Parameter searching device, method, and program
The parameter search device and method address the challenge of balancing manufacturing efficiency and performance by constructing an objective function and using a model-based approach to select parameter values that meet both criteria, ensuring stable and high-quality product manufacturing.
Patent Information
- Application Number
- JP2023207772
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-12-08
- Publication Date
- 2025-06-19
AI Technical Summary
Existing optimization algorithms struggle to select parameter values that balance manufacturing efficiency and performance specifications, often compromising on either efficiency or performance.
A parameter search device and method that sets target parameters and performance, constructs an objective function combining performance evaluation and parameter targeting, and selects parameter values using a model-based approach to satisfy both efficiency and performance criteria.
The solution enables the selection of reasonable parameter values that optimize both manufacturing efficiency and performance, ensuring stability and quality in product manufacturing.
Smart Images
Figure 2025092107000001_ABST
Abstract
Description
Technical Field
[0001] Embodiments of the present invention relate to a parameter search device, method, and program.
Background Art
[0002] When manufacturing a product using a manufacturing apparatus, it is necessary to set parameters related to manufacturing such as temperature, pressure, and processing time for the manufacturing apparatus, and the quality of the product depends on the setting. Therefore, in order to manufacture a product with stable quality, it is necessary to appropriately adjust the parameter values. Although the adjustment of parameter values is often performed by a technician based on experience and intuition, in recent years, in order to eliminate personalization and improve the efficiency of adjustment, optimization algorithms such as Bayesian optimization may be used to determine the parameter values of the manufacturing apparatus. In the following, it will be described on the assumption that quality can be ensured by satisfying a certain standard of performance, but it is not limited to performance as long as it is an index for ensuring quality.
[0003] Typical optimization algorithms search for one point with the best function value within the search region. However, when there are multiple candidate solutions that satisfy the performance specification (performance target value with a margin) for the objective function value (index related to the performance of the product), there is a demand to select manufacturing parameters with excellent manufacturing efficiency (manufacturing time, power consumption, etc.).
[0004] In the penalty function method, the constraints of optimization are incorporated into the objective function and solved as an unconstrained optimization problem. If optimization is performed by applying the performance of the product to the original objective function and the manufacturing efficiency to the constraint term in this method, parameter values that are somewhat good for both the performance of the product and the manufacturing efficiency can be obtained. However, in this method, since the objective function itself changes due to the constraint term, there is a possibility that the performance specification may not be satisfied when selecting the point with the best objective function value. Further, when using an optimization method that uses a model such as Bayesian optimization as the optimization algorithm, since the objective function including the constraint term is modeled, it is not possible to refer to the model of the performance of the product or a function related only to the performance.
Prior Art Documents
Non-Patent Documents
[0005]
Non-Patent Document 1
Summary of the Invention
Problems to be Solved by the Invention
[0006] The problem to be solved by the present invention is to provide a parameter search device, method, and program capable of selecting a reasonable parameter value from the viewpoints of both parameter values and performance values.
Means for Solving the Problems
[0007] The parameter search device according to the embodiment includes a first setting unit that sets a target parameter representing a target of a parameter value, a second setting unit that sets a target performance representing a target of a performance value, a third setting unit that sets an objective function based on a first function that evaluates a parameter value based on a performance value and a second function that evaluates a parameter value based on the target parameter, a calculation unit that calculates a first objective function value by applying a first performance value corresponding to a first parameter value to be evaluated or both the first parameter value and the first performance value to the objective function, a first storage unit that stores the first objective function value in association with the first parameter value, a second storage unit that stores the first performance value in association with the first parameter value, a search unit that constructs a first model representing the objective function based on a combination of the first parameter value and the first objective function value stored in the first storage unit and determines a first parameter value of the next evaluation target based on the first model, and a selection unit that selects a first parameter value that satisfies a selection criterion based on the target performance and the target parameter from among the first parameter values stored in the second storage unit.
Brief Description of the Drawings
[0008]
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Embodiment for Carrying Out the Invention
[0009] Hereinafter, a parameter search device, method, and program according to this embodiment will be described with reference to the drawings.
[0010] FIG. 1 is a diagram showing a functional configuration example of a parameter adjustment system 1 according to the present embodiment. As shown in FIG. 1, the parameter adjustment system 1 is a computer system having a parameter search device 10 and an external device 20. The parameter search device 10 and the external device 20 are communicably connected via wire or wireless. The parameter search device 10 is a computer that searches for an optimal parameter value. The parameter search device 10 supplies a parameter value to the external device 20. The external device 20 is a computer that outputs a performance value corresponding to the supplied parameter value. Specifically, the external device 20 receives a parameter value from the parameter search device 10 and outputs a performance value, which is a value obtained by evaluating the results of simulations and experiments performed based on the received parameter value. The performance value is supplied to the parameter search device 10. Note that since the parameter search device 10 only needs to be able to acquire the performance value output by the external device 20, the parameter search device 10 and the external device 20 do not necessarily need to be connected by wire or wireless.
[0011] As shown in FIG. 1, the parameter search device 10 includes a first storage unit 101, a search unit 102, a calculation unit 103, a second storage unit 104, a selection unit 105, an objective function setting unit 106, a target parameter setting unit 107, a target performance setting unit 108, a display control unit 109, and a control unit 110.
[0012] The first storage unit 101 stores a combination of a parameter value and an objective function value acquired from the calculation unit 103. The parameter value is not particularly limited, but as an example, parameter values related to manufacturing such as the set temperature, set pressure, and processing time of a manufacturing apparatus are assumed. The objective function value is a value of the objective function calculated by the calculation unit 103 based on the parameter value.
[0013] FIG. 2 is a diagram showing an example of a combination of parameter values and objective function values stored in the first storage unit 101. There is one or more parameter values for one ID. In the example of FIG. 2, there are four parameter values from parameter 1 to parameter 4. For example, for a certain manufacturing apparatus, there are four types of parameter values to be set, namely, set temperature, set pressure, set humidity, and processing time, and a combination of these four types of parameter values is associated with one ID. Each parameter value may be represented by any of a continuous value, a discrete value, and a categorical variable. One objective function value is associated with the combination of the four types of parameter values corresponding to this one ID.
[0014] Based on the parameter values and objective function values stored in the first storage unit 101, the search unit 102 constructs a first model representing the objective function, and determines the parameter value of the next evaluation target based on the first model. As the parameter value of the next evaluation target, a parameter value with a high possibility of maximizing or minimizing the objective function is searched for. In this embodiment, maximization of the objective function will be described, but the objective function may be minimized according to the problem setting. The first model represents the relationship between the parameter value and the objective function value. The parameter value of the next evaluation target is input to the external device 20 by the calculation unit 103, and a simulation or an experiment is executed.
[0015] FIG. 3 is a diagram showing the relationship between the input and output of the first model. The search unit 102 uses the combination of the parameter values stored in the first storage unit 101 and the objective function values corresponding thereto as learning data, and trains the first model based on a training method such as Gaussian process regression, random forest, or neural network. Thereby, a first model having the parameter value as an input and the objective function value as an output is constructed.
[0016] The calculation unit 103 applies a performance value corresponding to a parameter value to be evaluated, or both the parameter value to be evaluated and the performance value corresponding to the parameter value, to the objective function set by the objective function setting unit 106 to calculate an objective function value. Specifically, the calculation unit 103 supplies the parameter value to be observed next to the external device 20. The external device 20 outputs an observed value that is the simulation and / or experimental result of the parameter value. The calculation unit 103 acquires the observed value from the external device 20. Assume that the observed value is a value (performance value) representing the performance of the product. The calculation unit 103 sends the combination of the parameter value and the objective function value to the first storage unit 101, and the combination of the parameter value and the performance value to the second storage unit 104.
[0017] The second storage unit 104 stores the combination of the parameter value and the performance value acquired from the calculation unit 103.
[0018] FIG. 4 is a diagram showing an example of the combination of the parameter value and the performance value stored in the second storage unit 104. The IDs of FIG. 2 and FIG. 4 are linked. If the IDs of both are the same, the parameter value shown in FIG. 4 is the same as the parameter value shown in FIG. 2. One performance value is associated with the combination of four types of parameter values for one ID. The performance value is a value obtained by evaluating the performance exhibited by the combination of four types of parameter values for one ID. Note that a plurality of performance values may be associated with the combination of parameter values.
[0019] The selection unit 105 selects a parameter value that satisfies the selection criteria based on the target performance and the target parameter from among the combinations of the parameter value and the performance value stored in the second storage unit 104. The selected parameter value is output. As an example of the parameter value that satisfies the selection criteria, the selection unit 105 selects the parameter value that is closest to the target parameter among the parameter values whose performance value satisfies the target performance.
[0020] The objective function setting unit 106 sets an objective function based on a first function that evaluates parameter values based on performance values and a second function that evaluates parameter values based on target parameters. The first function is, for example, a function that evaluates a parameter value higher as the performance value is higher, and is defined based on one or more performance values. The second function is, for example, a function that evaluates the distance between the parameter value and the target parameter. The second function may be a function that gives a larger penalty as the parameter value is farther from the target parameter. For example, the objective function is defined by the sum of the first function and the second function with a weight coefficient. Note that the design of the first function is not limited to the above, and for example, the first function may be designed to evaluate a parameter value higher as the performance value is lower. Also, the design of the second function is not limited to the above, and for example, it may be a function that evaluates the similarity between the parameter value and the target parameter or other metrics.
[0021] The target parameter setting unit 107 sets target parameters. The target parameters are given as a scalar value, a range, a function, etc. The range specifically means between an upper limit value and a lower limit value. The target parameters are referred to, for example, by the objective function setting unit 106 to set the objective function and by the selection unit 105 to select parameter values.
[0022] The target performance setting unit 108 sets target performance. The target performance is given as a scalar value, a range, a function, etc. The range specifically means between an upper limit value and a lower limit value. The target performance means performance specifications. The target performance is referred to, for example, by the objective function setting unit 106 to create the objective function and by the selection unit 105 to select parameter values.
[0023] The display control unit 109 displays various information on a display device. For example, the display control unit 109 may display the parameter value selected by the selection unit 105 on the display device. As another example, the display control unit 109 may display the relationship between the parameter value and the performance value in a graph.
[0024] The control unit 110 comprehensively controls the parameter search device 10. Specifically, the control unit 110 repeats, until the termination condition is satisfied, the construction of the first model by the search unit 102, the determination of the next parameter value by the search unit 102, the acquisition of the performance value and the calculation of the objective function value by the calculation unit 103, the storage of the combination of the parameter value and the performance value in the second storage unit 104, and the storage of the combination of the parameter value and the objective function value in the first storage unit 101. The control unit 110 may control the setting by the objective function setting unit 106.
[0025] FIG. 5 is a diagram showing an example of parameter search processing by the parameter search device 10 according to the present embodiment. As shown in FIG. 5, first, the target parameter setting unit 107 sets a target parameter, and the target performance setting unit 108 sets a target performance (step S101). The target parameter may be a set of scalar values having the same number as the number of parameters to be searched by the parameter search device 10, or may be a certain range or function. For example, the parameter value that minimizes the manufacturing cost may be set as the target parameter, or the range of parameter values within which the manufacturing cost is within an allowable range may be set. Further, the target performance setting unit 108 sets the performance specifications to be satisfied regarding the performance value of the product as the target performance. The target performance may be a scalar value or a certain range. Also, the target performance may be one performance value or a plurality of performance values. The target parameter and the target performance may be set according to an instruction by the user or according to an arbitrary algorithm.
[0026] When step S101 is performed, the objective function setting unit 106 sets an objective function for parameter optimization (step S102). In step S102, the objective function setting unit 106 acquires the target parameter and the target performance set in step S101, and sets an objective function based on the target parameter and the target performance. The objective function setting unit 106 sets the form, coefficients, etc. of the objective function according to an instruction by the user.
[0027] As an example, the objective function f´(x) is represented by the following formula (1).
[0028]
Number
[0029] The first function f(x) of the objective function is a function that evaluates the parameter value x based on the performance value that is the response of the external device 20 to the parameter value x. The parameter x assumes a vector value of multiple dimensions such as four dimensions, as illustrated in FIGS. 2 and 4, for example. The first function f(x) may use a function that evaluates the parameter value x higher as the performance value is better, such as the squared error or the absolute error between the target performance and the performance value corresponding to the parameter value x. The first function f(x) may be calculated based on one or more performance values. When there are multiple performance values, for example, it may be a function in which multiple terms corresponding to the multiple performance values are added or the like.
[0030] The second function g(x) of the objective function is a term that evaluates the current parameter value x based on the target parameter. The second function g(x) may use a function that evaluates the parameter value x higher as the parameter value x is closer to the target parameter, such as the Euclidean distance between the parameter value and the target parameter, that is, a function that penalizes the objective function more as they are farther apart.
[0031] The coefficient w for the second function g(x) is a coefficient that controls the influence degree of the second term. The smaller the value of the coefficient w, the relatively greater the influence of f(x), and the larger the value of the coefficient w, the relatively greater the influence of g(x). That is, the smaller the value of the coefficient w, the more the performance value is emphasized, and the larger the value of the coefficient w, the more the target parameter is emphasized.
[0032] When step S102 is performed, the calculation unit 103 determines the initial parameter value x in order to create the initial data for building the first model (step S103). init It may be determined randomly or specified by the user. Also, the number of elements of the initial parameter value is (x init 1 init, x2 init, There may be a plurality of them, such as ···).
[0033] When step S103 is performed, the calculation unit 103 supplies the initial parameter value x determined in step S103 init to the external device 20 and acquires the performance value corresponding to the parameter value, which is the output from the external device 20 (step S104). Here, the performance value may be a single value p or a vector (p1, p2, ···) consisting of a plurality of values. Also, when a plurality of initial parameter values (x 1 init, x 2 init, ···) are given, the corresponding performance values (p 1 , p 2 , ···) are acquired for all the initial parameter values. When the performance value is a vector, ((p 1 1, p 1 2, ···), (p 2 1, p 2 2, ···), ···) is obtained.
[0034] When step S104 is performed, the calculation unit 103 stores the combination of the initial parameter value x determined in step S103 init or the observation parameter value x determined in step S109 described later next and the corresponding performance value p in the second storage unit 104 (step S105). The combination of the observation parameter value and the performance value is (x init , p) or (x next , p) when the performance value is a single value, and (x init , (p1, p2)) or (x next , (p1, p2)) when the performance value is a plurality of values.
[0035] When step S105 is performed, the calculation unit 103 calculates the value f´(x next or the observation parameter value x next and the corresponding performance value p, and calculates the value f´(x init ) or f´(xnext ) and calculate the initial parameter value x init and the objective function value f´(x init ) and combination (x init ,f´(x init )), or the observed parameter value x next and the objective function value f´(x next ) and combination (x next ,f´(x next ) is stored in the first storage unit 101 (S106).
[0036] After step S106, the control unit 110 judges whether or not an end condition for the optimization is satisfied (step S107). The end condition may be, for example, whether or not a predetermined number of iterations has been exceeded, or whether or not a predetermined elapsed time has been exceeded, or whether or not the objective function value f'(x) exceeds a predetermined value, or whether or not a predetermined number or more of performance values that satisfy the target performance have been collected, or a combination of these conditions may be used.
[0037] If it is determined that the termination condition is not satisfied (step S107: NO), the search unit 102 acquires a combination (x, f'(x)) of the parameter value and the objective function value from the first storage unit 101, and constructs a first model representing the objective function from the acquired combination (step S108). The first model may be constructed by Gaussian process regression, random forest, neural network, or the like.
[0038] When step S108 is performed, the search unit 102 searches for the next parameter value x next For example, the search unit 102 determines the next observed parameter value x next The next parameter value x nextTo determine it, it is possible to use EI (Expected Improvement) or UCB (Upper Confidential Bound), which are acquisition functions used in Bayesian optimization. EI and UCB are functions represented by the following equations (2) and (3), respectively.
[0039]
Equation
[0040] μ is the expected value of the first model, and σ is the standard deviation of the first model. In Equation (2), f´ max is the maximum value of the objective function f´ observed so far, Φ is the cumulative distribution function of the normal distribution, φ is the probability density function of the normal distribution, and ξ is an adjustment constant specified by the user. In Equation (3), β is a weight specified by the user indicating the degree of considering σ.
[0041] When step S109 is performed, for the parameter value determined in step S109, acquisition of the performance value (S104), storage of the combination of the parameter value and the performance value (S105), storage of the combination of the parameter value and the objective function value (S106), and determination of the end condition (S107) are performed. In this way, the processing from step S104 to S109 is repeated until the end condition is satisfied.
[0042] And when it is determined in step S107 that the end condition is satisfied (step S107: YES), the selection unit 105 selects the output parameter value (step S110). In step S110, the selection unit 105 first acquires the combination of the parameter value and the performance value from the second storage unit 104. The selection unit 105 selects a parameter value that satisfies the selection criterion based on the target performance and the target parameter from among these combinations. In this embodiment, it is assumed that the selection criterion is the parameter value that is closest to the target parameter among the parameter values whose performance value satisfies the target performance.
[0043] FIG. 6 is a diagram schematically showing an example of output parameter selection by the selection unit 105. The left vertical axis in FIG. 6 represents the objective function value, the right vertical axis represents the performance value, and the horizontal axis represents the parameter value. The thick line represents the true function of the performance value with respect to the parameter value, and the thin line represents the true function of the objective function value with respect to the parameter value. The circles represent the parameter value supplied by the calculation unit 103 to the external device 20, the performance value acquired from the external device 20 with respect to the parameter value, and the objective function value calculated from the acquired performance value. A target parameter 51 is set for the parameter value, and a performance specification 52 which is the target performance specified within a range is set for the performance value.
[0044] In the case of a comparative example in which the output parameter value is searched based only on the objective function value, in the example of FIG. 6, since the parameter value P1 at which the objective function value becomes maximum is set as the output parameter value, the performance value does not satisfy the performance specification 52. On the other hand, in the case of the present embodiment, since the parameter P2 which is closest to the target parameter 51 among the parameter values satisfying the performance specification 52 is selected as the output parameter value, the performance specification 52 is surely satisfied.
[0045] FIG. 7 is a diagram showing the parameter search results of the first comparative example (standard Bayesian optimization; hereinafter referred to as standard BO), the second comparative example (standard BO + penalty function method), and the present embodiment. Standard BO is a method of using, as an objective function, a first function f(x) that evaluates a parameter value x based on a performance value in Equation (1), and searching for a parameter value that maximizes the objective function value. The first comparative example uses the standard BO. The penalty function method is a method of using an objective function with a constraint term added. The second comparative example uses a method in which the penalty function method is combined with standard BO, that is, a method of searching for a parameter value that maximizes the objective function value of the objective function with a penalty term added. As described above, the present embodiment is a method in which the selection method of selecting the parameter value closest to the target parameter among the parameter values whose performance values satisfy the target performance is further combined with the standard BO + penalty function method. The target performance value (target performance specified as a scalar value) is 16.800, and the performance specification (target performance specified as a range) is 16.783 to 16.817. The target parameter is set to a value in which each parameter value constituting them is 0.
[0046] As shown in FIG. 7, in the case of the first comparative example, the performance value is close to the target performance value, but the output parameter value cannot be made close to the target parameter. In the case of the second comparative example, the output parameter value can be made close to the target parameter, but since the objective function value changes by the amount of the penalty, there may be a case where the performance value does not satisfy the performance specification as in the example. In the case of the present embodiment, the performance value satisfies the performance specification, and the output parameter value can be made close to the target parameter.
[0047] When step S110 is performed, the parameter search process according to the present embodiment ends.
[0048] The parameter search process shown in FIG. 5 is an example, and the present embodiment is not limited thereto. Various additions, deletions, and / or changes to the process are possible without departing from the spirit of the invention.
[0049] As an example, the parameter search device 10 according to the above embodiment has the first storage unit 101 that stores combinations of parameter values and objective function values and the second storage unit 104 that stores combinations of parameter values and performance values. However, the present embodiment is not limited to this. For example, the parameter search device 10 may have a storage unit that stores combinations of parameter values, objective function values, and performance values. As another example, the parameter search device 10 may have a storage unit that stores both combinations of parameter values and objective function values and combinations of parameter values and performance values.
[0050] As another example, the output parameter value selected in step S110 may be displayed by the display control unit 109 in a predetermined layout. In this case, the display control unit 109 may display only the output parameter value, or may display a combination of the output parameter value, the performance value, and the objective function value. The output parameter value or the like may be displayed numerically, or may be displayed as a graph as shown in FIG. 6. When displayed as the graph of FIG. 6, the display control unit 109 may display the target parameter, the performance specification, etc. together with the output parameter value or the like.
[0051] In the above embodiment, the objective function f´(x) is the sum of the first function f(x) and the second function g(x) with the weight coefficient w as in the above formula (1). However, the present embodiment is not limited to this. As another example, the objective function f´(x) may be the sum of f(x) and g(x) with the weight coefficient w.
[0052] In the above embodiment, the selection unit 105 selects, as an example of the parameter value that satisfies the selection criteria, the parameter value that is closest to the target parameter value among the parameter values whose performance value satisfies the target performance value. However, this embodiment is not limited thereto. As another example, the selection unit 105 may select, as the parameter value that satisfies the selection criteria, the parameter value whose performance value is closest to the target performance value among the parameter values whose parameter value satisfies the target parameter value. In this case, it is preferable that the target parameter value is specified within a range and the performance value is specified as a scalar value. By designing the selection criteria in this way, it is possible to surely select a parameter value that satisfies the target parameter value as compared with the above embodiment. The selection criteria may be designed according to the user's requirements.
[0053] (Application Example) Next, a parameter adjustment system according to an application example of this embodiment will be described. In the following description, components having substantially the same functions as those in the above embodiment are denoted by the same reference numerals, and duplicate descriptions will be given only when necessary.
[0054] FIG. 8 is a diagram showing a functional configuration example of a parameter adjustment system 2 according to an application example. As shown in FIG. 8, the parameter adjustment system 2 includes a parameter search device 30 and an external device 20. The parameter search device 10 is a computer that searches for an optimal parameter value. The parameter search device 30 includes a first storage unit 101, a search unit 102, a calculation unit 103, a second storage unit 104, a selection unit 105, an objective function setting unit 106, a target parameter setting unit 107, a target performance setting unit 108, a display control unit 109, and a control unit 110, and further includes a model construction unit 111 and a visualization image generation unit 112.
[0055] The model construction unit 111 constructs a second model that infers a performance value from a parameter value based on the combination of the parameter value and the performance value stored in the second storage unit 104.
[0056] FIG. 9 is a diagram showing the relationship between the input and output of the second model. The model construction unit 111 uses the combination of the parameter values stored in the second storage unit 104 and the performance values corresponding thereto as learning data, and trains the second model based on a training method such as Gaussian process regression, random forest, or neural network. Thereby, a second model is constructed in which the parameter value is used as an input and the performance value is used as an output.
[0057] The visualization image generation unit 112 generates an image (hereinafter referred to as a visualization image) that visualizes the correspondence between the parameter value and the performance value based on the second model constructed by the model construction unit 111. For example, the visualization image generation unit 112 inputs a set of parameter values into the second model to calculate a set of performance values, and generates a visualization image based on the set of parameter values and the set of performance values. The visualization image may be displayed on a display device by the display control unit 109.
[0058] By using the second model, the parameter search device 30 can predict the performance value corresponding to the parameter value without using the external device 20. By generating a visualization image based on the second model, the user can easily grasp the correspondence between the parameter value and the performance value. The visualization image may be displayed so as to be observable from outside the parameter search device 30.
[0059] <Hardware Configuration> FIG. 10 is a diagram showing a hardware configuration example of the parameter search devices 10 and 30. As shown in FIG. 10, the parameter search devices 10 and 30 are computers having a processor 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, an auxiliary storage device 14, an input device 15, a display device 16, and a communication device 17. The data and various signals of the processor 11, ROM 12, RAM 13, auxiliary storage device 14, input device 15, display device 16, and communication device 17 are transmitted and received via a bus.
[0060] Processor 11 is an integrated circuit that controls the overall operation of the parameter search devices 10 and 30. For example, processor 11 has a CPU (Central Processing Unit), GPU (Graphics Processing Unit), DSP (Digital Signal Processor), and / or FPU (Floating-Point Unit). Processor 11 may include an internal memory and an I / O interface. Processor 11 executes various processes shown in the above embodiments by interpreting and computing programs pre-stored in ROM 12 or auxiliary storage device 14, etc. Note that part or all of processor 11 may be realized by hardware such as an ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array).
[0061] ROM 12 is a non-volatile memory that stores various data. For example, ROM 12 stores data, setting values, etc. used when processor 11 executes various processes. ROM 12 may have a non-transitory computer-readable storage medium that stores programs executed by processor 11.
[0062] RAM 13 is a volatile memory used for reading and writing data. RAM 13 temporarily stores data used when processor 11 executes various processes. RAM 13 provides a work area for processor 11.
[0063] Auxiliary storage device 14 is a non-volatile memory that stores various data. For example, auxiliary storage device 14 stores data, setting values, data generated by various processes in processor 11, etc. used when processor 11 executes various processes. Auxiliary storage device 14 is composed of an HDD (Hard Disk Drive), SSD (Solid State Drive), an integrated circuit storage device, etc. Note that auxiliary storage device 14 may have a non-transitory computer-readable storage medium that stores programs executed by processor 11.
[0064] The input device 15 receives inputs of various operations from the operator. As the input device 15, a keyboard, a mouse, various switches, a touch pad, a touch panel display, etc. can be used. An electrical signal (hereinafter referred to as an operation signal) corresponding to the received input of the operation is supplied to the processor 11.
[0065] The display device 16 displays various data according to the control by the processor 11. As the display device 16, a CRT (Cathode-Ray Tube) display, a liquid crystal display, an organic EL (Electro Luminescence) display, an LED (Light-Emitting Diode) display, a plasma display, or any other arbitrary display can be appropriately used. The display device 16 may be a projector.
[0066] The communication device 17 includes a communication interface such as a network interface card (NIC: Network Interface Card) for performing data communication with various devices connected to the parameter search devices 10, 30 via a network. Note that an operation signal may be supplied from a computer connected via the communication device 17 or an input device included in the computer, or various data may be displayed on a display device or the like included in the computer connected via the communication device 17. The input device 15 can be replaced with a computer connected via the communication device 17 or an input device included in the computer, and the display device 16 can be replaced with a display device or the like included in the computer connected via the communication device 17.
[0067] The parameter search devices 10 and 30 do not necessarily include a part of the processor 11, ROM 12, RAM 13, auxiliary storage device 14, input device 15, display device 16, and communication device 17. The parameter search devices 10 and 30 may be provided with any additional hardware devices useful for executing the processing according to the present embodiment. The parameter search devices 10 and 30 do not necessarily have to be physically constituted by one computer, and may be constituted by a computer system having a plurality of computers communicably connected via a wired or network line or the like. The assignment of a series of processing according to the present embodiment to the plurality of processors 11 respectively mounted on the plurality of computers can be arbitrarily set. All the processors 11 may execute all the processing in parallel, or a specific processing may be assigned to one or some of the processors 11, and a series of processing according to the present embodiment may be executed as the entire computer system.
[0068] Thus, according to the present embodiment, it becomes possible to provide a parameter search device, method, and program capable of selecting a parameter value appropriate from both viewpoints of the parameter value and the performance value.
[0069] Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalent scope thereof.
Description of Reference Numerals
[0070] 1…Parameter adjustment system, 2…Parameter adjustment system, 10…Parameter search device, 11…Processor, 12…ROM, 13…RAM, 14…Auxiliary storage device, 15…Input device, 16…Display device, 17…Communication device, 20…External device, 30…Parameter search device, 51…Target parameter, 52…Target performance (performance specification), 101…First storage unit, 102…Search unit, 103…Calculation unit, 104…Second storage unit, 105…Selection unit, 106…Target function setting unit, 107…Target parameter setting unit, 108…Target performance setting unit, 109…Display control unit, 110…Control unit, 111…Model construction unit, 112…Visualization image generation unit.
Claims
1. A first setting unit that sets a target parameter representing a target of a parameter value; A second setting unit that sets a target performance representing a target of a performance value; A third setting unit that sets an objective function based on a first function that evaluates a parameter value based on a performance value and a second function that evaluates a parameter value based on the target parameter; A calculation unit that calculates a first objective function value by applying a first performance value corresponding to a first parameter value to be evaluated, or both the first parameter value and the first performance value, to the objective function; A first storage unit that stores the first objective function value in association with the first parameter value; A second storage unit that stores the first performance value in association with the first parameter value; A search unit that constructs a first model representing the objective function based on a combination of the first parameter value and the first objective function value stored in the first storage unit, and determines a first parameter value of the next evaluation target based on the first model; A selection unit that selects a first parameter value that satisfies a selection criterion based on the target performance and the target parameter from among the first parameter values stored in the second storage unit; A parameter search device comprising:
2. The parameter search device according to claim 1, wherein the selection unit selects, as the first parameter value that satisfies the selection criterion, the first parameter value that is closest to the target parameter among the first parameter values for which the first performance value satisfies the target performance.
3. The parameter search device according to claim 1, wherein the target parameter is given as a scalar value, a range, or a function.
4. The parameter search device according to claim 1, wherein the target performance is given as a scalar value, a range, or a function.
5. The parameter search device according to claim 1, wherein one or more target performances are given.
6. The parameter search device according to claim 1, wherein the first function evaluates the first parameter value higher as the first performance value is higher.
7. The parameter search device according to claim 1, wherein the first function is defined based on one or more performance values.
8. The parameter search device according to claim 1, wherein the second function evaluates the distance between the first parameter value and the target parameter.
9. The parameter search device according to claim 1, wherein the second function gives a larger penalty as the first parameter value is farther from the target parameter.
10. The parameter search device according to claim 1, wherein the objective function is a weighted sum of the first function and the second function.
11. The parameter search device according to claim 1, wherein the calculation unit supplies the first parameter value to an external device, and acquires, as the first performance value, a performance value output in response to the supply of the first parameter value to the external device.
12. The parameter search device according to claim 1, wherein the first model is based on Gaussian process regression, random forest, or a neural network.
13. The parameter search device according to claim 1, further comprising a construction unit that constructs a second model for inferring a second performance value from a second parameter value based on a combination of the first parameter value and the first performance value stored in the second storage unit.
14. A visualization image generation unit that generates an image visualizing the correspondence between the second parameter value and the second performance value based on the second model, and a display control unit that displays the image on a display device. The parameter search device according to claim 13.
15. The parameter search device according to claim 1, further comprising a display control unit that displays the selected first parameter value on a display device.
16. Set a target parameter representing the target of the parameter value, Set a target performance representing the target of the performance value, Set an objective function based on a first function for evaluating a parameter value based on a performance value and a second function for evaluating a parameter value based on the target parameter, Calculate a first objective function value by applying a first performance value corresponding to the first parameter value to be evaluated, or both the first parameter value and the first performance value, to the objective function, Store the first objective function value in a first storage unit in association with the first parameter value, Store the first performance value in a second storage unit in association with the first parameter value, Construct a first model representing the objective function based on the combination of the first parameter value and the first objective function value stored in the first storage unit, determine the first parameter value of the next evaluation target based on the first model, Select a first parameter value that satisfies a selection criterion based on the target performance and the target parameter from among the first parameter values stored in the second storage unit. A parameter search method comprising the above.
17. Cause a computer to Have a function of setting a target parameter representing the target of the parameter value, Have a function of setting a target performance representing the target of the performance value, Have a function of setting an objective function based on a first function for evaluating a parameter value based on a performance value and a second function for evaluating a parameter value based on the target parameter, Have a function of calculating a first objective function value by applying a first performance value corresponding to the first parameter value to be evaluated, or both the first parameter value and the first performance value, to the objective function, A function of storing the first objective function value in the first storage unit in association with the first parameter value; A function of storing the first performance value in the second storage unit in association with the first parameter value; A function of constructing a first model representing the objective function based on the combination of the first parameter value and the first objective function value stored in the first storage unit, and determining the first parameter value of the next evaluation target based on the first model; A function of selecting a first parameter value that satisfies the selection criteria based on the target performance and the target parameter from among the first parameter values stored in the second storage unit; A parameter search program for realizing the above.