Dielectric constant prediction method and system, network training method and system

Abstract

The present invention discloses a dielectric constant prediction method and system, as well as a network training method and system. The method comprises: obtaining material characteristics of a material to be tested; constructing multiple descriptors based on the material characteristics of the material to be tested; screening the multiple descriptors using a genetic algorithm to determine M primary descriptors, and determining descriptors other than the primary descriptors as other descriptors; inputting the primary descriptors and other descriptors into a preset dielectric constant prediction model, and outputting dielectric constant prediction data. The present invention uses machine learning technology to efficiently and accurately predict the dielectric constant of microwave high-frequency materials. Compared with traditional neural network models and descriptor selection methods, the method significantly improves prediction accuracy, model efficiency, and generalization ability, and solves the problem of high-dimensional descriptors not matching small data sets.

Description

Technical Field

[0001] The present application relates to the fields of materials science and artificial intelligence technology, and more specifically, to a dielectric constant prediction method and system, and a network training method and system. Background Art

[0002] The dielectric constant is an important physical quantity that describes the relative ability of a dielectric to store electrostatic energy in an electric field. Its research is of vital importance in fields such as energy, electronics, communications, and new material design. For example, in microwave high-frequency devices, low dielectric loss materials have higher signal transmission efficiency, while in energy storage materials, high dielectric constant materials can significantly increase energy density. However, despite the important position of the dielectric constant in materials science, the process of experimentally measuring the dielectric constant is often long and costly, and is greatly affected by equipment, environment, etc. In addition, existing theoretical calculation methods (such as density functional theory, DFT) also face many challenges in accurately predicting the dielectric constant.

[0003] Currently, the study of dielectric constants relies primarily on two methods: experimental measurement and theoretical calculation. Experimental methods can provide highly accurate dielectric constant data, but are limited by numerous factors, such as sample preparation, test environment, and equipment accuracy, resulting in low repeatability and predictability of experimental results. Furthermore, for microwave and high-frequency materials, experimental measurements often struggle to obtain accurate dielectric constants. Theoretical calculation methods (such as simulation studies based on Debye dielectric theory) predict dielectric constants using analytical formulas, but their results rely on multiple assumptions and approximations. This significantly increases the error in theoretical calculations, particularly when the material has strong intermolecular interactions or a complex lattice structure.

[0004] In recent years, with the rapid development of machine learning and artificial intelligence technologies, data-driven models have been widely used in the field of materials science. By learning patterns from large amounts of experimental or computational data, machine learning techniques can efficiently construct a mapping relationship between dielectric constants and material descriptors. These methods can not only significantly reduce computational costs, but also achieve high-precision predictions on limited data sets. In particular, deep learning models (such as neural networks) have gradually become an important tool for studying material properties due to their powerful fitting capabilities and ability to capture nonlinear characteristics. However, existing machine learning models still face the following challenges when dealing with dielectric constant problems:

[0005] Data scarcity: Experimental data in the field of materials science is limited, especially the dielectric constant data of microwave high-frequency materials. This makes it difficult for existing models to fully learn the data characteristics.

[0006] The contradiction between high-dimensional descriptors and small datasets: Constructing a material property prediction model requires a large number of descriptors, which may include elemental properties, structural parameters, etc. However, high-dimensional descriptors often lead to model overfitting and reduce generalization ability.

[0007] Single model structure: Existing neural network structures are unable to effectively process complex high-dimensional data, resulting in insufficient prediction accuracy.

[0008] Therefore, developing an efficient and accurate dielectric constant prediction tool can overcome the problems of data scarcity and high dimensionality while improving the generalization ability of the model, which is of great significance for accelerating the discovery and optimization of new materials. Summary of the Invention

[0009] In light of the above issues, the present invention aims to provide a dielectric constant prediction method and system, as well as a network training method and system, that utilize machine learning techniques to efficiently and accurately predict the dielectric constant of microwave high-frequency materials. Compared to traditional neural network models and descriptor selection methods, this method significantly improves prediction accuracy, model efficiency, and generalization capabilities, resolving the problem of high-dimensional descriptors being incompatible with small datasets.

[0010] By adopting the statistical ensemble hierarchical neural network (HNN) architecture and multi-level sub-model decomposition and integration, the information in small data sets is fully mined, significantly improving the generalization ability of the model.

[0011] Using a genetic algorithm (GA), 909 basic descriptors were screened and optimized, ultimately extracting the six most important descriptors for dielectric constant prediction: d-block metal presence, atomic weight, p-block electron number, polarizability, Wigner-Seitz electron cloud density, and electronegativity. This dimensionality reduction method, based on the order of descriptor importance, significantly reduced model complexity while retaining key information.

[0012] By introducing a hierarchical neural network architecture, the complex descriptor space is decomposed into multiple sub-models, and the prediction accuracy and generalization ability are significantly improved through layer-by-layer optimization and integration.

[0013] A first aspect of the present invention provides a dielectric constant prediction method, comprising:

[0014] Obtaining material characteristics of the material to be tested;

[0015] constructing a plurality of descriptors according to the material characteristics of the material to be tested;

[0016] Screening the plurality of descriptors using a genetic algorithm to determine M primary descriptors, and determining descriptors other than the primary descriptors as other descriptors; the primary descriptors include whether it is a d-block metal, atomic weight, number of p-block electrons, polarizability, electron cloud density, and electronegativity;

[0017] The main descriptor and the other descriptors are input into a preset dielectric constant prediction model, and dielectric constant prediction data is output.

[0018] In this solution, the multiple descriptors are constructed according to the material characteristics of the material to be tested, including:

[0019] Calculate the configurational entropy of a material:

[0020] ;

[0021] Where, ΔS con is the configuration entropy of the material, k B is the Boltzmann constant, T is the Kelvin temperature, c i is the proportion of element i in all elements;

[0022] Determine the electronic structure properties by calculating the ratio of the s, p, d, and f valence electrons to the sum of the s, p, d, and f valence electrons of the elements present in the material, respectively;

[0023] Determine the properties of ionic compounds based on the electronegativity of the material;

[0024] Through a variety of preset statistical methods, the various elemental properties of the material are statistically analyzed to determine the statistical properties of the elemental properties of the material;

[0025] Configurational entropy, electronic structure properties, ionic compound properties and elemental property statistics of the material are determined as descriptors.

[0026] In this solution, the genetic algorithm is used to screen the multiple descriptors to determine M main descriptors, including:

[0027] Step 1: Create individuals randomly or based on some prior knowledge of the problem to build an initial population;

[0028] Step 2: Evaluate the fitness of each individual in each group using pre-set evaluation criteria; the pre-set evaluation criteria include the coefficient of determination R 2 , mean square error MSE and root mean square error RMSE;

[0029] Step 3, select individuals from various populations to create a new population based on their fitness;

[0030] Step 4: Select individuals from the new population and perform genetic operations to generate new individuals; the genetic operations include selection, mutation, and crossover;

[0031] Step 5, replace the old population with the new population;

[0032] Repeat steps 2-5 until the evaluation criteria converge, and use the converged descriptor dimensions and high determination coefficient as R 2 The descriptor corresponding to the population is determined as the main descriptor.

[0033] This plan also includes:

[0034] The preset dielectric constant prediction model is a hierarchical neural network structure composed of multiple sub-networks with the same structure and arranged in a hierarchical manner. Each sub-network includes m input nodes and 1 output node. The output data of the n-th layer sub-network is the input data of the n+1-th layer sub-network; M main descriptors and mM other descriptors are input into each sub-network in the first layer respectively, and the other descriptors input into each sub-network are determined by random selection.

[0035] A second aspect of the present invention provides a method for training a network for predicting the dielectric constant of a material, the method comprising:

[0036] Obtain dielectric constant data of sample materials;

[0037] performing data cleaning on the dielectric constant data of the sample material to obtain a sample data set;

[0038] constructing a plurality of descriptors based on material characteristics of sample materials in the sample data set;

[0039] Screening the plurality of descriptors using a genetic algorithm to determine M primary descriptors, and determining descriptors other than the primary descriptors as other descriptors; the primary descriptors include whether it is a d-block metal, atomic weight, number of p-block electrons, polarizability, electron cloud density, and electronegativity;

[0040] The main descriptor and the other descriptors are trained through a hierarchical neural network to establish a preset dielectric constant prediction model.

[0041] In this solution, the step of constructing multiple descriptors based on the material characteristics of the sample materials in the sample data set includes:

[0042] Calculate the configurational entropy of a material:

[0043] ;

[0044] Where, ΔS con is the configuration entropy of the material, k B is the Boltzmann constant, T is the Kelvin temperature, c i is the proportion of element i in all elements;

[0045] Determine the electronic structure properties by calculating the ratio of the s, p, d, and f valence electrons to the sum of the s, p, d, and f valence electrons of the elements present in the material, respectively;

[0046] Determine the properties of ionic compounds based on the electronegativity of the material;

[0047] Through a variety of preset statistical methods, the various elemental properties of the material are statistically analyzed to determine the statistical properties of the elemental properties of the material;

[0048] Configurational entropy, electronic structure properties, ionic compound properties and elemental property statistics of the material are determined as descriptors.

[0049] In this solution, the genetic algorithm is used to screen the multiple descriptors to determine M main descriptors, including:

[0050] Step 1: Create individuals randomly or based on some prior knowledge of the problem to build an initial population;

[0051] Step 2: Evaluate the fitness of each individual in each group using pre-set evaluation criteria; the pre-set evaluation criteria include the coefficient of determination R 2 , mean square error MSE and root mean square error RMSE;

[0052] Step 3, select individuals from various populations to create a new population based on their fitness;

[0053] Step 4: Select individuals from the new population and perform genetic operations to generate new individuals; the genetic operations include selection, mutation, and crossover;

[0054] Step 5, replace the old population with the new population;

[0055] Repeat steps 2-5 until the evaluation criteria converge, and use the converged descriptor dimensions and high determination coefficient as R 2 The descriptor corresponding to the population is determined as the main descriptor.

[0056] This plan also includes:

[0057] The preset dielectric constant prediction model is a hierarchical neural network structure composed of multiple sub-networks with the same structure and arranged in a hierarchical manner. Each sub-network includes m input nodes and 1 output node. The output data of the n-th layer sub-network is the input data of the n+1-th layer sub-network; M main descriptors and mM other descriptors are input into each sub-network in the first layer respectively, and the other descriptors input into each sub-network are determined by random selection.

[0058] A third aspect of the present invention provides a dielectric constant prediction system configured to predict the dielectric constant of a material, comprising:

[0059] at least one storage medium storing at least one instruction set; and

[0060] at least one processor, in communication with the at least one storage medium;

[0061] When the prediction system is running, the at least one processor reads the at least one instruction set and executes any one of the dielectric constant prediction methods according to instructions of the at least one instruction set.

[0062] A fourth aspect of the present invention provides a training system configured to train a prediction network for predicting a dielectric constant of a material, comprising:

[0063] at least one storage medium storing at least one instruction set; and

[0064] at least one processor, in communication with the at least one storage medium;

[0065] Wherein, when the training system is running, the at least one processor reads the at least one instruction set and executes any one of the network training methods according to the instructions of the at least one instruction set.

[0066] The present invention discloses a dielectric constant prediction method and system, as well as a network training method and system. The method comprises: obtaining material characteristics of a material to be tested; constructing multiple descriptors based on the material characteristics of the material to be tested; screening the multiple descriptors using a genetic algorithm to determine M primary descriptors, and determining descriptors other than the primary descriptors as other descriptors; inputting the primary descriptors and other descriptors into a preset dielectric constant prediction model, and outputting dielectric constant prediction data. The present invention uses machine learning technology to efficiently and accurately predict the dielectric constant of microwave high-frequency materials. Compared with traditional neural network models and descriptor selection methods, the method significantly improves prediction accuracy, model efficiency, and generalization ability, and solves the problem of high-dimensional descriptors not matching small data sets. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A flow chart of a dielectric constant prediction method provided by the present invention is shown;

[0068] Figure 2 A flow chart of a network training method provided by the present invention is shown;

[0069] Figure 3 A schematic diagram showing the material characteristics provided by the present invention;

[0070] Figure 4 A flowchart showing the operation of the iterative descriptor of the hierarchical neural network model provided by the present invention is shown;

[0071] Figure 5 A schematic diagram showing the prediction effect of the dielectric constant prediction model built using a traditional neural network provided by the present invention;

[0072] Figure 6 A schematic diagram showing the prediction effect of the dielectric constant prediction model built by the hierarchical neural network provided by the present invention. DETAILED DESCRIPTION

[0073] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0074] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0075] Figure 1 A flow chart of a dielectric constant prediction method provided by the present invention is shown.

[0076] like Figure 1 As shown, the present invention discloses a dielectric constant prediction method, comprising:

[0077] S102, obtaining material characteristics of the material to be tested;

[0078] S104, constructing multiple descriptors according to the material characteristics of the material to be tested;

[0079] S106, screening the multiple descriptors using a genetic algorithm to determine M primary descriptors, and determining descriptors other than the primary descriptors as other descriptors; the primary descriptors include whether it is a d-block metal, atomic weight, number of p-block electrons, polarizability, electron cloud density, and electronegativity;

[0080] S108 , inputting the main descriptor and other descriptors into a preset dielectric constant prediction model, and outputting dielectric constant prediction data.

[0081] According to an embodiment of the present invention, Figure 3As shown, there are 53 material characteristics of the material. The material characteristics of the material to be tested can be obtained through one or more detection methods preset by the system (such as X-ray detection, table lookup method, etc.). The system analyzes the material characteristics of the material to be tested by selecting the preset descriptor calculation method, and determines 909 descriptors including 1 configuration entropy, 4 electronic structure properties, 3 ionic compound properties and 901 elemental property statistical properties. The 909 descriptors are screened by genetic algorithm to determine the most important main descriptors for dielectric constant prediction. There are 6 main descriptors, namely whether it is a d-block metal, atomic weight, p-block electron number, polarizability, electron cloud density and electronegativity. The 6 main descriptors have clear physical meanings and are closely related to the physical nature of the dielectric constant:

[0082] Whether it is a d-block metal: represents the charge separation ability of the material. D-block metal oxides usually have a higher dielectric constant.

[0083] Atomic weight and number of electrons in the p-region: closely related to the chemical composition and electronic structure of the material, affecting the polarizability and dielectric constant.

[0084] Polarizability and electron cloud density: directly reflect the material's ability to respond under the action of an electric field and are key determinants of the dielectric constant.

[0085] Electronegativity: Characterizes the ability of an atom to attract electrons and affects the electrostatic energy storage capacity of the material.

[0086] The main descriptors and other descriptors are input into the preset dielectric constant prediction model. All the main descriptors are input into each sub-network of the first level of the preset dielectric constant prediction model respectively, and other descriptors are randomly input through the remaining idle input nodes. The output data of each level is used as the input data of the next level. Through hierarchical analysis, combined with the dielectric constant data and material characteristics of the sample materials in the database, the corresponding dielectric constant prediction data is finally output.

[0087] According to an embodiment of the present invention, multiple descriptors are constructed based on the material characteristics of the material to be tested, including:

[0088] Calculate the configurational entropy of a material:

[0089] ;

[0090] Where, ΔS con is the configuration entropy of the material, k B is the Boltzmann constant, T is the Kelvin temperature, c i is the proportion of element i in all elements;

[0091] Determine the electronic structure properties by calculating the ratio of the s, p, d, and f valence electrons to the sum of the s, p, d, and f valence electrons of the elements present in the material, respectively;

[0092] Determine the properties of ionic compounds based on the electronegativity of the material;

[0093] Through a variety of preset statistical methods, the various elemental properties of the material are statistically analyzed to determine the statistical properties of the elemental properties of the material;

[0094] Configurational entropy, electronic structure properties, ionic compound properties, and statistical properties of elemental properties of materials are identified as descriptors.

[0095] It should be noted that there are 909 descriptors in total, including 1 configuration entropy, 4 electronic structure properties, 3 ionic compound properties and 901 element property statistical properties. First, determine the element i contained in the material to be tested, and the proportion c of element i in all elements. i The proportion of element i among all elements is input into the system's preset configurational entropy calculation formula to determine the material's configurational entropy. Configurational entropy represents the spatial arrangement of atoms in a material. Including configurational entropy in the descriptor can characterize key properties of the material being tested and improve the accuracy of the predicted dielectric constant of the agent being tested.

[0096] The electronic structure properties are calculated as:

[0097] ;

[0098] Among them, F x represents the electronic structure property corresponding to the xth valence electron, c i is the proportion of the i-th element in all elements, E x Indicates the number of x-th valence electrons (s, p, d or f valence electrons) in the i-th element in the sample to be tested, E n represents the total number of s, p, d, and f valence electrons of the i-th element.

[0099] The properties of ionic compounds are calculated using the formula:

[0100] ;

[0101] Where I is the property of ionic compounds, e is the natural constant, c i is the proportion of the i-th element in all elements, f i is the electronegativity corresponding to the i-th element, is the maximum value or average value of the electronegativity of each element. When is the maximum value among the electronegativity corresponding to each element, determine the first ionic compound property; when When I is the average value of the electronegativity corresponding to each element, the second ionic compound property is determined. At the same time, the third ionic compound property is determined by the second ionic compound property. When I>1.7, the third ionic compound property is 1; when I≤1.7, the third ionic compound property is 0.

[0102] The preset statistical methods include 17 statistical methods such as calculating the minimum value, maximum value and range of element properties; the (weighted) minimum value, maximum value and range of element properties after weighted sorting; the (weighted) minimum value, maximum value, range and mean of the absolute percentage of element properties. Figure 3 The properties of 53 elements shown in the statistic are statistically analyzed to determine the statistical properties of 901 elements.

[0103] According to an embodiment of the present invention, a plurality of descriptors are screened by a genetic algorithm to determine M main descriptors, including:

[0104] Step 1: Create individuals randomly or based on some prior knowledge of the problem to build an initial population;

[0105] Step 2: Evaluate the fitness of each individual in each group using pre-set evaluation criteria; the pre-set evaluation criteria include the coefficient of determination R 2 , mean square error MSE and root mean square error RMSE;

[0106] Step 3, select individuals from various populations to create a new population based on their fitness;

[0107] Step 4: Select individuals from the new population and perform genetic operations to generate new individuals; genetic operations include selection, mutation, and crossover;

[0108] Step 5, replace the old population with the new population;

[0109] Repeat steps 2-5 until the evaluation criteria converge, and use the converged descriptor dimensions and high determination coefficient as R 2 The descriptor corresponding to the population is determined as the main descriptor.

[0110] It should be noted that Genetic Algorithm (GA) is classified as a family of evolutionary algorithms. Its theory draws on the idea of ​​"survival of the fittest" in the biological world and generates high-quality solutions to optimization and search problems through operators (selection, mutation, and crossover) inspired by biology. The schematic diagram of the genetic algorithm is as follows Figure 4 The basic steps of the genetic algorithm are as follows:

[0111] Initialization: Individuals (including one or more descriptors) are created randomly based on a preset number of individuals in the system or based on some prior knowledge of the problem to build the initial population. Each individual represents a combination of these descriptors and is represented by a binary code, where 1 indicates that the descriptor at the corresponding position is selected, and 0 indicates the opposite.

[0112] Evaluation: Evaluate the fitness of each individual in the population. Fitness represents the ability of an individual to solve a problem and is usually determined by a fitness function that quantifies the goal or cost of the solution. For neural networks, the coefficient of determination can be used. R 2 , mean square error (MSE) and root mean square error (RMSE) are used as evaluation criteria for the descriptor combination represented by the individual.

[0113] Selection: Create a new population by selecting individuals from the population based on their fitness. Individuals with higher fitness are more likely to be selected, mimicking the concept of "survival of the fittest."

[0114] Reproduction: Producing new individuals (offspring) by performing genetic operations such as crossover (recombination) and mutation on selected individuals. Crossover involves combining the genetic material of two individuals to create a new solution, while mutation introduces random changes to maintain diversity in the population.

[0115] Replacement: The replacement of an old population with a new population that includes both selected individuals and newly produced offspring.

[0116] Termination: Repeat the above steps until the evaluation criteria converge. In practice, the descriptor dimension and determination coefficient are used. R 2 As the evolution direction of the population, we finally get the converged descriptor dimension and high determination coefficient R 2 The population, that is, determining the main descriptors.

[0117] According to an embodiment of the present invention, the further embodiment includes:

[0118] The preset dielectric constant prediction model is a hierarchical neural network structure composed of multiple sub-networks with the same structure and arranged in a hierarchical manner. Each sub-network includes m input nodes and 1 output node. The output data of the n-th sub-network is the input data of the n+1-th sub-network. M main descriptors and mM other descriptors are input into each sub-network in the first layer respectively, and the other descriptors input into each sub-network are determined by random selection.

[0119] It should be noted that other descriptors for each sub-network can be input using , where 909-M represents other descriptors, i.e., descriptors other than the primary descriptor; and mM represents the number of other descriptors that can be input into the subnetwork in addition to the primary descriptor. mM other descriptors are randomly (and unorderedly) selected from the 909-M other descriptors and input into the subnetwork along with the M primary descriptors. The other descriptors input into each subnetwork are randomly selected, so they are not identical in each subnetwork.

[0120] In a hierarchical neural network, the number of subnetworks included in each layer of the network gradually decreases. That is, the number of subnetworks included in the current layer is less than that of the previous layer, and the highest layer contains only one subnetwork. The output data of each subnetwork in the current layer serves as the input data for each subnetwork in the next layer. The network structure of each subnetwork is the same, and those skilled in the art can modify the network structure of any subnetwork in any layer according to actual needs.

[0121] Figure 2 The flowchart of a network training method provided by the present invention is shown.

[0122] like Figure 2 As shown, the second aspect of the present invention provides a method for training a network for predicting the dielectric constant of a material, the method comprising:

[0123] S202, obtaining dielectric constant data of the sample material;

[0124] S204, performing data cleaning on the dielectric constant data of the sample material to obtain a sample data set;

[0125] S206, constructing a plurality of descriptors according to material characteristics of the sample materials in the sample data set;

[0126] S208, screening the multiple descriptors using a genetic algorithm to determine M primary descriptors, and determining descriptors other than the primary descriptors as other descriptors; the primary descriptors include whether it is a d-block metal, atomic weight, number of p-block electrons, polarizability, electron cloud density, and electronegativity;

[0127] S210, training the main descriptor and other descriptors through a hierarchical neural network to establish a preset dielectric constant prediction model.

[0128] It should be noted that 1124 sample materials, mainly microwave high-frequency materials, were collected through literature and other channels, and the dielectric constant data and material characteristics of each sample material were determined. Among them, there are 53 types of material characteristics, as follows: Figure 3 The training process of the preset dielectric constant prediction model is as follows Figure 4As shown, first, the dielectric constant data of the obtained sample materials are cleaned. The data cleaning steps include deleting duplicate values, supplementing missing values, and processing outliers, so as to obtain a high-quality sample data set (containing 1124 sample data in total). Each sample data in the sample data set is analyzed in turn to construct the corresponding descriptor. The main descriptors are screened out by genetic algorithm, and the hierarchical neural network (HNN) architecture based on convolutional neural network (CNN) is adopted. 880 data are selected from the sample data set as training sets and 224 data as test sets. By hierarchical input of different combinations of descriptors (including all main descriptors and some other descriptors), hierarchical training is performed to gradually improve the model prediction ability and establish a preset dielectric constant prediction model. At the same time, 880 data are selected as training sets and 224 data are selected as test sets to train the traditional convolutional neural network model for comparative experiments. The traditional convolutional neural network model uses 145 descriptors to build a dielectric constant prediction model, and its training results are shown as follows. Figure 5 As shown, the training set R 2 =0.990, test set R 2 =0.789. The hierarchical neural network used in the present invention uses 909 descriptors to build a dielectric constant prediction model, and its training results are as follows: Figure 6 As shown, the training set R 2 =0.994, test set R 2 =0.903.

[0129] Results show that the HNN model's prediction accuracy on the test set increased by 14.5%, achieving a high level of prediction accuracy. Furthermore, by optimizing the model layer by layer through a hierarchical neural network, the complex nonlinear relationship between descriptors and dielectric constants can be fully exploited, resulting in more reliable predictions and lower errors.

[0130] In addition, the present invention uses a genetic algorithm (GA) to filter descriptors and optimizes the original 909 descriptors into 6 main descriptors, significantly reducing the model complexity while retaining information that is crucial for dielectric constant prediction. The optimized model exhibits the following advantages:

[0131] Improved computational efficiency: After descriptor screening, model parameters are significantly reduced, and training time is reduced by an average of 30%-40%. The genetic algorithm reduces redundant descriptors and avoids the burden of high-dimensional input on model training.

[0132] The physical meaning of the key descriptors is clear: The six key descriptors selected by the genetic algorithm, including d-block metal presence, atomic weight, p-block electron number, polarizability, electron cloud density, and electronegativity, are all closely related to the physical nature of the dielectric constant. These descriptors demonstrate a high contribution to model training, ensuring the accuracy and reliability of model predictions.

[0133] In addition, the hierarchical neural network model of the present invention significantly enhances the generalization ability of the model on small data sets through layered input and statistical ensemble optimization, effectively solving the problem of data scarcity:

[0134] Solving the data scarcity problem: Using only 1,124 data samples (880 training sets and 224 test sets), the HNN model achieved an R² value of 0.903 on the test set, fully verifying its excellent performance on small datasets.

[0135] Avoiding overfitting: By using layered input descriptors and optimizing layer by layer, the model effectively avoids the overfitting problem caused by the mismatch between high-dimensional descriptors and small datasets. The ensemble statistics method integrates multiple training results into the output of each layer of the model, further improving the robustness and generalization ability of the model.

[0136] A third aspect of the present invention provides a dielectric constant prediction system configured to predict the dielectric constant of a material, comprising:

[0137] at least one storage medium storing at least one instruction set; and

[0138] at least one processor, in communication with at least one storage medium,

[0139] When the prediction system is running, at least one processor reads at least one instruction set and executes any one of the dielectric constant prediction methods according to the instructions of the at least one instruction set.

[0140] A fourth aspect of the present invention provides a training system configured to train a prediction network for predicting a dielectric constant of a material, comprising:

[0141] at least one storage medium storing at least one instruction set; and

[0142] at least one processor, in communication with at least one storage medium,

[0143] When the training system is running, at least one processor reads at least one instruction set and executes any one of the network training methods according to the instructions of the at least one instruction set.

[0144] The information involved in this application (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals (including but not limited to signals transmitted between user terminals and other devices, etc.) are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions. For example, the "material characteristics of the material to be tested" and "dielectric constant data of the sample material" involved in this disclosure are all obtained with full authorization.

[0145] The present invention discloses a dielectric constant prediction method and system, as well as a network training method and system. The method comprises: obtaining material characteristics of a material to be tested; constructing multiple descriptors based on the material characteristics of the material to be tested; screening the multiple descriptors using a genetic algorithm to determine M primary descriptors, and determining descriptors other than the primary descriptors as other descriptors; inputting the primary descriptors and other descriptors into a preset dielectric constant prediction model, and outputting dielectric constant prediction data. The present invention uses machine learning technology to efficiently and accurately predict the dielectric constant of microwave high-frequency materials. Compared with traditional neural network models and descriptor selection methods, the method significantly improves prediction accuracy, model efficiency, and generalization ability, and solves the problem of high-dimensional descriptors not matching small data sets.

[0146] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0147] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0148] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0149] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0150] Alternatively, if the integrated units described above are implemented as software modules and sold or used as standalone products, they can also be stored on a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product, stored on a storage medium, includes instructions for enabling a computer device (such as a personal computer, server, or network device) to execute all or part of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as removable storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A dielectric constant prediction method, characterized in that: include: Obtaining material characteristics of the material to be tested; constructing a plurality of descriptors according to the material characteristics of the material to be tested; Screening the plurality of descriptors using a genetic algorithm to determine M primary descriptors, and determining descriptors other than the primary descriptors as other descriptors; the primary descriptors include whether it is a d-block metal, atomic weight, number of p-block electrons, polarizability, electron cloud density, and electronegativity; Inputting the main descriptor and the other descriptors into a preset dielectric constant prediction model, and outputting dielectric constant prediction data; The step of constructing a plurality of descriptors according to the material characteristics of the material to be tested comprises: Calculate the configurational entropy of a material: ； Where, ΔS con is the configuration entropy of the material, k B is the Boltzmann constant, T is the Kelvin temperature, c i is the proportion of element i in all elements; Determine the electronic structure properties by calculating the ratio of the s, p, d, and f valence electrons to the sum of the s, p, d, and f valence electrons of the elements present in the material, respectively; Determine the properties of ionic compounds based on the electronegativity of the material; Through a variety of preset statistical methods, the various elemental properties of the material are statistically analyzed to determine the statistical properties of the elemental properties of the material; Configurational entropy, electronic structure properties, ionic compound properties and elemental property statistics of the material are determined as descriptors.

2. The dielectric constant prediction method according to claim 1, wherein: The screening of the plurality of descriptors by a genetic algorithm to determine M main descriptors includes: Step 1: Create individuals randomly or based on some prior knowledge of the problem to build an initial population; Step 2: Evaluate the fitness of each individual in each group using a preset evaluation standard; the preset evaluation standard includes the coefficient of determination R 2 , mean square error MSE and root mean square error RMSE; Step 3, select individuals from various populations to create a new population based on their fitness; Step 4: Select individuals from the new population and perform genetic operations to generate new individuals; the genetic operations include selection, mutation, and crossover; Step 5, replace the old population with the new population; Repeat steps 2-5 until the evaluation criteria converge, and use the converged descriptor dimension and high determination coefficient R 2 The descriptor corresponding to the population is determined as the main descriptor.

3. The dielectric constant prediction method according to claim 1, wherein: Also includes: The preset dielectric constant prediction model is a hierarchical neural network structure composed of multiple sub-networks with the same structure and arranged in layers, each sub-network includes m input nodes and 1 output node, and the output data of the n-th layer sub-network is the input data of the n+1-th layer sub-network; M main descriptors and mM other descriptors are input to each sub-network in the first layer respectively, and the other descriptors input to each sub-network are determined by random selection.

4. A method for training a network for predicting the dielectric constant of a material, the method comprising: Obtain dielectric constant data of sample materials; performing data cleaning on the dielectric constant data of the sample material to obtain a sample data set; constructing a plurality of descriptors based on material characteristics of sample materials in the sample data set; Screening the plurality of descriptors using a genetic algorithm to determine M primary descriptors, and determining descriptors other than the primary descriptors as other descriptors; the primary descriptors include whether it is a d-block metal, atomic weight, number of p-block electrons, polarizability, electron cloud density, and electronegativity; Training the main descriptor and the other descriptors through a hierarchical neural network to establish a preset dielectric constant prediction model; The constructing of a plurality of descriptors according to the material characteristics of the sample materials in the sample data set comprises: Calculate the configurational entropy of a material: ； Where, ΔS con is the configuration entropy of the material, k B is the Boltzmann constant, T is the Kelvin temperature, c i is the proportion of element i in all elements; Determine the electronic structure properties by calculating the ratio of the s, p, d, and f valence electrons to the sum of the s, p, d, and f valence electrons of the elements present in the material, respectively; Determine the properties of ionic compounds based on the electronegativity of the material; Through a variety of preset statistical methods, the various elemental properties of the material are statistically analyzed to determine the statistical properties of the elemental properties of the material; Configurational entropy, electronic structure properties, ionic compound properties and elemental property statistics of the material are determined as descriptors.

5. The training method according to claim 4, characterized in that The screening of the plurality of descriptors by a genetic algorithm to determine M main descriptors includes: Step 1: Create individuals randomly or based on some prior knowledge of the problem to build an initial population; Step 2: Evaluate the fitness of each individual in each group using a preset evaluation standard; the preset evaluation standard includes the coefficient of determination R 2 , mean square error MSE and root mean square error RMSE; Step 3, select individuals from various populations to create a new population based on their fitness; Step 4: Select individuals from the new population and perform genetic operations to generate new individuals; the genetic operations include selection, mutation, and crossover; Step 5, replace the old population with the new population; Repeat steps 2-5 until the evaluation criteria converge, and use the converged descriptor dimension and high determination coefficient R 2 The descriptor corresponding to the population is determined as the main descriptor.

6. The training method according to claim 4, characterized in that Also includes: The preset dielectric constant prediction model is a hierarchical neural network structure composed of multiple sub-networks with the same structure and arranged in layers, each sub-network includes m input nodes and 1 output node, and the output data of the n-th layer sub-network is the input data of the n+1-th layer sub-network; M main descriptors and mM other descriptors are input to each sub-network in the first layer respectively, and the other descriptors input to each sub-network are determined by random selection.

7. A dielectric constant prediction system, characterized in that: Configured to predict the dielectric constant of materials, including: at least one storage medium storing at least one instruction set; and at least one processor, in communication with the at least one storage medium; Wherein, when the prediction system is running, the at least one processor reads the at least one instruction set and executes the method according to any one of claims 1 to 3 according to the instructions of the at least one instruction set.

8. A training system, characterized in that: A prediction network configured to train a prediction network for predicting the dielectric constant of a material, comprising: at least one storage medium storing at least one instruction set; and at least one processor, in communication with the at least one storage medium; Wherein, when the training system is running, the at least one processor reads the at least one instruction set and executes the method according to any one of claims 4 to 6 according to the instructions of the at least one instruction set.

Images