dimensional design parameters after inputting the simulation effect index value.

[0082] In the step S3, the artificial neural network (ANN) is a mathematical model that simulates the structure and function of a biological neural network and is used for information processing and pattern recognition. The artificial neural network is composed of a large number of nodes (or "neurons") connected to each other. Each node represents a specific output function, called an activation function. The connection between every two nodes represents a weighted value, called a weight, which is equivalent to the memory of the artificial neural network. The output of the network depends on the connection method, weights, and activation function of the network. Therefore, based on a certain amount of the historical simulation sample data, through a conventional calibration and verification modeling process (specifically including the calibration process and verification process of the model, that is, first comparing the model simulation results with the measured data, and then adjusting the model parameters according to the comparison results to make the simulation results coincide with the actual situation), the simulation effect prediction model or the reverse structure prediction model can be trained. Specifically, the artificial neural network can but is not limited to using a residual network, where the number of layers of the residual network is 6 to 10 layers, and for example, it can be 8 layers. In addition, the sample data set can be specifically divided into a training set, a verification set, and a test set, etc. during the calibration and verification modeling process, and randomly selected according to a ratio of 3:1:1.

[0083] S4. Apply the simulation effect prediction model or the reverse structure prediction model to obtain the design parameters of the target integrated photon object and output them.

[0084] In the step S4, since the model input item of the reverse structure prediction model is the simulation effect index value, the target effect index value of the target integrated photon object can be directly input into the reverse structure prediction model, and then the design parameters of the target integrated photon object can be directly obtained and output. And since the model output item of the simulation effect prediction model is the simulation effect index value, it is necessary to continuously adjust the design parameters of the target integrated photon object and import them into the simulation effect prediction model until the output simulation effect index value reaches the target effect index value, and then the model input data at this time can be used as the design parameters of the target integrated photon object and output. For the latter, in order to automatically optimize and obtain the optimal design parameters of the target integrated photon object, preferably, when the design parameter configuration information also includes a target function related to the simulation effect index of the target integrated photon object, applying the simulation effect prediction model to obtain the design parameters of the target integrated photon object includes but is not limited to the following steps S41 to S43.

[0085] S41. Apply the simulation effect prediction model, optimize the multi-dimensional design parameters based on a swarm optimization algorithm, and obtain the optimization search result of the multi-dimensional design parameters for optimizing the objective function.

[0086] In step S41, the swarm optimization algorithm can be but is not limited to a particle swarm optimization algorithm, a genetic algorithm, a grey wolf algorithm, etc. The specific optimization process can be obtained by referring to the conventional derivation of existing algorithms.

[0087] S42. Use the optimization search result of the multi-dimensional design parameters as the optimal design parameters of the target integrated photon object and import them into an integrated optoelectronic link simulation software for simulation to obtain the simulation effect index value corresponding to the optimal design parameters.

[0088] In step S42, the integrated optoelectronic link simulation software is existing software, such as the existing pSim Plus software.

[0089] S43. Determine whether the simulation effect index value corresponding to the optimal design parameters meets the preset simulation effect expectation. If so, output the optimal design parameters; otherwise, use the optimal design parameters and the simulation effect index value corresponding to the optimal design parameters as a discarded historical simulation sample data.

[0090] In step S43, the simulation effect expectation can be specifically exemplified as exceeding the target effect index value of the target integrated photon object. In addition, after the optimal design parameters and the simulation effect index value corresponding to the optimal design parameters are discarded, they can be added to the sample dataset through the aforementioned step S2 for secondary utilization.

[0091] Thus, based on the integrated photon prediction method described in the foregoing steps S1 to S4, a new solution for integrated photon design based on historical invalid simulation data is provided. That is, after obtaining the configuration data for predicting the simulation effect of a target integrated photon object, first use a data acquisition algorithm to collect sample data from all historical simulation sample data that have been discarded and match the target object, and add the collection result to the sample dataset. Then, use the dataset to calibrate and verify the modeling of a machine learning model based on an artificial neural network to obtain a simulation effect prediction model or a reverse structure prediction model. Finally, apply the prediction model to obtain and output the design parameters of the target integrated photon object. In this way, all the invalid simulation data generated in the design process can be fully utilized, saving simulation design time for users to achieve different functions under the same integrated photon structure, and facilitating practical application and promotion.

[0092] Such as Figure 2As shown in the figure, the second aspect of this embodiment provides a virtual device for implementing the integrated photon prediction method described in the first aspect, including a configuration data acquisition unit, a sample data collection unit, a calibration verification modeling unit, and a design parameter output unit;

[0093] The configuration data acquisition unit is used to acquire configuration data for simulating effect prediction or reverse structure prediction of a target integrated photon object. Among them, the target integrated photon object includes an integrated photon device or an integrated optoelectronic system link to be designed, and the configuration data contains design parameter configuration information of the target integrated photon object, and the design parameter configuration information contains multi-dimensional design parameters to be optimized;

[0094] The sample data collection unit is communicatively connected to the configuration data acquisition unit, and is used to collect the historical simulation sample data from all historical simulation sample data that have been discarded and match the target integrated photon object by using a data collection algorithm, and add the collection result to the sample data set. Among them, the historical simulation sample data contains historical simulation parameter values and historical simulation effect index values of the multi-dimensional design parameters used as model input items and model output items for each other;

[0095] The calibration verification modeling unit is communicatively connected to the sample data collection unit, and is used to apply the sample data set to perform calibration verification modeling on a machine learning model based on an artificial neural network, and obtain a simulation effect prediction model for outputting corresponding simulation effect index values after inputting parameter values of the multi-dimensional design parameters, or a reverse structure prediction model for outputting parameter values of the corresponding multi-dimensional design parameters after inputting simulation effect index values;

[0096] The design parameter output unit is communicatively connected to the calibration verification modeling unit, and is used to apply the simulation effect prediction model or the reverse structure prediction model to obtain and output the design parameters of the target integrated photon object.

[0097] For the working process, working details and technical effects of the foregoing device provided in the second aspect of this embodiment, reference can be made to the integrated photon prediction method described in the first aspect, which will not be elaborated here.

[0098] As Figure 3As shown in the figure, the third aspect of this embodiment provides a computer device for executing the integrated photon prediction method described in the first aspect, including a memory, a processor, and a transceiver that are communicatively connected in sequence. Among them, the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer programs and execute the integrated photon prediction method described in the first aspect. Specifically, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, first input first output (FIFO), and / or first input last output (FILO), etc.; the processor may include, but is not limited to, a microprocessor of the STM32F105 series. In addition, the computer device may also include, but is not limited to, a power supply module, a display screen, and other necessary components.

[0099] For the working process, working details, and technical effects of the aforementioned computer device provided in the third aspect of this embodiment, reference may be made to the integrated photon prediction method described in the first aspect, which will not be elaborated here.

[0100] The fourth aspect of this embodiment provides a computer-readable storage medium storing instructions including the integrated photon prediction method described in the first aspect, that is, instructions are stored on the computer-readable storage medium, and when the instructions are run on a computer, the integrated photon prediction method described in the first aspect is executed. Among them, the computer-readable storage medium refers to a carrier for storing data, and may include, but is not limited to, computer-readable storage media such as floppy disks, optical discs, hard disks, flash memories, USB flash drives, and / or memory sticks. The computer may be a general-purpose computer, a dedicated computer, a computer network, or other programmable devices.

[0101] For the working process, working details, and technical effects of the aforementioned computer-readable storage medium provided in the fourth aspect of this embodiment, reference may be made to the integrated photon prediction method described in the first aspect, which will not be elaborated here.

[0102] The fifth aspect of this embodiment provides a computer program product including a computer program or instructions, and the computer program or the instructions, when executed by a computer, implement the integrated photon prediction method described in the first aspect. Among them, the computer may be a general-purpose computer, a dedicated computer, a computer network, or other programmable devices.

[0103] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An integrated photon prediction method, characterized in that: include: Acquiring configuration data for performing simulation effect prediction or reverse structure prediction on a target integrated photonic object, wherein the target integrated photonic object includes an integrated photonic device or an integrated optoelectronic system link to be designed, and the configuration data contains design parameter configuration information of the target integrated photonic object, and the design parameter configuration information contains multi-dimensional design parameters to be optimized; Using a data acquisition algorithm to acquire the historical simulation sample data from all historical simulation sample data that have been discarded and match the target integrated photon object, and adding the acquisition results to the sample data set, wherein the historical simulation sample data contains historical simulation parameter values ​​and historical simulation effect index values ​​of the multi-dimensional design parameters that are used as model input items and model output items for each other; Applying the sample data set to perform calibration and verification modeling on a machine learning model based on an artificial neural network, a simulation effect prediction model for outputting a corresponding simulation effect index value after inputting a parameter value of the multidimensional design parameter or a reverse structure prediction model for outputting a corresponding parameter value of the multidimensional design parameter after inputting a simulation effect index value is obtained; The simulation effect prediction model or the inverse structure prediction model is applied to obtain the design parameters of the target integrated photonic object and output them.

2. The integrated photon prediction method according to claim 1, characterized in that: When the design parameter configuration information further includes an objective function related to a simulation effect index of the target integrated photonic object, applying the simulation effect prediction model to obtain the design parameters of the target integrated photonic object includes: Applying the simulation effect prediction model, optimizing the multi-dimensional design parameters based on a group optimization algorithm, and obtaining the multi-dimensional design parameters and an optimization search result for optimizing the objective function; The optimization search result of the multi-dimensional design parameters is used as the optimal design parameters of the target integrated photonic object and is imported into integrated optoelectronic link simulation software for simulation to obtain simulation effect index values ​​corresponding to the optimal design parameters; Determine whether the simulation effect index value corresponding to the optimal design parameter meets the preset simulation effect expectation. If so, output the optimal design parameter; otherwise, treat the optimal design parameter and the simulation effect index value corresponding to the optimal design parameter as a discarded historical simulation sample data.

3. The integrated photon prediction method according to claim 1, characterized in that: The historical simulation sample data is collected from all the historical simulation sample data that have been discarded and match the target integrated photon object using a data collection algorithm, and the collection results are added to the sample data set, including: Processing and analyzing the current sample data set to determine the current sample data gap direction, wherein the sample data set includes a plurality of simulation sample data, and the simulation sample data includes simulation parameter values ​​and simulation effect index values ​​of the multidimensional design parameters that are mutually model input items and model output items; A data acquisition algorithm is configured according to the sample data gap direction, and the historical simulation sample data is collected from all historical simulation sample data that have been discarded and match the target integrated photon object using the data acquisition algorithm, wherein the historical simulation sample data contains historical simulation parameter values ​​and historical simulation effect index values ​​of the multi-dimensional design parameters that are used as model input items and model output items for each other; Determine whether the collected historical simulation sample data matches the sample data gap direction. If so, add the collected historical simulation sample data to the sample data set; otherwise, give up adding the collected historical simulation sample data to the sample data set.

4. The integrated photon prediction method according to claim 3, characterized in that: A data acquisition algorithm is configured according to the sample data gap direction, and the historical simulation sample data is collected from all historical simulation sample data that have been discarded and match the target integrated photon object using the data acquisition algorithm, including the following steps S221-S229: S221. According to the direction of the gap in the sample data, the initial position corresponding to the initial parameter value of the multidimensional design parameter is initialized at the position around the data gap, and then step S222 is performed, wherein the initial position and the position around the data gap are respectively located in the multidimensional design parameter space constructed based on the multidimensional design parameter, and the position around the data gap is determined according to the direction of the gap in the sample data; S222. According to the initial parameter value of the multi-dimensional design parameter, a direct binary search algorithm is used to search for the first historical simulation sample data whose model input item matches the initial parameter value from all historical simulation sample data that have been discarded and match the target integrated photon object, and then execute step S223, wherein the historical simulation sample data contains historical simulation parameter values ​​and historical simulation effect index values ​​of the multi-dimensional design parameters that are model input items and model output items for each other; S223. Initialize the parameter serial number variable i to 1, and then execute step S224; S224. Based on the current parameter value of the multi-dimensional design parameter, change the parameter value of the i-th design parameter in the multi-dimensional design parameter to obtain a new parameter value of the multi-dimensional design parameter, and then execute step S225; S225. According to the new parameter value of the multi-dimensional design parameter, the direct binary search algorithm is used to search for the second historical simulation sample data whose model input item matches the new parameter value from all the historical simulation sample data, and then step S226 is executed; S226. Compare the historical simulation effect index values ​​in the second historical simulation sample data and the first historical simulation sample data and used as the model output item. If the historical simulation effect index value in the second historical simulation sample data is better than the historical simulation effect index value in the first historical simulation sample data, execute step S227, otherwise execute step S228; S227. Update the first historical simulation sample data to the second historical simulation sample data, and if the parameter sequence number variable i is less than the total number of dimensions of the multi-dimensional design parameter, increment the parameter sequence number variable i by 1, and then return to step S224, otherwise execute step S229; S228. Restore the parameter value of the i-th design parameter to the value before the change, and if the parameter sequence number variable i is less than the total number of dimensions of the multi-dimensional design parameter, increment the parameter sequence number variable i by 1, and then return to step S224, otherwise execute step S229; S229. Take the first historical simulation sample data as the collection result and end this collection.

5. The integrated photon prediction method according to claim 3, characterized in that: Process and analyze the current sample data set to determine the direction of the current sample data gap, including: According to the definition domain of the simulation effect index value of the target integrated photon object, statistical analysis is performed on the current sample data set to determine the distribution interval of the simulation effect index value in the sample data set in the definition domain; Eliminating the distribution interval in the definition domain to obtain a distribution missing interval in the definition domain; According to the distribution missing interval, the direction of the current sample data gap is determined.

6. The integrated photon prediction method according to claim 1, characterized in that: The artificial neural network adopts a residual network, wherein the number of layers of the residual network is 6 to 10.

7. An integrated photon prediction device, characterized in that: It includes a configuration data acquisition unit, a sample data collection unit, a calibration verification modeling unit and a design parameter output unit; The configuration data acquisition unit is used to acquire configuration data for simulation effect prediction or reverse structure prediction of a target integrated photonic object, wherein the target integrated photonic object includes an integrated photonic device or an integrated optoelectronic system link to be designed, and the configuration data contains design parameter configuration information of the target integrated photonic object, and the design parameter configuration information contains multi-dimensional design parameters to be optimized; The sample data acquisition unit is communicatively connected to the configuration data acquisition unit, and is used to adopt a data acquisition algorithm to acquire the historical simulation sample data from all historical simulation sample data that have been discarded and match the target integrated photon object, and add the acquisition results to the sample data set, wherein the historical simulation sample data contains historical simulation parameter values ​​and historical simulation effect index values ​​of the multi-dimensional design parameters that are used as model input items and model output items for each other; The calibration verification modeling unit is communicatively connected to the sample data acquisition unit, and is used to apply the sample data set to perform calibration verification modeling on the machine learning model based on the artificial neural network, so as to obtain a simulation effect prediction model for outputting a corresponding simulation effect index value after inputting a parameter value of the multidimensional design parameter, or a reverse structure prediction model for outputting a corresponding parameter value of the multidimensional design parameter after inputting a simulation effect index value; The design parameter output unit is communicatively connected to the calibration verification modeling unit, and is used to apply the simulation effect prediction model or the inverse structure prediction model to obtain and output the design parameters of the target integrated photonic object.

8. A computer device, characterized in that: It includes a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the integrated photon prediction method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores instructions, and when the instructions are executed on a computer, the integrated photon prediction method as described in any one of claims 1 to 6 is executed.

10. A computer program product comprising a computer program or instructions, characterized in that When the computer program or the instruction is executed by a computer, the integrated photon prediction method according to any one of claims 1 to 6 is implemented.

Images