Information processing system, information processing method, and information processing program

Abstract

To search for a feature amount of a candidate component for constituting a desired composition, at a higher speed.SOLUTION: An information processing system according to one example comprises at least one processor. The at least one processor repeats optimization using a regression model for calculating a physical property value of a composition from feature amount vectors indicating feature amounts of a candidate component group based on a plurality of candidate components, while executing an exclusion process for excluding at least one candidate component from the candidate component group, and specifies feature amounts of one or more candidate components remaining after the repetition. The exclusion process includes steps of: selecting a feature amount vector corresponding to a solution to the physical property value obtained by the optimization; and excluding at least one candidate component from the candidate component group on the basis of the selected feature amount vector.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] One aspect of the present disclosure relates to an information processing system, an information processing method, and an information processing program.

Background Art

[0002] Conventionally, a method for searching for characteristic quantities of candidate components for constituting a desired composition has been known. For example, Patent Document 1 describes a method for creating a blending plan for mixing a plurality of types of blending raw materials using a mathematical programming method. Patent Document 2 discloses a method for planning the blending ratio of raw materials by solving a raw material blending plan formulated as a mixed integer programming problem including non-linear variables.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the methods described in Patent Documents 1 and 2, when the number of types of candidate components is extremely large, a great deal of calculation time is required to obtain the characteristic quantities of the candidate components for constituting the desired composition. Therefore, a method for more quickly searching for the characteristic quantities is desired.

Means for Solving the Problems

[0005] An information processing system relating to one aspect of this disclosure comprises at least one processor. The at least one processor repeatedly performs optimization using a regression model that calculates physical properties of a composition from feature vectors representing the features of each of a group of candidate components based on multiple candidate components, while performing an exclusion process that excludes at least one candidate component from the group of candidate components, and identifies the features of each of the one or more candidate components that remain after the iteration. The exclusion process includes the steps of selecting a feature vector corresponding to the solution of physical properties obtained by optimization, and excluding at least one candidate component from the group of candidate components based on the selected feature vector.

[0006] An information processing method relating to one aspect of this disclosure is performed by an information processing system comprising at least one processor. This information processing method includes the steps of: repeatedly performing an optimization using a regression model that calculates physical property values ​​of a composition from feature vectors representing the respective feature quantities of a group of candidate components based on a plurality of candidate components, while performing an exclusion process to exclude at least one candidate component from the group of candidate components; and identifying the respective feature quantities of the one or more candidate components that remain after the repeated steps. The exclusion process includes the steps of selecting a feature vector corresponding to the solution of physical property values ​​obtained by optimization; and excluding at least one candidate component from the group of candidate components based on the selected feature vector.

[0007] An information processing program relating to one aspect of this disclosure causes a computer to perform the following steps: an optimization using a regression model that calculates the physical properties of a composition from feature vectors representing the features of each of a group of candidate components based on multiple candidate components, while repeatedly performing an exclusion process to exclude at least one candidate component from the group of candidate components; and a step of identifying the features of each of the one or more candidate components that remain after the repeated steps. The exclusion process includes the steps of selecting a feature vector corresponding to the solution of the physical properties obtained by the optimization, and excluding at least one candidate component from the group of candidate components based on the selected feature vector.

[0008] In this respect, optimization using a regression model that calculates the physical properties of a composition from the feature vectors of candidate components is repeatedly performed while reducing the number of candidate components. By performing optimization step by step while narrowing down the candidate components, it is possible to quickly search for the feature vectors of candidate components necessary to constitute the desired composition. [Effects of the Invention]

[0009] According to one aspect of the present invention, the characteristic quantities of candidate components for constituting a desired composition can be searched more quickly. [Brief explanation of the drawing]

[0010] [Figure 1] This figure shows an example of the functional configuration of the information processing system according to the embodiment. [Figure 2] A flowchart illustrating an example of the operation of the information processing system according to the embodiment. [Figure 3] A flowchart illustrating an example of optimization. [Modes for carrying out the invention]

[0011] Embodiments of this disclosure will be described in detail below with reference to the attached drawings. In the description of the drawings, the same or equivalent elements are denoted by the same reference numerals, and redundant descriptions are omitted.

[0012] [System Configuration] The information processing system 10 according to this embodiment is a computer system that searches for feature quantities of candidate components for constituting a desired composition. A composition is a substance composed by processing one or more components. Typically, a composition is obtained using two or more components. For example, a composition may be an inorganic composition, an organic composition such as a resin composition, or a feed composition. Components of a composition are substances that are intentionally used to produce the composition. Components may or may not retain their original form when the composition is completed. In this disclosure, components that can constitute a desired composition are also referred to as "candidate components." Individual candidate components are prepared in correspondence with the composition and may be various materials such as inorganic materials or organic materials.

[0013] In one example, the information processing system 10 repeatedly performs optimization using a pre-trained regression model to finally identify feature quantities of two or more candidate components for constituting a desired composition. In this disclosure, “desired composition” means a composition having desired physical properties. Examples of feature quantities of candidate components include the amount of each component used. This amount may be expressed as an absolute quantity, a ratio (composition ratio), or by other indicators. Individual feature quantities may be expressed as a weighted average or as a linear weighted sum. For example, the amount of each component may be expressed as a weighted average. Examples of physical properties of a composition include the glass transition temperature, elastic modulus, and coefficient of thermal expansion.

[0014] A regression model is a computational model for determining the value of one or more dependent variables y based on the values ​​of one or more independent variables x. In one example, a regression model accepts feature vectors based on the features of each of a group of candidate components as independent variables, and calculates the physical properties of a composition as the dependent variable. In this disclosure, the phrase "multiple candidate components" means multiple types of candidate components. A group of candidate components can also be said to be a collection of multiple candidate components. A feature vector is an n-dimensional vector that shows the features of each of the n candidate components as vector elements, and can be represented as a one-dimensional array.

[0015] In a single optimization, the information processing system 10 modifies at least a portion of the feature vectors and inputs each feature vector into a regression model to obtain the physical properties of the composition. The information processing system 10 selects a provisional solution from the multiple physical properties and selects the feature vector corresponding to that solution, i.e., the feature vector input into the regression model to obtain that solution. The information processing system 10 performs an exclusion process to exclude at least one candidate component from the candidate component group based on the selected feature vector. Then, the information processing system 10 performs optimization again, assuming that some candidate components have been excluded. The information processing system 10 repeats the optimization while performing the exclusion process, and after each iteration, outputs the features of each of the one or more remaining candidate components as the final processing result.

[0016] In this way, the information processing system 10 performs optimization step by step while narrowing down the candidate components. As the number of candidate components considered in the optimization problem gradually decreases, the time required for each optimization is shortened with each iteration. Therefore, the time required to obtain the final processing result can be shortened compared to optimization that considers all candidate components from start to finish. This speedup of processing can be particularly noticeable when there are a very large number of candidate components, such as in materials development.

[0017] The information processing system 10 consists of one or more computers. When multiple computers are used, these computers are connected via a communication network such as the Internet or an intranet, thereby logically constructing a single information processing system 10.

[0018] The computer that constitutes the information processing system 10 generally includes a processor, a memory, and a communication interface as hardware devices. The processor is, for example, a CPU, and the memory is composed of a flash memory, a hard disk, etc., but the types of hardware devices that constitute the information processing system 10 are not limited to this and may be arbitrarily selected. Each function of the information processing system 10 is realized by the processor executing a program stored in the memory.

[0019] The information processing program for causing the computer to function as the information processing system 10 includes program codes for realizing each function module of the information processing system 10. This information processing program may be provided after being non-temporarily recorded on a tangible recording medium such as a CD-ROM, a DVD-ROM, or a semiconductor memory. Alternatively, the information processing program may be provided via a communication network as a data signal superimposed on a carrier wave. The provided information processing program is stored, for example, in a memory.

[0020] FIG. 1 is a diagram showing an example of the functional configuration of the information processing system 10. In this example, the information processing system 10 includes a processor 101. In one example, the processor 101 functions as an acquisition unit 11, an optimization unit 12, an exclusion unit 13, a determination unit 14, and an output unit 15. The acquisition unit 11 is a functional module that acquires a reference physical property value, which is a desired physical property value, and a candidate component group. The optimization unit 12 is a functional module that performs optimization using a regression model. The exclusion unit 13 is a functional module that excludes at least one candidate component from the candidate component group. The determination unit 14 is a functional module that determines whether to end the repetition of optimization. The output unit 15 is a functional module that outputs a processing result.

[0021] In one example, the information processing system 10 accesses a candidate component database 21 via a given communication network. The communication network may be configured to include at least one of the Internet and an intranet. The communication network can be configured using at least one of a wired network and a wireless network. The candidate component database 21 may be a component of the information processing system 10 or may be provided in a computer system different from the information processing system 10.

[0022] The candidate component database 21 is a database that stores candidate component data indicating individual candidate components. In one example, the candidate component data indicates a plurality of available components.

[0023] [Operation of the System] Referring to FIGS. 2 and 3, the operation of the information processing system 10 will be described and the information processing method according to the present embodiment will be described. FIG. 2 is a flowchart showing an example of processing in the information processing system 10. FIG. 3 is a flowchart showing an example of optimization processing.

[0024] In step S11, the acquisition unit 11 acquires a reference physical property value. In one example, the acquisition unit 11 receives a reference physical property value input by the user. Alternatively, the acquisition unit 11 may read a reference physical property value from a previously prepared physical property value database or may receive a reference physical property value transmitted from another computer.

[0025] In step S12, the acquisition unit 11 acquires a group of candidate components. In one example, the acquisition unit 11 accesses the candidate component database 21, reads candidate component data indicating multiple candidate components, and acquires these multiple candidate components as a group of candidate components. The acquisition unit 11 may acquire all candidate components recorded in the candidate component database 21 as a group of candidate components, or it may acquire at least two of those candidate components as a group of candidate components. Alternatively, the acquisition unit 11 may acquire multiple candidate components entered by the user as a group of candidate components without accessing the candidate component database 21. Alternatively, the acquisition unit 11 may receive a group of candidate components transmitted from another computer.

[0026] In step S13, the optimization unit 12 uses a regression model to optimize the features of candidate components. "Optimization of candidate component features" refers to a method of searching for one or more candidate components and the features of each candidate component in order to obtain a composition having physical properties that match or approximate the reference physical properties.

[0027] An example of the optimization process will be explained with reference to Figure 3. In step S131, the optimization unit 12 generates a feature vector. In one example, the optimization unit 12 generates a feature vector that represents the feature of each of the candidate component groups. The number of dimensions of the feature vector corresponds to the number of types of candidate components. That is, if the number of types of candidate components is n, the generated feature vector is an n-dimensional vector that represents n features as vector elements. The optimization unit 12 sets the feature of each candidate component, for example, using random search.

[0028] In step S132, the optimization unit 12 calculates the physical properties of the composition from the feature vector using a regression model. In this disclosure, the physical properties calculated by the regression model are also referred to as "estimated physical properties". The optimization unit 12 calculates the estimated physical properties by inputting the feature vector into the regression model. The regression model used by the optimization unit 12 may be Support Vector Regression (SVR) or Gaussian Process Regression (GPR). The optimization unit 12 may use a regression model stored in a given memory, or it may access a regression model existing outside the information processing system 10 via a given communication network.

[0029] In step S133, the optimization unit 12 determines whether the termination condition for calculating the estimated physical properties is met. For example, the termination condition may be set based on the number of times the regression model has been run. The number of runs may be set by the user or predetermined.

[0030] If the termination condition is not met (NO in step S133), the process returns to step S131, and the optimization unit 12 generates a new feature vector. The optimization unit 12 modifies at least one vector element of the feature vector to generate a new feature vector. To generate a new feature vector, the optimization unit 12 may use random search, a combination of random search and gradient descent, or a combination of random search and Newton's method. Alternatively, the optimization unit 12 may generate a new feature vector using a Bayesian optimization acquisition function.

[0031] If the termination condition is met (YES in step S133), the process proceeds to step S134, and the optimization unit 12 obtains at least one estimated physical property from among the multiple estimated physical properties that corresponds to the reference physical property as the solution. The "estimated physical property corresponding to the reference physical property" may be, for example, an estimated physical property that matches the reference physical property, or an estimated physical property that most closely approximates the reference physical property.

[0032] Returning to Figure 2, in step S14, the exclusion unit 13 excludes at least one candidate component from the candidate component group. That is, step S14 is an exclusion process. The exclusion unit 13 selects a feature vector corresponding to the estimated physical property values ​​obtained as a solution in step S13. Subsequently, the exclusion unit 13 applies at least one exclusion requirement to each feature shown in the feature vector and excludes at least one candidate component that satisfies any of the exclusion requirements from the candidate component group. For example, the exclusion unit 13 may exclude a candidate component that satisfies the exclusion requirement that its feature is less than or equal to a given threshold. Hereinafter, this exclusion requirement will also be referred to as the "first exclusion requirement". Alternatively, the exclusion unit 13 may exclude a candidate component that has the smallest feature in the feature vector. Hereinafter, this exclusion requirement will also be referred to as the "second exclusion requirement". Alternatively, the exclusion unit 13 may perform a process to exclude at least one candidate component using the first exclusion requirement, and if no candidate components are excluded in this process, it may use the second exclusion requirement to exclude at least one candidate component. Alternatively, the exclusion unit 13 may exclude the candidate component having the largest feature in the feature vector. In one example, the exclusion unit 13 excludes the candidate component from the group of candidate components by fixing the feature of the candidate component to be excluded to zero in the feature vector.

[0033] In step S15, the determination unit 14 determines whether the termination condition for the optimization iteration is met. For example, this termination condition may be set based on the number of types of candidate components remaining in the candidate component group, that is, the number of candidate components in the feature vector whose feature quantities are not fixed at zero. In this case, the determination unit 14 determines that the termination condition is met if that number is less than or equal to a given threshold. Alternatively, the termination condition may be that the optimization process and the exclusion process have been executed a given number of times.

[0034] If the termination condition is not met (NO in step S15), the process returns to step S13. In step S13, the optimization unit 12 executes the optimization process again, and in step S14, the exclusion unit 13 executes the exclusion process again.

[0035] In the repeated step S13, at least some of the features of the feature vector are fixed to zero. In step S131, the optimization unit 12 sets the features of the remaining candidate components using various methods such as random search, gradient descent, Newton's method, and acquisition function, while keeping the features of the excluded candidate components fixed to zero, to generate a new feature vector. At this time, the dimensionality of the generated feature vector is effectively the initial dimensionality of the feature vector minus the number of excluded candidate components. For example, if the initial feature vector is n dimensional and 3 candidate components are excluded, the newly generated feature vector will effectively be (n-3) dimensional. Therefore, the dimensionality of the newly generated feature vector effectively decreases each time the optimization and exclusion process is repeated.

[0036] If the termination condition is met (YES in step S15), the process proceeds to step S16, and the output unit 15 outputs the processing result. For example, the output unit 15 outputs the feature quantities of each of the remaining one or more candidate components as processing results. Alternatively, the output unit 15 may output estimated physical properties along with the feature quantities as processing results. For example, the output unit 15 may store the processing result in a given database, transmit it to another computer or computer system, or display it on a display device. Alternatively, the output unit 15 may output the processing result to another functional module for subsequent processing of the information processing system 10.

[0037] As described above, until the termination condition in step S15 is satisfied, the information processing system 10 repeats the optimization using a regression model that calculates the physical property values of the composition from the feature vectors indicating the respective feature amounts of the candidate component groups based on a plurality of candidate components, while executing an exclusion process of excluding at least one candidate component from the candidate component groups. When searching for p or fewer candidate components (where p < n) that constitute a desired composition from a candidate component group that is a set of n candidate components, the information processing system 10 obtains a provisional solution considering all of the n candidate components in the first optimization. This can be said to be a process of calculating a relaxed problem in which the constraints regarding the feature vector are relaxed and searching for a relaxed solution. That the exclusion process excludes some candidate components and fixes the corresponding some feature amounts to zero means that constraints are added. That is, repeating the optimization while executing the exclusion process can be said to be a process of searching for an optimal solution while gradually increasing the constraints.

[0038] [Effect] As described above, the information processing system according to one aspect of the present disclosure includes at least one processor. The at least one processor repeats the optimization using a regression model that calculates the physical property values of the composition from the feature vectors indicating the respective feature amounts of the candidate component groups based on a plurality of candidate components, while executing an exclusion process of excluding at least one candidate component from the candidate component groups, and specifies the respective feature amounts of one or more candidate components remaining after the repetition. The exclusion process includes a step of selecting a feature vector corresponding to the solution of the physical property value obtained by the optimization, and a step of excluding at least one candidate component from the candidate component group based on the selected feature vector.

[0039] An information processing method relating to one aspect of this disclosure is performed by an information processing system comprising at least one processor. This information processing method includes the steps of: repeatedly performing an optimization using a regression model that calculates physical property values ​​of a composition from feature vectors representing the respective feature quantities of a group of candidate components based on a plurality of candidate components, while performing an exclusion process to exclude at least one candidate component from the group of candidate components; and identifying the respective feature quantities of the one or more candidate components that remain after the repeated steps. The exclusion process includes the steps of selecting a feature vector corresponding to the solution of physical property values ​​obtained by optimization; and excluding at least one candidate component from the group of candidate components based on the selected feature vector.

[0040] An information processing program relating to one aspect of this disclosure causes a computer to perform the following steps: an optimization using a regression model that calculates the physical properties of a composition from feature vectors representing the features of each of a group of candidate components based on multiple candidate components, while repeatedly performing an exclusion process to exclude at least one candidate component from the group of candidate components; and a step of identifying the features of each of the one or more candidate components that remain after the repeated steps. The exclusion process includes the steps of selecting a feature vector corresponding to the solution of the physical properties obtained by the optimization, and excluding at least one candidate component from the group of candidate components based on the selected feature vector.

[0041] In this respect, optimization using a regression model that calculates the physical properties of a composition from the feature vectors of candidate components is repeatedly performed while reducing the number of candidate components. By performing optimization step by step while narrowing down the candidate components, the feature quantities of candidate components that constitute the desired composition can be searched for at high speed. For example, compared to conventional methods such as metaheuristic methods like Ant Colony Optimization (ACO) and random search, the information processing system according to this disclosure can search for features at a faster speed.

[0042] In information processing systems relating to other aspects, optimization may include the steps of inputting each of a plurality of feature vectors into a regression model to calculate the physical properties of the composition as estimated physical properties, and obtaining at least one of the plurality of estimated physical properties as a solution. Since the solution is obtained from multiple estimated physical properties obtained from multiple feature vectors, the accuracy of optimization can be improved.

[0043] In information processing systems relating to other aspects, at least one processor may acquire a reference physical property value to serve as the basis for optimization, and obtain at least one estimated physical property value corresponding to the reference physical property value from among a plurality of estimated physical properties as the solution. By using the estimated physical property value corresponding to the reference physical property value as the solution to the optimization, the feature quantities of candidate components for constituting the desired composition can be searched more accurately.

[0044] In information processing systems relating to other aspects, at least one processor may terminate the optimization iteration based on the number of types of candidate components remaining after the exclusion process. By setting the termination condition for the iteration based on the number of types of candidate components remaining, it is possible to search for feature quantities of candidate components while considering the preparation of candidate components to obtain the desired composition, the production process of the composition, etc.

[0045] In information processing systems relating to other aspects, at least one processor may exclude candidate components whose feature quantities are below a threshold during the exclusion step. Using this threshold allows for efficient exclusion of candidate components.

[0046] In information processing systems relating to other aspects, at least one processor may exclude the candidate component with the smallest feature quantity during the exclusion step. By excluding the candidate component with the smallest feature quantity, the exclusion process can be reliably performed.

[0047] In information processing systems relating to other aspects, the feature quantities may represent the amounts of candidate components to constitute the composition. By setting the feature quantities in this way, it is possible to search for the amounts of each candidate component that should be prepared to constitute the desired composition.

[0048] [Differentiation] The embodiments described above have been explained in detail. However, the disclosure is not limited to the embodiments described above. The disclosure can be modified in various ways without departing from its essence.

[0049] For example, the information processing methods performed by at least one processor are not limited to the examples above. For instance, some of the steps or processes described above may be omitted, or each step may be performed in a different order. Furthermore, any two or more of the steps described above may be combined, or some of the steps may be modified or deleted. Alternatively, other steps may be performed in addition to each of the steps described above.

[0050] When comparing the relative magnitudes of two numbers within the information processing system 10, either of the two criteria, "greater than or equal to" and "greater than," may be used, or either of the two criteria, "less than or equal to" and "less than," may be used. The choice of such criteria does not change the technical significance of the process of comparing the relative magnitudes of two numbers.

[0051] In this disclosure, the expression "at least one processor performs a first process, a second process, ... and the nth process," or a corresponding expression, refers to a concept that includes cases where the processor performing the n processes from the first to the nth process changes along the way. That is, this expression refers to a concept that includes both cases where all n processes are performed by the same processor and cases where the processor changes at an arbitrary rate for the n processes. [Explanation of Symbols]

[0052] 10... Information processing system, 11... Acquisition unit, 12... Optimization unit, 13... Exclusion unit, 14... Judgment unit, 15... Output unit, 21... Candidate component database.

Claims

1. Equipped with at least one processor, The at least one processor, The optimization process using a regression model, which calculates the physical properties of a composition from feature vectors representing the features of each of the candidate component groups based on multiple candidate components, is repeated while performing an exclusion process to remove at least one candidate component from the candidate component group. After the above iteration, identify the feature quantities of each of the one or more remaining candidate components. The exclusion process described above The steps include selecting the feature vector corresponding to the solution of the physical property values ​​obtained by the optimization, A step of excluding at least one candidate component from the candidate component group based on the selected feature vector, including, Information processing system.

2. The aforementioned optimization, For each of the multiple feature vectors, the step of inputting the feature vector into the regression model to calculate the physical properties of the composition as estimated physical properties, A step of obtaining at least one of the multiple estimated physical properties as a solution, including, The information processing system according to claim 1.

3. The at least one processor, In the optimization described above, a reference physical property value is obtained, Of the plurality of estimated physical properties, at least one estimated physical property corresponding to the reference physical property is obtained as the solution. The information processing system according to claim 2.

4. The at least one processor terminates the optimization iteration based on the number of types of candidate components remaining after the exclusion process. The information processing system according to any one of claims 1 to 3.

5. The at least one processor excludes candidate components whose feature quantities are below a threshold in the exclusion step. The information processing system according to any one of claims 1 to 4.

6. The at least one processor, in the exclusion step, excludes the candidate component whose feature quantity is the smallest. The information processing system according to any one of claims 1 to 5.

7. The aforementioned characteristic quantity represents the amount of the candidate component to be incorporated into the composition. An information processing system according to any one of claims 1 to 6.

8. An information processing method performed by an information processing system comprising at least one processor, The process involves repeatedly performing an exclusion process to remove at least one candidate component from the candidate component group, using a regression model to calculate the physical properties of a composition from feature vectors representing the features of each of the candidate components based on multiple candidate components. The steps include identifying the feature quantities of each of the one or more candidate components remaining after the repeated steps described above, Includes, The exclusion process described above The steps include selecting the feature vector corresponding to the solution of the physical property values ​​obtained by the optimization, A step of excluding at least one candidate component from the candidate component group based on the selected feature vector, including, Information processing methods.

9. The process involves repeatedly performing an exclusion process to remove at least one candidate component from the candidate component group, using a regression model to calculate the physical properties of a composition from feature vectors representing the features of each of the candidate components based on multiple candidate components. The steps include identifying the feature quantities of each of the one or more candidate components remaining after the repeated steps described above, Have the computer run it, The exclusion process described above The steps include selecting the feature vector corresponding to the solution of the physical property values ​​obtained by the optimization, A step of excluding at least one candidate component from the candidate component group based on the selected feature vector, including, Information processing program.

Images