Bayesian optimization method, device and equipment based on weighted Euclidean distance matrix
By using weighted Euclidean distance matrix and Gaussian process model in Bayesian optimization method, the weights are dynamically updated to optimize decision variables, solving the problem of inefficient optimization caused by constant weights in the prior art, and achieving more efficient decision variable optimization.
Patent Information
- Application Number
- CN202411831630.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-12
- Publication Date
- 2025-04-22
AI Technical Summary
When optimizing decision variables, the weights are fixed, and the time required to iteratively adjust the value of the decision variable increases, reducing optimization efficiency.
The Bayesian optimization method based on the weighted Euclidean distance matrix is adopted, and the distance weighted matrix is determined by obtaining the predetermined value of the decision variable, and weighted calculation is performed to obtain the distance weighted matrix. Then, the Gaussian process model is applied to update the preset weights until the objective function relationship meets the set conditions, and the final functional relationship is obtained to optimize the decision variables.
By dynamically updating the weights, the accuracy and optimization efficiency of the objective function relationship are improved, and the iteration time is reduced to find the value that meets the set conditions.
Smart Images
Figure CN119862968A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of black box optimization problems, and in particular to a Bayesian optimization method, device and equipment based on a weighted Euclidean distance matrix. Background Art
[0002] Black box optimization problem refers to the optimization problem In (among which is the decision variable, For decision space, is the objective function to be optimized), and There is no mathematical expression shown in the figure. When a decision variable is given When the Value. Mixed space refers to the decision space There are many types of variables in the ML model: numerical inputs and categorical inputs. Numerical inputs have natural numerical relationships between their values, such as variables with a range of real numbers. Categorical inputs have no obvious order, size, or numerical relationship between their values. The black box optimization problem is to find a solution that makes The value can meet the set conditions The existing technology uses the following method to solve : First, give the decision variable Encode several values, then calculate the Euclidean distance between any two values to construct a distance matrix, then assign weights to the distance matrix based on the prior distribution, and finally use the Bayesian algorithm to iteratively adjust the decision variables based on the weighted distance matrix to find the value of The value can meet the set conditions value.
[0003] Since the weights are fixed in the above iterative process, and the weights determined based on the prior distribution are not necessarily suitable for any decision variables, using inaccurate and fixed weights will result in more time being spent on iteratively adjusting the decision variables. To find the value of The value can meet the set conditions value.
[0004] In summary, the existing technology reduces the efficiency of optimizing decision variables.
[0005] Therefore, the prior art still needs to be improved and enhanced. Summary of the invention
[0006] To solve the above technical problems, the present invention provides a Bayesian optimization method, device and equipment based on a weighted Euclidean distance matrix, which solves the problem that the prior art reduces the efficiency of optimizing decision variables.
[0007] To achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a Bayesian optimization method based on a weighted Euclidean distance matrix, which includes: Obtaining respective predetermined values of the decision variables, and determining, based on the respective predetermined values, a distance matrix for characterizing differences between the respective predetermined values; Obtaining preset weights corresponding to the distance matrix, performing weighted calculation on the preset weights and the distance matrix to obtain a distance weighted matrix; Applying a Gaussian process model to the distance weighted matrix to obtain an objective function relationship with the decision variable as an independent variable, and updating the preset weight based on the objective function relationship to update the objective function relationship until the objective function relationship satisfies a set condition to obtain a final function relationship; The decision variables are optimized according to the final functional relationship.
[0008] In one implementation, determining, based on the respective predetermined values, a distance matrix for characterizing the differences between the respective predetermined values includes: Using several distance calculation methods to calculate the distance value between any two of the predetermined values, the distance value is used to characterize the difference between any two of the predetermined values; According to the distance value corresponding to each distance calculation method, a distance sub-matrix corresponding to each distance calculation method is obtained, and each distance sub-matrix is used as a distance matrix.
[0009] In one implementation, the preset weights include preset weights of each of the distance sub-matrices, and weighted calculation is performed on the preset weights and the distance matrix to obtain a distance weighted matrix, including: The preset weights of the distance sub-matrices are weightedly calculated with each distance sub-matrix to obtain a distance weighted matrix.
[0010] In one implementation, updating the preset weight based on the objective function relational expression to update the objective function relational expression until the objective function relational expression satisfies a set condition to obtain a final function relational expression includes: evaluating the accuracy of the objective function relationship; Based on the precision, determining the posterior distribution of each of the preset weights; Based on the posterior distribution, updating the preset weight; Based on the updated preset weights, updating the distance weighted matrix; Based on the updated distance weighted matrix, the parameters of the objective function relationship are updated until the objective function relationship meets the set conditions, thereby obtaining a final function relationship.
[0011] In one implementation, optimizing the decision variables according to the final functional relationship includes: A Bayesian optimization algorithm is applied to the final functional relationship to optimize the decision variables.
[0012] In one implementation, the iterative optimization is terminated when the iteration time reaches a set time.
[0013] In one implementation, the decision variables include categorical variables and numerical variables; the categorical variable is a gate-level conversion operation required for chip design, the predetermined value is a sequential coding value of the gate-level conversion operation, and the objective function relationship is a gate-level circuit scale; or, the categorical variable is a pesticide, the predetermined value is a sequential coding value of the pesticide, and the objective function relationship is the number of crops infested by pests.
[0014] In a second aspect, an embodiment of the present invention further provides a Bayesian optimization device based on a weighted Euclidean distance matrix, wherein the device comprises the following components: A distance matrix calculation module, used to obtain each predetermined value of the decision variable, and based on each of the predetermined values, determine a distance matrix for characterizing the differences between each of the predetermined values; A distance weighted matrix calculation module, used to obtain preset weights corresponding to the distance matrix, and perform weighted calculation on the preset weights and the distance matrix to obtain a distance weighted matrix; A function relationship updating module, used for applying a Gaussian process model to the distance weighted matrix to obtain an objective function relationship with the decision variable as an independent variable, and updating the preset weight based on the objective function relationship to update the objective function relationship until the objective function relationship satisfies a set condition to obtain a final function relationship; An optimization module is used to optimize the decision variables according to the final functional relationship.
[0015] In the third aspect, an embodiment of the present invention also provides a terminal device, which includes a memory, a processor, and a Bayesian optimization program based on a weighted Euclidean distance matrix stored in the memory and executable on the processor. When the processor executes the Bayesian optimization program based on the weighted Euclidean distance matrix, the steps of the above-mentioned Bayesian optimization method based on the weighted Euclidean distance matrix are implemented.
[0016] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a Bayesian optimization program based on a weighted Euclidean distance matrix is stored. When the Bayesian optimization program based on the weighted Euclidean distance matrix is executed by a processor, the steps of the above-mentioned Bayesian optimization method based on the weighted Euclidean distance matrix are implemented.
[0017] Beneficial effect: The present invention first sets various predetermined values for the decision variables, then uses a distance matrix to characterize the differences between various predetermined values, and assigns preset weights to the distance matrix to obtain a distance weighted matrix, then applies a Gaussian process model to the distance weighted matrix, and updates the preset weights according to the objective function relationship output by the Gaussian process model, and updates the objective function relationship based on the updated preset weights until a final function relationship that meets the set conditions is obtained, and finally optimizes the decision variables based on the final function relationship with the decision variables as independent variables. From the above analysis, it can be seen that the present invention does not use fixed preset weights to optimize decision variables, but updates the preset weights according to the performance of the objective function relationship corresponding to the preset weights, so as to improve the accuracy of the preset weights after the update, and then the final function relationship can be determined as soon as possible according to the updated preset weights, thereby improving the efficiency of optimizing decision variables. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 It is the overall flow chart of the present invention; Figure 2 A flowchart of optimizing decision variables in an embodiment of the present invention; Figure 3 A structural diagram of a Bayesian optimization device based on a weighted Euclidean distance matrix provided by the present invention; Figure 4 This is a block diagram of the internal structure principle of the terminal device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0019] The following is a clear and complete description of the technical solution of the present invention in combination with the embodiments and the accompanying drawings. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0020] After research, it was found that the black box optimization problem refers to the optimization problem In (among which is the decision variable, For decision space, is the objective function to be optimized), and There is no mathematical expression shown in the figure. When a decision variable is given When the Value. Mixed space refers to the decision space There are many types of variables in the ML model: numerical variables and categorical variables. Numerical variables have natural numerical relationships between their values, such as variables with a range of real numbers. Categorical variables have no obvious order, size, or numerical relationship between their values. The black box optimization problem is to find a solution that makes The value can meet the set conditions The existing technology uses the following method to solve : First, give the decision variable Encode several values, then calculate the Euclidean distance between any two values to construct a distance matrix, then assign weights to the distance matrix based on the prior distribution, and finally use the Bayesian algorithm to iteratively adjust the decision variables based on the weighted distance matrix to find the value of The value can meet the set conditions value.
[0021] Since the weights are fixed in the above iterative process, and the weights determined based on the prior distribution are not necessarily suitable for any decision variables, using inaccurate and fixed weights will result in more time being spent on iteratively adjusting the decision variables. To find the value of The value can meet the set conditions values, thus reducing the efficiency of optimizing decision variables.
[0022] In order to solve the above technical problems, the present invention provides a Bayesian optimization method, device, and equipment based on a weighted Euclidean distance matrix, which solves the problem that the prior art reduces the efficiency of optimizing decision variables. In specific implementation, firstly, each predetermined value of the decision variable is obtained, and based on each predetermined value, a distance matrix used to characterize the differences between each predetermined value is determined; then, the preset weights corresponding to the distance matrix are obtained, and the preset weights and the distance matrix are weighted to obtain a distance weighted matrix; then, a Gaussian process model is applied to the distance weighted matrix to obtain an objective function relationship with the decision variable as an independent variable, and the preset weights are updated based on the objective function relationship to update the objective function relationship until the objective function relationship meets the set conditions to obtain the final function relationship; finally, the decision variable is optimized according to the final function relationship.
[0023] For example, the application scenario is farmland pest control, and the decision variables are is the type of pesticide, and the objective function is the relationship between the pesticides applied to the farmland and the number of crops infested by pests in the farmland .
[0024] The types of pesticides include pesticide 1, pesticide 2, pesticide 3, pesticide 4, and pesticide 5. They are coded in sequence as pesticide 1, pesticide 2, pesticide 3, pesticide 4, and pesticide 5, respectively. That is, pesticide 1 may be coded as 1 or 2 or 3 or 4 or 5 (1, 2, 3, 4, 5 are the decision variables of pesticide type Each predetermined value of pesticide 1, pesticide 2, pesticide 3, pesticide 4, and pesticide 5 may be encoded as 1, 2, 3, 4, or 5, pesticide 3 may be encoded as 1, 2, 3, 4, or 5, and pesticide 5 may be encoded as 1, 2, 3, 4, or 5. Any sequence of pesticide 1, pesticide 2, pesticide 3, pesticide 4, and pesticide 5 corresponds to a distance matrix. For example, the first group of codes of pesticide 1, pesticide 2, pesticide 3, pesticide 4, and pesticide 5 are 2, 4, 5, 1, and 3, respectively, then the distance matrix :
[0025] The value of any element is calculated by the number corresponding to the position of the element among 2, 4, 5, 1, and 3, for example The 2 in the first row and second column is the absolute value of the difference between 2 and 4. The 3 in the first column of the third row is the absolute value of the difference between 2 and 5. The values of other elements are calculated in the same way.
[0026] There are 120 coding combinations for Pesticide 1, Pesticide 2, Pesticide 3, Pesticide 4, and Pesticide 5, that is, there are 120 distance matrices, which are , , ,..., , each distance matrix sets a preset weight according to the prior distribution, and performs weighted calculation to obtain the distance weighted matrix :
[0027] In the formula, For the Distance submatrix The preset weight of .
[0028] Distance weighted matrix Apply the Gaussian process model to obtain the objective function relationship , evaluation objective function relationship If the accuracy is low, all preset weights are updated. After the preset weights are updated, the distance weighted matrix is updated. , and then update the objective function relationship , update the objective function relation That is, update the parameters in the function and evaluate the updated objective function relationship again If the accuracy is still low, continue to iterate and update all the preset weights until the objective function relation is The accuracy reaches the set condition (the set condition is that the accuracy is greater than the threshold), and the final functional relationship is obtained . Then according to Optimizing decision variables , optimize the decision variables That is, which one or which insecticides can be selected from insecticide 1, insecticide 2, insecticide 3, insecticide 4, and insecticide 5 to make The value of is the smallest, that is, the number of crops damaged by pests in the farmland is the least. In this embodiment, the minimum value is the optimal solution.
[0029] In this embodiment, there can be more than one decision variable, namely the type of pesticide. , it can also include three decision variables: pesticide type, crop type, and farmland geographical location. The three decision variables correspond to three distance weighted matrices.
[0030] The Bayesian optimization method based on the weighted Euclidean distance matrix of this embodiment can be applied to a terminal device, which can be a terminal product with a data processing function, such as a computer. Figure 1 As shown in , the Bayesian optimization method based on the weighted Euclidean distance matrix specifically includes the following steps: S100, obtaining each predetermined value of a decision variable, and determining, based on each of the predetermined values, a distance matrix for characterizing differences between each of the predetermined values; S200, obtaining preset weights corresponding to the distance matrix, performing weighted calculation on the preset weights and the distance matrix, and obtaining a distance weighted matrix; S300, applying a Gaussian process model to the distance weighted matrix to obtain an objective function relationship with the decision variable as an independent variable, and updating the preset weight based on the objective function relationship to update the objective function relationship until the objective function relationship satisfies a set condition to obtain a final function relationship; Among them, satisfying the set conditions includes obtaining an optimal solution for the objective function relation.
[0031] S400: Optimize the decision variables according to the final functional relationship.
[0032] The Gaussian process model is a trained model. The trained Gaussian process model is used to generate an objective function relationship. The preset weights required for calculating the distance weighted matrix are reversely updated according to the objective function relationship until the objective function relationship is updated to obtain the final function relationship. The optimized decision variables are obtained using the final function relationship. In the process of optimizing the decision variables, the Gaussian process model can be trained in turn according to the optimized decision variables to update the Gaussian process model and the weights required for weighted calculation, and then the function relationship is updated using the new Gaussian process model and the new weights.
[0033] Steps S100 to S400 can be applied to chip automation design, and the categorical variables in the decision variables are The gate-level conversion operation required for chip design, the predetermined value is the sequential encoding value of the gate-level conversion operation, and the objective function relation is the gate-level circuit scale.
[0034] That is, during the logic synthesis of chip design, the abstract digital circuit description is converted into a gate-level circuit representation by selecting reasonable conversion operations. Represents a transformation operation, and its value can include, for example, {rewrite, refactor, resub, balance, fraig}, etc. Each value is a transformation operation. It represents the gate-level circuit scale after conversion, the smaller the better. In this embodiment, the smallest gate-level circuit scale is the optimal solution.
[0035] Embodiment 1, in this embodiment, step S100 includes the following specific steps: using several distance calculation methods to calculate the distance value between any two of the predetermined values, the distance value is used to characterize the difference between any two of the predetermined values; according to the distance value corresponding to each of the distance calculation methods, obtaining the distance sub-matrix corresponding to each of the distance calculation methods, and using each of the distance sub-matrices as the distance matrix.
[0036] In this embodiment, when the decision variable is a numerical variable, for example, the decision variable is sunlight, the predetermined value of the decision variable is the intensity of sunlight in each time period, and the objective function relationship is the relationship between the sunlight intensity in each time period and the plant growth rate. , that is, the distance matrix includes multiple distance sub-matrices. When different sequential encodings are used to encode the decision variables, different distance sub-matrices will be obtained. Sequential encoding is the prior art.
[0037] When the decision variable is a categorical variable, the predetermined value of the decision variable is the sequential code of the decision variable, and the Euclidean distance between the variable values is calculated based on the numerical value of the sequential code, that is, Each element in is the Euclidean distance between two codes.
[0038] Embodiment 2, based on embodiment 1, in this embodiment, the preset weights include the preset weights of each of the distance sub-matrices, and the preset weights and the distance sub-matrices are weighted to obtain a distance weighted matrix :
[0039] is the total number of distance sub-matrices. Figure 2 As shown, are preset weights set according to the prior distribution.
[0040] Embodiment 3, based on embodiment 2 or embodiment 1, step S300 of this embodiment includes the following specific steps S301 to S305: S301, evaluating the objective function relation accuracy.
[0041] S302: Determine the posterior distribution of each of the preset weights based on the precision.
[0042] S303: Update the preset weight based on the posterior distribution.
[0043] S304: Update the distance weighted matrix based on the updated preset weights.
[0044] S305, based on the updated distance weighted matrix, update the parameters of the objective function relationship until the objective function relationship meets the set conditions, thereby obtaining a final function relationship.
[0045] Steps S301 to S305 are to convert the objective function equation As the evaluation object, update the preset weight according to the evaluation results , to improve The accuracy of The accuracy of this embodiment is The accuracy is the standard, update , can quickly improve within a limited time accuracy.
[0046] In this embodiment, the preset weight is first determined based on the prior distribution (the prior distribution includes the semi-Cauchy distribution). , and then perform the first iteration update, which will be replaced by the preset weights Determined distance weighted matrix Embedded into the Gaussian process model, the Gaussian process model outputs the objective function relationship , evaluate the objective function relationship The Gaussian process model has some other model parameters, so the preset weights are updated according to the accuracy. You can also update the model parameters of the Gaussian process model. Based on the preset weights of the first update Perform the second iteration until The accuracy reaches the set conditions and is completed Updates, using the last updated The final functional relationship can be obtained .
[0047] In this embodiment, the semi-Cauchy distribution encourages most weights to be close to zero, thereby reducing the complexity of the model and the risk of overfitting. From the perspective of learning the spatial structure, this will make the distance weighted matrix finally learned , depends on a small number of bases with non-zero weights. These bases are more able to reflect the spatial structure of the variables represented in the data.
[0048] The method for updating the preset weights in this embodiment is the Hamiltonian Monte Carlo method.
[0049] Embodiment 4, based on embodiment 3 or embodiment 2 or embodiment 1, this embodiment applies a Bayesian optimization algorithm to the final functional relationship to optimize the decision variables.
[0050] When the decision variable is a categorical variable, this embodiment optimizes the decision variable, that is, selects one or several predetermined values from various predetermined values of the decision variable so that the final functional relationship meets the set conditions.
[0051] When the decision variable is a numerical variable, this embodiment optimizes the decision variable, that is, adjusts each predetermined value of the decision variable so that the calculation result of the final functional relationship based on the adjusted predetermined value meets the set condition.
[0052] Once the parameters of the relationship between the weights and the objective function are learned, the final function relationship is obtained. Bayesian optimization is then performed, and the expected improvement is selected as the criterion to iteratively select the decision variable values for each round of experiments.
[0053] In summary, when the black box problem is a minimization problem (the number of crops lost to pests and diseases is a minimization problem), the present invention can minimize the value under the same computing resources, and the present invention can obtain a better solution in the same time, thereby improving the efficiency of the algorithm.
[0054] In addition, since the present invention uses a data-driven approach to establish a distance weighted matrix, the accuracy of the distance weighted matrix is improved, thereby improving the final functional relationship. Accuracy.
[0055] This embodiment also provides a Bayesian optimization device based on a weighted Euclidean distance matrix, such as Figure 3 As shown, the device comprises the following components: The distance matrix calculation module 01 is used to obtain each predetermined value of the decision variable, and based on each of the predetermined values, determine a distance matrix for characterizing the differences between each of the predetermined values; The distance weighted matrix calculation module 02 is used to obtain the preset weights corresponding to the distance matrix, and perform weighted calculation on the preset weights and the distance matrix to obtain a distance weighted matrix; Functional relation updating module 03, used for applying Gaussian process model to the distance weighted matrix to obtain the objective functional relation with the decision variable as the independent variable, and updating the preset weight based on the objective functional relation to update the objective functional relation until the objective functional relation satisfies the set condition to obtain the final functional relation; The optimization module 04 is used to optimize the decision variables according to the final functional relationship.
[0056] Based on the above embodiments, the present invention further provides a terminal device, whose principle block diagram can be shown as follows: Figure 4 As shown. The terminal device includes a processor, a memory, a network interface, and a display screen connected through a system bus. Among them, the processor of the terminal device is used to provide computing and control capabilities. The memory of the terminal device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the terminal device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a Bayesian optimization method based on a weighted Euclidean distance matrix is implemented. The display screen of the terminal device can be a liquid crystal display screen or an electronic ink display screen.
[0057] Those skilled in the art will understand that Figure 4 The principle block diagram shown in the figure is only a block diagram of a partial structure related to the scheme of the present invention, and does not constitute a limitation on the terminal device to which the scheme of the present invention is applied. The specific terminal device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0058] In one embodiment, a terminal device is provided, the terminal device comprising a memory, a processor, and a Bayesian optimization program based on a weighted Euclidean distance matrix stored in the memory and executable on the processor, and when the processor executes the Bayesian optimization program based on the weighted Euclidean distance matrix, the following operation instructions are implemented: Obtaining respective predetermined values of the decision variables, and determining, based on the respective predetermined values, a distance matrix for characterizing differences between the respective predetermined values; Obtaining preset weights corresponding to the distance matrix, performing weighted calculation on the preset weights and the distance matrix to obtain a distance weighted matrix; Applying a Gaussian process model to the distance weighted matrix to obtain an objective function relationship with the decision variable as an independent variable, and updating the preset weight based on the objective function relationship to update the objective function relationship until the objective function relationship satisfies a set condition to obtain a final function relationship; The decision variables are optimized according to the final functional relationship.
[0059] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A Bayesian optimization method based on a weighted Euclidean distance matrix, characterized in that: include: Obtaining respective predetermined values of the decision variables, and determining, based on the respective predetermined values, a distance matrix for characterizing differences between the respective predetermined values; Obtaining preset weights corresponding to the distance matrix, performing weighted calculation on the preset weights and the distance matrix to obtain a distance weighted matrix; Applying a Gaussian process model to the distance weighted matrix to obtain an objective function relationship with the decision variable as an independent variable, and updating the preset weight based on the objective function relationship to update the objective function relationship until the objective function relationship satisfies a set condition to obtain a final function relationship; The decision variables are optimized according to the final functional relationship.
2. The Bayesian optimization method based on the weighted Euclidean distance matrix as claimed in claim 1, characterized in that: The determining, based on each of the predetermined values, a distance matrix for characterizing the differences between each of the predetermined values comprises: Using several distance calculation methods to calculate the distance value between any two of the predetermined values, the distance value is used to characterize the difference between any two of the predetermined values; According to the distance value corresponding to each distance calculation method, a distance sub-matrix corresponding to each distance calculation method is obtained, and each distance sub-matrix is used as a distance matrix.
3. The Bayesian optimization method based on the weighted Euclidean distance matrix as claimed in claim 2, characterized in that: The preset weights include the preset weights of each of the distance sub-matrices, and weighted calculation is performed on the preset weights and the distance matrix to obtain a distance weighted matrix, including: The preset weights of the distance sub-matrices are weightedly calculated with each distance sub-matrix to obtain a distance weighted matrix.
4. The Bayesian optimization method based on the weighted Euclidean distance matrix according to claim 1, characterized in that: Updating the preset weight based on the objective function relational expression to update the objective function relational expression until the objective function relational expression satisfies a set condition to obtain a final function relational expression, including: evaluating the accuracy of the objective function relationship; Based on the precision, determining the posterior distribution of each of the preset weights; Based on the posterior distribution, updating the preset weight; Based on the updated preset weights, updating the distance weighted matrix; Based on the updated distance weighted matrix, the parameters of the objective function relationship are updated until the objective function relationship meets the set conditions, thereby obtaining a final function relationship.
5. The Bayesian optimization method based on weighted Euclidean distance matrix according to claim 1, characterized in that: Optimizing the decision variables according to the final functional relationship includes: A Bayesian optimization algorithm is applied to the final functional relationship to optimize the decision variables.
6. The Bayesian optimization method based on weighted Euclidean distance matrix according to claim 5, characterized in that: The iterative optimization is terminated when the iteration time reaches the set time.
7. The Bayesian optimization method based on a weighted Euclidean distance matrix according to any one of claims 1 to 6, characterized in that: The decision variables include categorical variables and numerical variables; the categorical variable is a gate-level conversion operation required for chip design, the predetermined value is a sequential coding value of the gate-level conversion operation, and the objective function relationship is a gate-level circuit scale; or, the categorical variable is a pesticide, the predetermined value is a sequential coding value of the pesticide, and the objective function relationship is the number of crops infested by pests.
8. A Bayesian optimization device based on a weighted Euclidean distance matrix, characterized in that: The device comprises the following components: A distance matrix calculation module, used to obtain each predetermined value of the decision variable, and based on each of the predetermined values, determine a distance matrix for characterizing the differences between each of the predetermined values; A distance weighted matrix calculation module, used to obtain preset weights corresponding to the distance matrix, and perform weighted calculation on the preset weights and the distance matrix to obtain a distance weighted matrix; A function relationship updating module, used for applying a Gaussian process model to the distance weighted matrix to obtain an objective function relationship with the decision variable as an independent variable, and updating the preset weight based on the objective function relationship to update the objective function relationship until the objective function relationship satisfies a set condition to obtain a final function relationship; An optimization module is used to optimize the decision variables according to the final functional relationship.
9. A terminal device, characterized in that: The terminal device includes a memory, a processor, and a Bayesian optimization program based on a weighted Euclidean distance matrix stored in the memory and executable on the processor. When the processor executes the Bayesian optimization program based on the weighted Euclidean distance matrix, the steps of the Bayesian optimization method based on the weighted Euclidean distance matrix as described in any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a Bayesian optimization program based on a weighted Euclidean distance matrix. When the Bayesian optimization program based on a weighted Euclidean distance matrix is executed by a processor, the steps of the Bayesian optimization method based on a weighted Euclidean distance matrix as described in any one of claims 1 to 7 are implemented.