Program, condition search device, and condition search method

Abstract

The present invention causes a computer to function as: a condition storage unit for storing one or more conditions represented by coordinates in a multidimensional space; a condition candidate generation unit for generating a plurality of condition candidates represented by coordinates in the multidimensional space; a condition candidate extraction unit for calculating, for the coordinates of each condition candidate, the shortest distance to the coordinates of the one or more conditions, and extracting, as coordinates of a new condition, the coordinates of the condition candidate for which the shortest distance is greatest; and a new condition addition unit for adding the extracted coordinates of the new condition to the coordinates of the one or more conditions. The condition candidate extraction unit repeats the process of calculating, for the coordinates of each of the plurality of condition candidates, the shortest distance to the coordinates of the one or more conditions, and extracting, as coordinates of a new condition, the coordinates of the condition candidate for which the shortest distance is greatest, until a termination condition is satisfied.

Description

Program, condition search device, and condition search method

[0001] The present disclosure relates to a program, a condition search device, and a condition search method.

[0002] For example, optimization of material production conditions is an example of a multidimensional combinatorial problem. In a multidimensional combinatorial problem, the number of possible combinations increases exponentially with the number of conditions (for example, in the case of a multi-component alloy composition ratio, the number of elements that can be synthesized). One known method for optimizing a multidimensional combinatorial problem is to narrow down candidate experimental conditions by learning the correlation between experimental conditions and properties to be optimized based on past experimental results and predicting the properties that can be obtained under any conditions (see, for example, Patent Document 1).

[0003] International Publication No. WO2019 / 181313

[0004] For example, when extracting new experimental conditions based on past experimental results, it is expected that extracting experimental conditions at distant coordinates from previously tested experimental conditions expressed in spatial coordinates will lead to an efficient search for new experimental conditions.

[0005] An object of the present disclosure is to provide a program, a condition search device, and a condition search method that can efficiently search for new conditions from one or more conditions expressed by coordinates in a multidimensional space.

[0006] The present disclosure has the following configuration.

[0007] [1] A program causing a computer to function as: a condition storage unit that stores one or more conditions expressed by coordinates in a multidimensional space; a condition candidate generation unit that generates a plurality of condition candidates expressed by coordinates in the multidimensional space; a condition candidate extraction unit that calculates, for each coordinate of the condition candidate, the shortest distance to the coordinates of the one or more conditions, and extracts the coordinate of the condition candidate with the longest shortest distance as the coordinate of a new condition; and a new condition addition unit that adds the extracted coordinate of the new condition to the coordinates of the one or more conditions, wherein the condition candidate extraction unit repeats the process of calculating, for each coordinate of the plurality of condition candidates, the shortest distance to the coordinate of the one or more conditions to which the new condition has been added, and extracting the coordinate of the condition candidate with the longest shortest distance as the coordinate of a new condition, until a termination condition is satisfied.

[0008] [2] The program according to [1], further causing the computer to function as a threshold setting receiving unit that receives a threshold setting from a user, and the condition candidate extraction unit determines whether the termination condition is satisfied based on a comparison result between the longest shortest distance and the threshold.

[0009] [3] The program according to [2], characterized in that the condition candidate extraction unit determines that the termination condition is satisfied when both the shortest distance of the condition candidate extracted the Nth time and the shortest distance of the condition candidate extracted the N-1th time are equal to or less than a threshold, and the shortest distance of the condition candidate extracted the Nth time and the shortest distance of the condition candidate extracted the N-1th time are different.

[0010] [4] The program according to any one of [1] to [3], wherein the condition candidate generation unit generates the plurality of condition candidates using random numbers, lattice points, or experimental design.

[0011] [5] The program according to any one of [1] to [4], wherein the condition candidate extraction unit extracts coordinates of the new condition so that coordinates of the one or more conditions and coordinates of the new condition are scattered in the multidimensional space.

[0012] [6] The program according to any one of [1] to [5], wherein the conditions are experimental conditions for material composition.

[0013] [7] A condition search device comprising: a condition storage unit that stores one or more conditions expressed by coordinates in a multidimensional space; a condition candidate generation unit that generates a plurality of condition candidates expressed by coordinates in the multidimensional space; a condition candidate extraction unit that calculates, for each coordinate of the condition candidate, the shortest distance to the coordinates of the one or more conditions, and extracts the coordinate of the condition candidate with the longest shortest distance as the coordinate of a new condition; and a new condition addition unit that adds the extracted coordinate of the new condition to the coordinate of the one or more conditions, wherein the condition candidate extraction unit repeats the process of calculating, for each coordinate of the plurality of condition candidates, the shortest distance to the coordinate of the one or more conditions to which the new condition has been added, and extracting the coordinate of the condition candidate with the longest shortest distance as the coordinate of a new condition, until a termination condition is satisfied.

[0014] [8] A condition search method comprising: causing a computer to execute the following steps: a storage step of storing one or more conditions expressed by coordinates in a multidimensional space; a generation step of generating a plurality of condition candidates expressed by coordinates in the multidimensional space; an extraction step of calculating, for each coordinate of the condition candidate, the shortest distance to the coordinates of the one or more conditions, and extracting the coordinate of the condition candidate with the longest shortest distance as the coordinate of a new condition; and an addition step of adding the extracted coordinate of the new condition to the coordinate of the one or more conditions, and repeating the extraction step of calculating, for each coordinate of the plurality of condition candidates, the shortest distance to the coordinate of the one or more conditions to which the new condition has been added, and extracting the coordinate of the condition candidate with the longest shortest distance as the coordinate of a new condition, and the addition step of adding the extracted coordinate of the new condition to the coordinate of the one or more conditions, until a termination condition is satisfied.

[0015] According to the present disclosure, new conditions can be efficiently searched for from one or more conditions expressed by coordinates in a multidimensional space.

[0016] 1 is a configuration diagram of an example of an information processing system according to the present embodiment. FIG. 1 is a hardware configuration diagram of an example of a computer according to the present embodiment. FIG. 2 is a functional configuration diagram of an example of an information processing system according to the present embodiment. FIG. 2 is a diagram illustrating an example of prerequisites for searching for new experimental conditions according to the present embodiment. FIG. 3 is a diagram illustrating an example of prerequisites for searching for new experimental conditions according to the present embodiment. FIG. 4 is a diagram illustrating an example of an outline of the process of searching for new experimental conditions according to the present embodiment. FIG. 5 is a diagram illustrating an example of an outline of the process of searching for new experimental conditions according to the present embodiment. FIG. 6 is a flowchart of an example of a condition search process of the information processing system according to the present embodiment. FIG. 7 is a diagram illustrating an example of normalized existing experimental conditions. FIG. 8 is a diagram illustrating an example of normalized existing experimental conditions. FIG. 9 is a diagram illustrating an example of a plurality of experimental condition candidates using lattice points. FIG. 10 is a diagram illustrating an example of a plurality of experimental condition candidates using lattice points. FIG. 11 is a diagram illustrating an example of an experimental condition candidate with the longest nearest neighbor distance. FIG. 12 is a diagram illustrating an example of an experimental condition candidate with the longest nearest neighbor distance. FIG. 13 is a diagram illustrating an example of an experimental condition candidate with the longest nearest neighbor distance extracted third. FIG. 14 is a diagram illustrating an example of an experimental condition candidate with the longest nearest neighbor distance extracted third. FIG. 15 is a diagram illustrating an example of an experimental condition candidate with the longest nearest neighbor distance extracted fourth. 10 is a diagram illustrating an example of a candidate experimental condition for the longest nearest neighbor distance extracted fourth. FIG. 11 is a diagram illustrating an example of a candidate experimental condition for the longest nearest neighbor distance extracted tenth. FIG. 12 is a diagram illustrating an example of a candidate experimental condition for the longest nearest neighbor distance extracted tenth. FIG. 13 is a diagram illustrating an example of a candidate experimental condition for the longest nearest neighbor distance extracted tenth. FIG. 14 is a diagram illustrating a change in the longest nearest neighbor distance. FIG. 15 is a diagram illustrating a change in the longest nearest neighbor distance. FIG. 16 is a diagram illustrating an example of a data distribution for an experimental condition to which the seventh longest nearest neighbor distance was added. FIG. 17 is a diagram illustrating an example of a data distribution for an experimental condition to which the eighth longest nearest neighbor distance was added. FIG. 18 is a diagram illustrating an example of a data distribution for an experimental condition to which the ninth longest nearest neighbor distance was added. FIG. 19 is a diagram illustrating an example of a data distribution for an experimental condition to which the tenth longest nearest neighbor distance was added.1 is a diagram showing an example of the relationship between the longest nearest neighbor distance and data density; FIG. 2 is a diagram showing an example of the relationship between the longest nearest neighbor distance and data density; FIG. 3 is a diagram showing an example of the relationship between the longest nearest neighbor distance and data density; FIG. 4 is a diagram showing an example of the relationship between the longest nearest neighbor distance and data density;

[0017] Next, an embodiment of the present invention will be described in detail. However, the present invention is not limited to the following embodiment. In this embodiment, an experimental condition for a material composition expressed by coordinates in a two-dimensional space, which is an example of a multidimensional space, will be described as an example of one or more conditions expressed by coordinates in a multidimensional space.

[0018] [First embodiment] <System configuration> Fig. 1 is a configuration diagram of an example of an information processing system according to this embodiment. The information processing system 1 in Fig. 1 includes a condition search device 10 and a user terminal 12. The condition search device 10 and the user terminal 12 are connected to each other so as to be able to communicate data via a communication network 18 such as a local area network (LAN) or the Internet.

[0019] The user terminal 12 is an information processing terminal such as a PC, tablet terminal, or smartphone operated by a user. The user terminal 12 accepts input of information necessary for searching for experimental conditions for material compositions from the user and causes the condition searching device 10 to search for experimental conditions for material compositions. The user terminal 12 also receives information such as the experimental conditions for material compositions searched for by the condition searching device 10 and displays it on a display device, for example, for the user to confirm.

[0020] The condition search device 10 is an information processing device such as a PC or workstation that searches for experimental conditions for material composition. The condition search device 10 searches for new experimental conditions from existing experimental conditions using existing experimental data, as described below. The condition search device 10 transmits information such as the searched experimental conditions to a user terminal 12.

[0021] The information processing system 1 in Figure 1 is merely an example, and it goes without saying that there are various system configuration examples depending on the application and purpose. For example, the condition search device 10 may be realized by multiple computers, or may be realized as a cloud computing service. Furthermore, the information processing system 1 in Figure 1 may be realized by a standalone computer.

[0022] <Hardware Configuration> The condition search device 10 and the user terminal 12 in FIG. 1 are realized by, for example, a computer 500 having the hardware configuration shown in FIG.

[0023] Fig. 2 is a diagram showing an example of the hardware configuration of a computer according to this embodiment. The computer 500 in Fig. 2 includes an input device 501, a display device 502, an external I / F 503, a RAM 504, a ROM 505, a CPU 506, a communication I / F 507, and an HDD 508, all of which are interconnected by a bus B. The input device 501 and the display device 502 may be connected to each other for use.

[0024] The input device 501 includes a touch panel, operation keys, buttons, a keyboard, a mouse, etc., which are used by the user to input various signals. The display device 502 includes a display such as a liquid crystal or organic EL display for displaying a screen, a speaker for outputting audio data such as voice and sound, etc. The communication I / F 507 is an interface for the computer 500 to perform data communication.

[0025] The HDD 508 is an example of a non-volatile storage device that stores programs and data. The stored programs and data include an OS, which is basic software that controls the entire computer 500, and applications that provide various functions on the OS. Note that the computer 500 may use a drive device that uses flash memory as a storage medium (e.g., a solid-state drive (SSD)) instead of the HDD 508.

[0026] The external I / F 503 is an interface with an external device. The external device may be a recording medium 503a. This allows the computer 500 to read and / or write data from and to the recording medium 503a via the external I / F 503. The recording medium 503a may be a flexible disk, a CD, a DVD, an SD memory card, a USB memory, or the like.

[0027] The ROM 505 is an example of a non-volatile semiconductor memory (storage device) that can retain programs and data even when the power is turned off. The ROM 505 stores programs and data such as the BIOS, OS settings, and network settings that are executed when the computer 500 starts up. The RAM 504 is an example of a volatile semiconductor memory (storage device) that temporarily retains programs and data.

[0028] The CPU 506 is a computing device that reads programs and data from storage devices such as the ROM 505 and HDD 508 onto the RAM 504 and executes processing to realize overall control and functions of the computer 500. By executing programs, the computer 500 according to this embodiment can realize various functions of the condition search device 10 and the user terminal 12, which will be described later.

[0029] <Functional Configuration> The configuration of the information processing system 1 according to this embodiment will be described. Fig. 3 is a functional configuration diagram of an example of the information processing system according to this embodiment. Note that in the configuration diagram of Fig. 3, portions that are not necessary for explaining this embodiment are omitted as appropriate.

[0030] 3 includes a condition candidate generating unit 30, a condition candidate extracting unit 32, a new condition adding unit 34, a threshold setting receiving unit 36, an output unit 38, an experimental condition storing unit 40, and a threshold storing unit 42. The user terminal 12 includes an information display unit 20, an operation receiving unit 22, a request sending unit 24, and a response receiving unit 26.

[0031] The information display unit 20 of the user terminal 12 displays information for the user to confirm on the display device 502. The operation accepting unit 22 accepts various operations from the user, such as input of information necessary for searching for experimental conditions. The request sending unit 24 sends a request for processing, such as searching for experimental conditions, to the condition searching device 10. In addition, the response receiving unit 26 receives a response to the request for processing, such as searching for experimental conditions, sent by the request sending unit 24.

[0032] The experimental condition storage unit 40 of the condition search device 10 stores existing experimental data that has been tested, including information on existing experimental conditions as described below. The experimental condition storage unit 40 stores information on new experimental conditions as described below. The experimental condition storage unit 40 stores existing and new experimental conditions for material compositions expressed in coordinates in two-dimensional space as described below. In addition, the threshold storage unit 42 stores thresholds set by the user.

[0033] As described below, the condition candidate generation unit 30 generates a plurality of condition candidates represented by coordinates in two-dimensional space. As described below, the condition candidate extraction unit 32 calculates the shortest distance (nearest neighbor distance) between the coordinates of each of the generated plurality of condition candidates and the coordinates of the existing and new experimental conditions. As described below, the condition candidate extraction unit 32 extracts the coordinates of the condition candidate with the longest nearest neighbor distance as the coordinates of the new experimental condition. The new condition addition unit 34 stores the extracted coordinates of the new experimental condition in the experimental condition storage unit 40 so as to add them.

[0034] The processing by the condition candidate extraction unit 32 and the new condition addition unit 34 is repeated until the condition candidate extraction unit 32 determines that a termination condition, which will be described later, is satisfied. For example, the condition candidate extraction unit 32 determines whether the termination condition is satisfied, as will be described later, based on the result of comparing the longest nearest neighbor distance with a threshold value.

[0035] The threshold setting receiving unit 36 ​​receives a threshold setting from the user and stores it in the threshold storage unit 42. The output unit 38 transmits the searched experimental conditions as a response to a request for searching experimental conditions for material composition, for example, from the user terminal 12. Note that the configuration diagram in Fig. 3 is an example. Various configurations of the information processing system 1 according to this embodiment are possible.

[0036] <Prerequisites for the Invention> The information processing system 1 according to this embodiment searches for new experimental conditions from existing experimental conditions included in existing experimental data that has already been tested, as shown in Figures 4A, 4B, 5A, and 5B. Figures 4A and 4B are diagrams showing an example of prerequisites for searching for new experimental conditions according to this embodiment. Figures 5A and 5B are diagrams showing an example of an outline of the process for searching for new experimental conditions according to this embodiment.

[0037] 4A, 4B, 5A, and 5B show experimental conditions represented by coordinates x1 and x2 in a two-dimensional space. x1 and x2 are parameters of the experimental conditions. In this embodiment, an example of a two-dimensional space with two types of parameters is described for ease of understanding, but a multidimensional space with two or more types of parameters may be used. Each of the points displayed in FIGS. 4A, 4B, 5A, and 5B represents a different experimental condition according to its coordinates.

[0038] 4A shows an example of a point cloud of existing experimental conditions included in existing experimental data that has already been tested. Here, it is assumed that the experimental conditions represented by the point cloud in FIG. 4A do not contain any experimental conditions that satisfy the required characteristics. It is also assumed that even if a search continues in the area 1000 where the point cloud in FIG. 4A exists, it is unlikely that an experimental condition that satisfies the required characteristics will be found.

[0039] If it is assumed that the possibility of finding experimental conditions that satisfy the required characteristics is low even if the search in the region 1000 is continued, it is necessary to search for experimental conditions in a region 1002 other than the region 1000, as shown in Fig. 4B. Methods for generating the experimental conditions in the region 1002 include a method using experimental design, a method using random numbers, a method using grid points, and the like.

[0040] When searching for experimental conditions in region 1002, it is expected that searching for experimental conditions from points in region 1002 that are far from the point cloud present in region 1000 is likely to lead to more efficient searching. However, as the number of parameters of the experimental conditions increases and the number of dimensions of the multidimensional space that represents the experimental conditions with coordinates increases, it becomes more difficult to extract points in region 1002 that are far from the point cloud present in region 1000.

[0041] 5A and 5B, the point with the longest shortest distance (nearest neighbor distance) to the point cloud of the existing experimental conditions included in the existing experimental data that has been tested is defined as the point farthest from the existing experimental conditions. The nearest neighbor distance is the shortest distance among the distances to each of the point clouds of the existing experimental conditions that have been included in the existing experimental data that has been tested, and is the distance from the point of the new experimental condition to the closest point (nearest neighbor point) of the existing experimental condition.

[0042] Fig. 5A shows a circle 1102 whose radius is the distance from point 1100 of a new experimental condition to the nearest point. Fig. 5B shows a circle 1106 whose radius is the distance from point 1104 of a new experimental condition to the nearest point. In the example of Fig. 5B, point 1104 of the new experimental condition is the point with the longest nearest distance. As shown in Figs. 5A and 5B, in this embodiment, new experimental conditions are extracted that have the largest circle or sphere with the nearest distance as their radius.

[0043] Note that the calculation of the nearest neighbor distance does not necessarily need to cover all of the point clouds of existing experimental conditions; for example, the amount of calculation can be reduced by using an algorithm that adopts points that appear to be close from the point clouds of existing experimental conditions.

[0044] <Processing> Hereinafter, details of the processing performed by the information processing system 1 according to this embodiment to search for new experimental conditions from existing experimental conditions included in existing experimental data on which experiments have been performed will be described.

[0045] 6 is a flowchart showing an example of a condition search process of the information processing system according to this embodiment. In step S10, the condition search device 10 converts the existing experimental conditions stored in the experimental condition storage unit 40 so that they have a maximum value of "1" and a minimum value of "0." Step S10 is a process for normalizing the parameters of the experimental conditions.

[0046] 7A and 7B are diagrams illustrating an example of normalized existing experimental conditions. FIG. 7A shows an example of a point cloud 1200 of normalized existing experimental conditions. The condition search device 10 can calculate the nearest neighbor distance for each of the point cloud 1200. Some of the nearest neighbor distances are indicated by arrows in FIG. 7A. FIG. 7B shows a histogram of the nearest neighbor distances calculated using FIG. 7A.

[0047] In step S12, the condition search device 10 generates multiple experimental condition candidates. For example, the condition search device 10 generates multiple experimental condition candidates using experimental design, random numbers, or lattice points. Here, a process for generating multiple experimental condition candidates using lattice points will be described.

[0048] 8A and 8B are diagrams illustrating an example of multiple experimental condition candidates using grid points. In Fig. 8A, grid points 1202 are generated for the experimental condition parameters x1 and x2 from 0 to 1.0 (in increments of 0.1). Each grid point 1202 represents multiple experimental condition candidates.

[0049] In step S14, the condition searching device 10 calculates the shortest distance (nearest neighbor distance) between each of the multiple experimental condition candidates represented by the lattice points 1202 in Fig. 8A and the point cloud 1200 of the existing experimental conditions. Step S14 is the first calculation of the nearest neighbor distance.

[0050] In step S16, the condition searching device 10 extracts the candidate experimental condition with the longest first nearest neighbor distance calculated in step S14. Figures 9A and 9B are diagrams illustrating an example of the candidate experimental condition with the longest nearest neighbor distance.

[0051] Fig. 9A shows an example in which, of the nearest neighbor distances calculated for the multiple experimental condition candidates represented by lattice points 1202 in Fig. 8A, the nearest neighbor distance of the experimental condition candidate represented by point 1204 is the longest. In Fig. 9A, the nearest neighbor distance of the experimental condition candidate represented by point 1204 is indicated by an arrow. Fig. 9B is an example of a histogram to which the nearest neighbor distance of point 1204 shown in Fig. 9A has been added. Arrow 1205 in Fig. 9B indicates the nearest neighbor distance of point 1204.

[0052] In step S18, the condition search device 10 accepts the setting of a threshold value based on the first nearest neighbor distance. The threshold value in step S18 may be set by the user while checking the histogram in FIG. 9B, or may be set automatically using a table in which nearest neighbor distances correspond to threshold values. The threshold value is used to determine whether the termination condition in step S24 is satisfied.

[0053] In step S20, the condition searching device 10 calculates the shortest distance (nearest neighbor distance) between the point group of existing experimental conditions and the point of the experimental information candidate with the longest first-time nearest neighbor distance for each experimental condition candidate, excluding point 1204 of the experimental information candidate with the longest first-time nearest neighbor distance, among the multiple experimental condition candidates represented by lattice points 1202 in Fig. 8A. Step S20 is the calculation of the nearest neighbor distance from the second time onwards.

[0054] In step S22, the condition searching device 10 extracts the candidate experimental condition with the longest nearest neighbor distance calculated in step S20 from the second time onward. Figures 10A and 10B are diagrams illustrating an example of the candidate experimental condition with the longest nearest neighbor distance.

[0055] Fig. 10A shows an example in which the nearest neighbor distance of the experimental condition candidate represented by point 1206 was the longest in step S22. In Fig. 10A, the nearest neighbor distance of the experimental condition candidate represented by point 1206 and the nearest neighbor distance of the experimental condition candidate represented by point 1204 are indicated by arrows. Fig. 10B is an example of a histogram to which the nearest neighbor distances of points 1204 and 1206 shown in Fig. 10A have been added. Arrow 1205 in Fig. 10B indicates the nearest neighbor distance of point 1204. Arrow 1207 in Fig. 10B indicates the nearest neighbor distance of point 1206. In the histogram in Fig. 10B, the nearest neighbor distance of point 1204 indicated by arrow 1205 has become smaller than that in Fig. 9B due to the second calculation of the nearest neighbor distance.

[0056] 8B and 9B, the shapes of the histograms change as the processes of steps S20 and S22 are repeated. In step S24, the condition search device 10 uses the threshold set in step S18 to determine whether the termination condition is met.

[0057] For example, the condition search device 10 determines whether the termination condition is met based on the comparison result between the longest nearest neighbor distance extracted in step S22 and the threshold value set in step S18. The condition search device 10 repeats the processes of steps S20 to S26 until it determines that the termination condition is met.

[0058] An example of the termination condition is whether the longest nearest neighbor distance extracted in step S22 is below a threshold value. If the longest nearest neighbor distance extracted in step S22 is below the threshold value, the condition search device 10 determines that the termination condition is satisfied.

[0059] Another example of a termination condition is whether or not the longest nearest neighbor distance extracted Nth and the longest nearest neighbor distance extracted N-1 are both below a threshold, and whether or not the longest nearest neighbor distance extracted Nth and the longest nearest neighbor distance extracted N-1 are different.

[0060] If both the longest nearest neighbor distance extracted in the Nth place and the longest nearest neighbor distance extracted in the N-1th place are below a threshold value, and the longest nearest neighbor distance extracted in the Nth place and the longest nearest neighbor distance extracted in the N-1th place are different, the condition search device 10 determines that the termination condition is satisfied.

[0061] If the termination condition is not satisfied, the condition search device 10 proceeds to the processing of step S26. The condition search device 10 adds the (N-1)th extracted experimental condition candidate with the longest nearest neighbor distance to the experimental condition group, then adds "1" to N, and returns to the processing of step S20.

[0062] 11A and 11B are diagrams illustrating an example of the third extracted candidate experimental condition for the longest nearest neighbor distance. Fig. 11A shows point 1208, which represents the third extracted candidate experimental condition for the longest nearest neighbor distance. Fig. 11B is an example of a histogram to which the nearest neighbor distances of points 1204 to 1208 shown in Fig. 11A have been added.

[0063] 12A and 12B are diagrams illustrating an example of the fourth extracted candidate experimental condition for the longest nearest neighbor distance. FIG. 12A shows point 1210, which represents the fourth extracted candidate experimental condition for the longest nearest neighbor distance. FIG. 12B is an example of a histogram to which the nearest neighbor distances of points 1204 to 1210 shown in FIG. 12A have been added. The shape of the histogram in FIG. 12B has changed due to the fourth calculation of the nearest neighbor distance.

[0064] Assume that the processing of steps S20 to S26 in Fig. 6 is repeated to extract up to the tenth longest nearest neighbor distance. Figs. 13A to 13C are diagrams illustrating an example of a candidate experimental condition for the tenth longest nearest neighbor distance extracted. Fig. 13A shows point 1222, which represents the candidate experimental condition for the tenth longest nearest neighbor distance extracted. Fig. 13B shows a histogram of nearest neighbor distances calculated using Fig. 13A.

[0065] FIG. 13C is a diagram showing the change in the longest nearest neighbor distance extracted first through tenth. The first through tenth points from the left on the horizontal axis represent the longest nearest neighbor distance extracted first through tenth. The eleventh and subsequent points from the left on the horizontal axis represent the nearest neighbor distances of existing experimental conditions included in existing experimental data that has already been tested. As shown in FIG. 13C, the longest nearest neighbor distance becomes smaller by repeating the processing of steps S20 through S26, and the termination condition is met. When the termination condition is met, the condition search device 10 terminates the processing of the flowchart in FIG. 6.

[0066] The process of determining whether the termination condition of step S24 is satisfied will be further described. Figures 14A and 14B are diagrams showing changes in the longest nearest neighbor distance. Figure 14A is a diagram showing changes in the longest nearest neighbor distance extracted from the 1st to 40th points. The 1st to 40th points from the left on the horizontal axis represent the longest nearest neighbor distance extracted from the 1st to 40th points. The 41st and subsequent points from the left on the horizontal axis represent the nearest neighbor distances of existing experimental conditions included in existing experimental data that has already been tested. Figure 14B also shows the values ​​of the longest nearest neighbor distance extracted from the 1st to 12th points.

[0067] For example, if the longest nearest neighbor distance extracted in step S22 falls below a threshold, the condition search device 10 determines that the termination condition has been met. If the threshold is "0.4", the seventh longest nearest neighbor distance determines that the termination condition has been met.

[0068] Furthermore, if both the longest nearest neighbor distance extracted the Nth time and the longest nearest neighbor distance extracted the N-1th time are below the threshold value, and the longest nearest neighbor distance extracted the Nth time and the longest nearest neighbor distance extracted the N-1th time are different, the condition search device 10 determines that the termination condition is met. If the threshold value is "0.4", the condition search device 10 determines that the termination condition is met based on the longest nearest neighbor distance of the 11th time.

[0069] For example, the condition search device 10 determines that the termination condition is not met because the longest nearest neighbor distance for the sixth search, "0.483," is greater than the threshold value. Also, the condition search device 10 determines that the termination condition is not met because the longest nearest neighbor distance for the seventh search, "0.392," is smaller than the threshold value, but the longest nearest neighbor distances for the eighth to tenth searches are the same as the longest nearest neighbor distance for the seventh search, "0.392."

[0070] Then, the condition search device 10 determines that the termination condition is met because the 11th longest nearest neighbor distance, "0.370", is smaller than the threshold value and the 10th longest nearest neighbor distance, "0.392", is different from the 11th longest nearest neighbor distance, "0.370".

[0071] If both the longest nearest neighbor distance extracted in the Nth position and the longest nearest neighbor distance extracted in the N-1th position are below the threshold, and the longest nearest neighbor distance extracted in the Nth position and the longest nearest neighbor distance extracted in the N-1th position are different, the condition search device 10 determines that the termination condition is satisfied. The termination condition is for obtaining a data distribution of experimental conditions that meet the threshold, as follows:

[0072] 15A to 15D are diagrams showing examples of data distributions for experimental conditions in which the longest nearest neighbor distances for the 7th to 10th runs have been added. FIG. 15A shows the data distribution for the experimental condition in which the longest nearest neighbor distance for the 7th run is "0.392." FIGS. 15B to 15D show the data distribution for the experimental condition in which the longest nearest neighbor distance for the 8th to 10th runs is "0.392." Points 1216 to 1222 in FIGS. 15A to 15D represent the experimental condition candidates extracted for the 7th to 10th runs.

[0073] If it is determined that the termination condition is met when the longest nearest neighbor distance for the seventh time shown in Figure 15A is "0.392," there is a possibility that the threshold value "0.4" and the longest nearest neighbor distance "0.483" of the data distribution of the extracted experimental conditions will not match, even though the points for the experimental conditions for the eighth to tenth times shown in Figures 15B to 15D with the longest nearest neighbor distance of "0.392" remain.

[0074] Therefore, by utilizing a termination condition in which the condition search device 10 determines that the termination condition has been met when both the longest nearest neighbor distance extracted in the Nth position and the longest nearest neighbor distance extracted in the N-1th position are below the threshold, and when the longest nearest neighbor distance extracted in the Nth position and the longest nearest neighbor distance extracted in the N-1th position are different, the condition search device 10 of this embodiment can search for a data distribution of experimental conditions that meets the set threshold.

[0075] For example, the relationship between the longest nearest neighbor distance and data density is as shown in Figures 16A to 16D. Figures 16A to 16D are diagrams showing an example of the relationship between the longest nearest neighbor distance and data density. Figure 16A is an example of a point cloud representing existing experimental conditions contained in existing experimental data that has already been tested. Regardless of the distribution of the point cloud in Figure 16A, the data distribution of the point cloud representing new experimental conditions can be estimated from the longest nearest neighbor distance. For example, if the point cloud representing the existing experimental conditions is spread out, the longest nearest neighbor distance will be small. If the point cloud representing the existing experimental conditions is not spread out, the longest nearest neighbor distance will be large.

[0076] Therefore, the condition search device 10 according to this embodiment sets a small threshold value if it is desired to increase the data density in the data distribution of new experimental conditions, and sets a large threshold value if it is acceptable for the data density in the data distribution of new experimental conditions to be low.

[0077] According to the condition search device 10 of this embodiment, by extracting the point (coordinate) with the longest nearest neighbor distance from the point cloud representing the existing experimental conditions as the point representing the new experimental conditions, it is possible to continuously extract new experimental conditions that are significantly different from the existing experimental conditions. For example, even if there are existing experimental conditions that have already been searched, the condition search device 10 of this embodiment can easily search for new experimental conditions in an unsearched search space from the search space of experimental conditions represented by coordinates in a multidimensional space.

[0078] [Other Embodiments] The experimental conditions for material composition searched for by the condition search device 10 according to this embodiment may be used as input data for an experimental device that performs material composition based on the experimental conditions for material composition. Furthermore, the condition search device 10 according to this embodiment may acquire an evaluation result of whether the required characteristics are satisfied from an evaluation device that evaluates the required characteristics of the material composed by the experimental device, and may use the evaluation result as existing experimental data containing information on the existing experimental conditions to repeatedly search for experimental conditions for material composition.

[0079] For example, in terms of industrial applicability, the condition searching device 10 according to the present embodiment can be used in various fields related to combinatorial optimization, such as resin blend composition searching. As a more specific application example, for example, it can be used to search for the optimal combination of resin types for a flexible transparent film.

[0080] As described above, the information processing system 1 according to this embodiment can provide a program, a condition search device, and a condition search method that can efficiently search for new conditions from one or more conditions expressed by coordinates in a multidimensional space.

[0081] Although the present embodiment has been described above, it will be understood that various changes in form and details are possible without departing from the spirit and scope of the claims. While the present invention has been described above based on examples, the present invention is not limited to the above examples and various modifications are possible within the scope of the claims. This application claims priority from basic application No. 2021-153879 filed with the Japan Patent Office on September 22, 2021, the entire contents of which are incorporated herein by reference.

[0082] REFERENCE SIGNS LIST 1 Information processing system 10 Condition search device 12 User terminal 18 Communication network 20 Information display unit 22 Operation reception unit 24 Request transmission unit 26 Response reception unit 30 Condition candidate generation unit 32 Condition candidate extraction unit 34 New condition addition unit 36 ​​Threshold setting reception unit 38 Output unit 40 Experimental condition storage unit 42 Threshold storage unit

Claims

1. Computer, a condition storage unit that stores one or more conditions expressed by coordinates in a multidimensional space; a condition candidate generation unit that generates a plurality of condition candidates represented by coordinates in the multidimensional space; a condition candidate extraction unit that calculates the shortest distance between each of the coordinates of the condition candidate and the coordinates of the one or more conditions, and extracts the coordinates of the condition candidate with the longest shortest distance as the coordinates of a new condition; a new condition adding unit that adds the extracted coordinates of the new condition to the coordinates of the one or more conditions; It functions as The condition candidate extraction unit repeats a process of calculating the shortest distance between each of the coordinates of the plurality of condition candidates and the coordinates of the one or more conditions to which the new condition has been added, and extracting the coordinates of the condition candidate with the longest shortest distance as the coordinates of the new condition, until an end condition is satisfied. A program characterized by.

2. The computer further comprises: a threshold setting receiving unit that receives a threshold setting from a user; The condition candidate extraction unit determines whether the termination condition is satisfied based on a result of comparing the longest shortest distance with a threshold value. The program according to claim 1,

3. The condition candidate extraction unit determines that the termination condition is satisfied when both the shortest distance of the condition candidate extracted for the Nth time and the shortest distance of the condition candidate extracted for the (N-1)th time are equal to or less than a threshold value, and the shortest distance of the condition candidate extracted for the Nth time and the shortest distance of the condition candidate extracted for the (N-1)th time are different.

3. The program according to claim 2,

4. The condition candidate generation unit generates the plurality of condition candidates using random numbers, lattice points, or experimental design.

4. The program according to claim 1, wherein:

5. The condition candidate extraction unit extracts coordinates of the new condition such that coordinates of the one or more conditions and coordinates of the new condition are scattered in the multidimensional space.

4. The program according to claim 1, wherein:

6. The conditions are experimental conditions for the material composition.

4. The program according to claim 1, wherein:

7. a condition storage unit that stores one or more conditions expressed by coordinates in a multidimensional space; a condition candidate generation unit that generates a plurality of condition candidates represented by coordinates in the multidimensional space; a condition candidate extraction unit that calculates the shortest distance between each of the coordinates of the condition candidate and the coordinates of the one or more conditions, and extracts the coordinates of the condition candidate with the longest shortest distance as the coordinates of a new condition; a new condition adding unit that adds the coordinates of the extracted new condition to the coordinates of the one or more conditions; and The condition candidate extraction unit repeats a process of calculating the shortest distance between each of the coordinates of the plurality of condition candidates and the coordinates of the one or more conditions to which the new condition has been added, and extracting the coordinates of the condition candidate with the longest shortest distance as the coordinates of the new condition, until an end condition is satisfied. A condition search device characterized by the above.

8. a storing step of storing one or more conditions expressed in coordinates in a multidimensional space; a generation step of generating a plurality of condition candidates represented by coordinates in the multidimensional space; an extraction step of calculating the shortest distance between each of the coordinates of the condition candidate and the coordinates of the one or more conditions, and extracting the coordinates of the condition candidate with the longest shortest distance as the coordinates of a new condition; an adding step of adding the coordinates of the extracted new condition to the coordinates of the one or more conditions; on the computer, an extraction step of calculating the shortest distance between the coordinates of the one or more conditions to which the new condition has been added and the coordinates of the condition candidate with the longest shortest distance as the coordinates of the new condition, and an addition step of adding the extracted coordinates of the new condition to the coordinates of the one or more conditions, for each of the coordinates of the plurality of condition candidates, until a termination condition is satisfied; A condition search method characterized by the above.