Information processing system, program, and method

Abstract

To provide a method for effectively solving a scale-up problem.SOLUTION: A method is for improving output in a second system based on a first data set composed of a first parameter group which is adjustable in a first system and a second data set composed of a second parameter group which is adjustable in the second system. The first parameter group and the second parameter group include one or more parameters of different kinds. The method includes: processing for generating a third data set and a fourth data set by supplementing and / or deleting one or more parameters of different kinds from the first data set and the second data set; processing for generating a model which simulates input / output in the second system by using the third data set and the fourth data set; and processing for determining a parameter group for improving the output in the second system by using the third data set, the fourth data set, and the model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to an information processing system, a program, and a method for obtaining a set of parameters for obtaining a suitable output. [Background technology]

[0002] In recent years, machine learning technology has been used in various industrial fields. For example, in manufacturing plants and systems, attempts are being made to set process conditions such as pressure and temperature based on the results of experimental plants and simulations that use machine learning technology.

[0003] Here, optimization techniques can be used to set process conditions. For example, Patent Document 1 describes the use of a constrained optimization technique using Bayesian optimization for industrial product development or manufacturing process development. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2022-185927 Summary of the Invention [Problem to be solved by the invention]

[0005] On the other hand, when a manufacturing plant or system is operated under process conditions derived from an experimental plant or system or simulation results, it is known that there are cases where the expected performance cannot be achieved, that is, a so-called scale-up problem. As a result, even if production conditions are derived using an experimental plant or simulation, trial runs are often repeated on-site through trial and error, which can waste time, manpower, and materials.

[0006] Furthermore, there may be differences between parameters such as pressure and agitation speed that can be controlled in an experimental plant or simulation and parameters that can be controlled in an actual manufacturing plant or system.

[0007] The present disclosure has been made in view of the above, and its purpose is to provide a method for effectively solving the scale-up problem. [Means for solving the problem]

[0008] A method according to an embodiment is a method for improving an output in a second system based on a first dataset consisting of a first group of parameters adjustable in a first system and a second dataset consisting of a second group of parameters adjustable in a second system different from the first system, wherein the first group of parameters and the second group of parameters include one or more parameters of different types, and the method includes a first process of generating a third dataset and a fourth dataset consisting of a third group of parameters by supplementing and / or deleting the one or more parameters of different types from the first dataset and the second dataset, a second process of generating a model that simulates input / output in the second system using the third dataset and the fourth dataset, and a third process of determining a group of parameters for improving the output in the second system using the third dataset, the fourth dataset, and the model.

[0009] A program according to an embodiment is a program for improving an output in a second system based on a first dataset consisting of a first group of parameters that can be adjusted in a first system and a second dataset consisting of a second group of parameters that can be adjusted in a second system different from the first system, wherein the first group of parameters and the second group of parameters include one or more parameters of different types, and the program causes an information processing system to execute a first process of generating a third dataset and a fourth dataset consisting of a third group of parameters by supplementing and / or deleting the one or more parameters of different types from the first dataset and the second dataset; a second process of generating a model that simulates input / output in the second system using the third dataset and the fourth dataset; and a third process of determining the third group of parameters for improving the output in the second system using the third dataset, the fourth dataset, and the model.

[0010] An information processing system according to an embodiment is an information processing system for improving an output in a second system based on a first dataset consisting of a first group of parameters that can be adjusted in a first system and a second dataset consisting of a second group of parameters that can be adjusted in a second system different from the first system, wherein the first group of parameters and the second group of parameters include one or more parameters of different types, and the information processing system has: a first means for generating a third dataset and a fourth dataset consisting of a third group of parameters by supplementing and / or deleting the one or more parameters of different types from the first dataset and the second dataset; a second means for generating a model that simulates input / output in the second system using the third dataset and the fourth dataset; and a third means for determining the third group of parameters for improving the output in the second system using the third dataset, the fourth dataset, and the model. [Effects of the Invention]

[0011] According to the present disclosure, it is possible to provide a method for effectively solving the scale-up problem. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 2 is a diagram for explaining the overall processing of the system according to the embodiment. [Figure 2] FIG. 2 is a block diagram showing a functional configuration of a program according to the embodiment. [Figure 3] 10 is a flowchart illustrating a flow of adjusting a data set according to an embodiment. [Figure 4] FIG. 1 is a block diagram showing a specific example of a configuration of an information processing device according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0014] 1. Overall Concept First, an outline of the processing according to this embodiment will be described with reference to FIGS.

[0015] 1.1 Overview The information processing device (hereinafter also referred to as the "condition determination device") according to this embodiment is a device for determining conditions such as various parameters (e.g., temperature, pressure, stirring speed, flow rate, concentration, etc.) when implementing a manufacturing process to produce a target product. The final manufacturing process is realized in a manufacturing factory or the like where large-scale machinery is used to mass-produce products (hereinafter, a manufacturing factory or other place where a manufacturing process is implemented on a large scale is referred to as a "large-scale system" or "production system"). In such a large-scale system 120, a significant amount of cost is incurred due to repeated trial and error, in which parameters are finely adjusted to find optimal conditions and the results are confirmed. Therefore, in this embodiment, the manufacturing process is implemented while adjusting conditions in a small-scale location, such as a laboratory (hereinafter, a location such as a laboratory where a manufacturing process is implemented on a small scale is referred to as a "small-scale system" or "experimental system"), and the conditions for the large-scale system 120 are determined based on the results of the small-scale system 110. In other words, the conditions used in the large-scale system 120 are determined using the small-scale system 110 for manufacturing process experiments and the large-scale system 120 where the manufacturing process itself is actually implemented in the manufacturing factory. For example, in the case of a chemical manufacturer, the condition determination device may determine conditions such as temperature, pressure, flow rate, and concentration in unit operations such as reaction and separation. It is desirable that such conditions be determined so as to maximize an index showing a specific performance, such as a yield rate. Typically, various conditions are examined in a small-scale system 110 where trial costs are low, and the conditions for a large-scale system 120 such as a manufacturing plant are set based on the results of the examination.

[0016] Here, the set of input and output parameters tried in the small-scale system 110 is included for the number of trials and is called the small-scale system data set D. S The input and output parameter sets tested in the large-scale system 120 are included in the large-scale system data set D L As mentioned above, it is considered that the small-scale system 110 can be implemented with lower trial cost and accuracy than the large-scale system 120. The small-scale system data set D S and large-scale data set D LAmong these, the input (condition) parameters may be, for example, the temperature, pressure, flow rate, etc. of the equipment, and the output (result) parameters may be sales, profit, yield rate, refinement degree, manufacturing time, failure rate, etc. Alternatively, a small-scale data set D S is the actual factory dataset for product A, and the large-scale dataset D L may be a dataset of an actual factory for product B. In this case, input parameters may be part size, part position, part accuracy, etc., and output parameters may be sales, profit, yield rate, refinement level, manufacturing time, failure rate, etc.

[0017] As is known as the scale-up problem, the optimal conditions in the small-scale system 110 do not necessarily lead to optimal results in the large-scale system 120. For this reason, the condition determination device executes a condition determination process, which will be described in detail below, in order to determine the optimal conditions in the large-scale system 120 at the lowest possible cost.

[0018] The condition determination device is a small-scale system data set D S and large-scale dataset D L Based on the above, the trial conditions for the large-scale system 120 are determined using Bayesian optimization. That is, the condition determination device regards the large-scale system 120 as a black-box function, and searches for trial conditions for the black-box function using the surrogate model 150 and an acquisition function.

[0019] According to Bayesian optimization, the condition determination device performs the following steps: 1) approximating the large-scale system 120 using a surrogate model 150; and 2) designing an acquisition function from the surrogate model 150. For example, the condition determination device sets a surrogate model 150 that approximates (simulates) the large-scale system 120. For example, Gaussian process regression can be used as the surrogate model 150, but this is not limiting. When Gaussian process regression is used, n inputs x1, x2,...x n Given, the output y i =f(x i ) joint probability p(y1,y2,yn ) follows an n-dimensional Gaussian (normal) distribution, then the function f(x i ) is said to follow a Gaussian process.

[0020] As mentioned above, the surrogate model 150 is based on the small-scale dataset D S and large-scale dataset D L However, in order to generate the surrogate model 150 and the acquisition function, the number of dimensions (types of parameters included in both) of both systems must be the same. However, there may be differences in the controllable parameters between the small-scale system 110, such as a laboratory, and the large-scale system 120, such as a manufacturing plant. More specifically, for example, when a manufacturing process is carried out on a small scale in the small-scale system 110, mixing occurs naturally (no parameter setting is required), but in the large-scale system 120, the mixing speed must be explicitly set. In this case, the small-scale system data set D S does not include the parameter of the stirring speed, but the large-scale data set D L Conversely, if the pressure was controllable in the small-scale system 110 but not in the large-scale system 120, the small-scale system data set D S contains pressure parameters, while the large-scale data set D L does not include the pressure parameter.

[0021] Therefore, in this embodiment, the small-scale system data set D S and large-scale data set D L Regarding the small-scale dataset D, we confirmed the difference in the types of parameters (hereinafter, "type" and "classification" may be used interchangeably) between the two datasets, and if there are parameters that are included only in one dataset, we complement or delete them to match the types of parameters included in both datasets. S ' and large-scale data set D L In the previous example, for example, the stirring speed is S Large-scale dataset D, which is not included inL Since it is included in the small-scale data set D S Alternatively, the parameter of the stirring speed can be supplemented in the large-scale data set D L The parameter for the stirring speed is deleted from the small-scale data set D. S Although it is included in the large-scale data set D L Since it is not included in the small-scale dataset D S Remove the pressure parameter from the large-scale data set D L This complements the pressure parameter.

[0022] In this way, the small-scale system data set D, which is a group of parameters that can be controlled in the small-scale system 110, S and large-scale system data set D, which is a set of controllable parameters in large-scale system 120. L On the other hand, if there is a parameter type that is included in only one of the two, the condition determination device complements or deletes it to make the parameters included in both the small-scale system data set D S ' and large-scale data set D L Then, the condition determination device generates a surrogate model 150 based on these data sets and attempts to optimize the parameters (search for conditions).

[0023] 1.2 How to adjust (delete / supplement) parameters Referring to Figure 2, small-scale data set D S or large-scale data set D L The processing for the parameter x included in at least one of the above will be described.

[0024] First, the parameter x is the small-scale data set D S and large-scale data set D L If the parameter x is included in the small-scale system data set D, it is determined whether or not the parameter x is included in the small-scale system data set D (S201). S and large-scale data set D LIf the parameter x is included in both of the small-scale system data set D, the processing for the parameter x is terminated (the processing moves to the next parameter). S Included only in the large-scale dataset D L If the parameter x is not included in the small-scale system data set D, it is determined whether or not to use the parameter x as additional information for generating the surrogate model 150 (S203). Whether or not to use the parameter x as additional information can be determined, for example, based on whether the parameter has a large influence on the output. If the parameter x is not to be used as additional information (No in S203), S The parameter x is deleted from the small-scale data set D S ' is generated (S205).

[0025] If the parameter x is used as additional information (Yes in S203), it is determined whether the amount of additional information of the parameter x is large (S207). This can be determined by whether the influence of the parameter x on the output is large. If it is determined that the amount of additional information of the parameter x is large (Yes in S207), the small-scale system data set D S Parameter x (here parameter x S ) from the large-scale data set D L parameter x (here parameter x L ) and use this item in the large-scale data set D L By adding it to the large-scale data set D L ' is generated (S209). Parameter x L There are various methods for estimating . For example, the small-scale data set D S In the parameter x S and one or more other parameters (here, parameter y S If a correlation or other relationship can be found between x S =f(y S ) can be written as y Scan be a scalar or a vector of two or more parameters), and a large-scale system dataset D L One or more parameters y contained in L By applying this relationship to the parameter x L can be estimated.

[0026] On the other hand, if it is determined that there is not much additional information about the parameters (No in S207), the small-scale system data set D S The average value of parameter x contained in parameter x L As a result, this item is called the large-scale dataset D L By adding it to the large-scale data set D L ' is generated (S211).

[0027] In S201, the parameter x is a large-scale data set D L Only included in the small-scale dataset D S If the parameter x is not included in the large-scale system dataset D, it is determined whether or not to use the parameter x as additional information for generating the surrogate model 150 (S213). Whether or not to use the parameter x as additional information can be determined, for example, based on whether the parameter has a large influence on the output. If the parameter x is not to be used as additional information (No in S213), L The parameter x is deleted from the large-scale data set D L ' is generated (S215).

[0028] If the parameter x is used as additional information (Yes in S213), it is determined whether the amount of additional information of the parameter x is large (S217). In other words, this can be determined by whether the influence of the parameter x on the output is large. If it is determined that the amount of additional information of the parameter x is large (Yes in S217), the large-scale system data set D L Parameter x contained in L From the small-scale data set D S The parameter x that should be included in Sand this item is used in the small-scale data set D S By adding to the small-scale data set D S ' is generated (S219). As in the process of S209, the parameter x L There are various methods for estimating . For example, for a large-scale data set D L In the parameter x L and one or more other parameters (here, parameter y L If a correlation or other relationship can be found between x L =f(y L ) can be written as y L can be a scalar or a vector of two or more parameters), and the small-scale system dataset D S One or more parameters y contained in S By applying this relationship to the parameter x S can be estimated.

[0029] On the other hand, if it is determined that there is not much additional information about the parameters (No in S217), the large-scale system dataset D L The average value of parameter x contained in parameter x S This item is called small-scale data set D S By adding to the small-scale data set D S ' is generated (S221).

[0030] 1.3 How to optimize parameters (how to search for trial conditions) A specific example of a method for searching for suitable trial conditions (a method for optimizing parameters) in the large-scale system 120 will be described below.

[0031] As described above, the condition determination device according to this embodiment determines trial conditions (adjusts / optimizes parameters) using Bayesian optimization. According to Bayesian optimization, the condition determination device sets a surrogate model 150 that approximates (simulates) the large-scale system 120.

[0032] Here, we assume that the problem is a maximization problem that searches for the input that maximizes the output. However, Bayesian optimization according to the present disclosure is not limited to this. For example, minimization can also be considered by determining the sign. Furthermore, if we want the output to approach a specific value, we can consider the difference between the output and that specific value, or the modulus or absolute value of the difference, and minimize that difference.

[0033] When Gaussian process regression is used as the surrogate model 150, the condition determination device uses the small-scale system data set D obtained from the small-scale system 110 and the large-scale system 120, respectively. S and large-scale dataset D L As mentioned above, the parameter types included are aligned to create the small-scale data set D S ' and large-scale data set D L ', and then the predictive distribution of the Gaussian process based on these datasets

number

number

number

number

number

number

number

number

[0034] Also, for a positive definite kernel, for a set X,

number

[0035] Symmetry

number

number

number

number

[0036] In this case, the following kernel trick is known to hold.

[0037]

number

[0038] where:

number

[0039] In this way, the small-scale data set D S Predicting the distribution of Gaussian processes using surrogate models and 150

number

[0040]

number

[0041] where Φ(z) is the cumulative distribution of the standard normal distribution in the range [-∞,z].

number

[0042] The condition determination device determines the trial conditions to be sampled next from the acquisition function derived in this way. By implementing the determined trial conditions (parameter set) in the large-scale system 120, results such as the yield rate are obtained as output.

[0043] That is, the condition determination device calculates the small-scale system data set D for the small-scale system 110. S and large-scale dataset D L The condition determination device determines the trial conditions (parameter set) to be implemented in the large-scale system 120 based on the above. Note that initially, the condition determination device may determine the trial conditions to be implemented in the large-scale system 120 based on data related to the small-scale system 110.

[0044] The data generation process of the small-scale system 110 can be modeled as follows.

number

[0045] That is, the parameter W of the small-scale system 110 S has mean 0 and variance

number

number

[0046]

number

[0047] That is, the parameter W of the large-scale system 120 Lis the parameter W of the small-scale system 110 S It is assumed that there is a deviation of δW from the mean.

number

number

number

[0048]

number

number

[0049] Here, the following small-scale data set D S and large-scale dataset D L Think about it.

[0050]

number

[0051] Here, n and m are the sample sizes of each data set. If the large-scale system 120 data set is empty, D L ={}. The joint distribution is

number

number

[0052] First, the joint distribution can be written as follows:

[0053]

number

[0054] Now, to make the formula easier to read,

number

[0055]

number

[0056] That is, the predictive distribution as the surrogate model 150 is

number

[0057] When the trial results for the large-scale system 120 under the trial conditions do not satisfy a predetermined termination condition, the condition determination device determines further trial conditions for the large-scale system 120. Specifically, if the results (e.g., yield rate) of the trial in the large-scale system 120 under the trial conditions determined by the trial condition determination device are not equal to or greater than a predetermined threshold, the condition determination device can instruct the determination of the next trial conditions for the large-scale system 120. On the other hand, if the trial results are equal to or greater than the predetermined threshold, the condition determination device determines that the trial conditions are suitable and terminates the condition determination process. That is, the condition determination device sequentially determines trial conditions and tries out the work process in the large-scale system 120 under the trial conditions until the results of the trial conditions (parameter set) determined by the condition determination device satisfy the predetermined termination condition. Here, using a threshold as the satisfying condition is merely an example, and the number of trials, for example, may also be used.

[0058] In the above example, the small-scale data set D S and / or large-scale dataset D L is composed of pairs of trial conditions and trial results, but is not limited to this and may include a prediction model that predicts trial results from trial conditions of the small-scale system 110 and / or the large-scale system 120. Such a prediction model enables the condition determination device to predict trial results from trial conditions for the small-scale system 110 and / or the large-scale system 120, and the small-scale system data set D S and / or large-scale dataset D L There is no need to obtain

[0059] 2. System Configuration An example of the system configuration of the condition determination device used in this embodiment will be described with reference to the block diagrams of FIGS.

[0060] 2.1 Software Configuration First, the software configuration of the condition determination device will be described with reference to Fig. 3. Note that, although Fig. 3 shows the condition determination program 300 as if it were a single program, in reality, two or more program modules may work together to achieve the function. Furthermore, it does not have to be executed by only one computer (hardware), and may be realized as a program that operates on an information processing system consisting of two or more computers that can work together.

[0061] As shown in FIG. 3, the condition determination program 300 includes a first parameter group input unit 301, a second parameter group input unit 303, a first parameter adjustment unit 305, a second parameter adjustment unit 307, a model generation unit 309, a parameter search unit 311, and a database 330.

[0062] The first parameter group input unit 301 receives a small-scale system data set D, which is an input / output result when a manufacturing process is tried one or more times in the small-scale system 110. S The input of the small-scale data set D is received. S is stored in the database 330. S can include one or more parameters that can be adjusted (controlled) in the small-scale system 110, which are input (trial conditions), i.e., one or more parameters such as temperature, pressure, flow rate, concentration, etc., and one or more parameters that are output, i.e., one or more parameters such as sales, profit, yield rate, refinement degree, production time, failure rate, etc. The group of parameters that are input and output in one trial in the small-scale system 110 constitutes one data set, and the data for the number of trials is the small-scale system data set D S will be included in the

[0063] The second parameter group input unit 303 receives a large-scale system data set D, which is an input / output result when a manufacturing process is tried one or more times in the large-scale system 120. L The input is received. The large-scale data set D L is stored in the database 330. L The small-scale dataset DS Similarly, the large-scale system data set D can include one or more input (trial conditions) parameters that can be adjusted (controlled) in the small-scale system 110, such as temperature, pressure, flow rate, concentration, etc., and one or more output parameters, such as sales, profit, yield rate, refinement level, production time, failure rate, etc. The set of parameters that are input and output for one trial in the large-scale system 120 constitutes one data set, and the data for the number of trials is the large-scale system data set D L will be included in the

[0064] As mentioned above, there is a difference in adjustable (controllable) parameters between the small-scale system 110 and the large-scale system 120. Therefore, the small-scale system data set D S and large-scale dataset D L and can contain one or more parameters of different types. S and large-scale dataset D L It may contain one or more parameters that are included in only one of the above.

[0065] Therefore, the first parameter adjustment unit 305 and the second parameter adjustment unit 307 adjust the small-scale system data set D S and large-scale dataset D L By deleting / complementing parameters that are included only in one of the datasets so that the parameter types included are the same, we can obtain the small-scale dataset D S ' and large-scale data set D L The parameter deletion / complement methods performed by first parameter adjustment section 305 and second parameter adjustment section 307 have been described in detail in 1.2 above, and therefore will not be described here.

[0066] The model generation unit 309 and the parameter search unit 311 each generate a small-scale system data set D S ' and large-scale data set D L' is used to generate a surrogate model 150 and an acquisition function that simulates the large-scale system 120, and to determine a set of parameters for improving the output of the large-scale system 120. The method for generating the surrogate model 150 and the acquisition function, and the method for determining a set of parameters for improving the output of the large-scale system 120 have been explained in detail in 1.3 above, so further explanation will be omitted here.

[0067] The database 330 is for storing various data managed by the condition determination program 300. Specifically, the small-scale system data set D S and large-scale data set D L , and a small-scale dataset D adjusted to include the same types of parameters as these datasets. S ' and large-scale data set D L ' is managed on the database 330. The surrogate model 150 for simulating the large-scale system 120 and an acquisition function (not shown) can also be managed on the database 330.

[0068] 2.2 Hardware Configuration An example of the hardware configuration of a condition determination device 400, which is an information processing device (computer / information processing system) that executes the condition determination program illustrated in Fig. 3, will be described below with reference to the block diagram of Fig. 4. Note that, although Fig. 4 illustrates the condition determination device 400 as if it were a single physical device, in reality, it can also be realized by an information processing system in which two or more computers work together.

[0069] The condition determination device 400 illustrated in FIG. 4 includes a control unit 401, a memory unit 405, a communication interface (I / F) unit 407, an input unit 409, and an output unit 411, and each unit can be connected to each other via a bus line 415 so that they can communicate with each other.

[0070] The control unit 401 may include a central processing unit (CPU), a random access memory (RAM) 403, a read only memory (ROM), etc., and controls each component in accordance with information processing. More specifically, for example, the CPU that may be included in the control unit 401 can perform various processes by loading an operating system (OS) (not shown) and various application programs including the condition determination program 300 from the storage unit 405 into the RAM 403 and executing the various programs such as the condition determination program 300. Note that the control unit 401 may also include a graphic processing unit (GPU), an application specific integrated circuit (ASIC), a field programmable array (FPGA), etc. instead of or in addition to the CPU.

[0071] The memory unit 405 is an auxiliary storage device such as a hard disk drive (HDD) or a solid state drive (SSD), and stores various programs such as the condition determination program 300 executed by the control unit 401, as well as various data used therein.

[0072] The communication I / F unit 407 is, for example, a communication module for communicating with other devices via wired or wireless communication. The communication method used by the communication I / F unit 407 to communicate with other devices is arbitrary, but may be, for example, a public line network such as the Internet.

[0073] The input unit 409 is a device for receiving various input operations from the user, which can be realized by, for example, a mouse, a keyboard, a touch panel, etc. The output unit 411 is a device for notifying the user of various information by display, sound, etc., such as a display or a speaker.

[0074] 3. Summary As described above, when there is a difference in the controllable parameters between the small-scale system 110 and the large-scale system 120, the condition determination device 400 and the condition determination program delete or store the different parameters to create a small-scale system data set D S ' and large-scale data set D L ' and, based on this, generate a surrogate model 150 and an acquisition function. Then, the condition determination device 400 and the condition determination program use the generated surrogate model 150 and acquisition function to search for a set of parameters that will optimize the output in the large-scale system 120. This makes it possible to suitably solve the scale-up problem even if there is a difference in the controllable parameters between the small-scale system 110 and the large-scale system 120.

[0075] 4.Other The above-described embodiments are merely examples in all respects and should not be construed as limiting. The present invention is not limited to the above-described embodiments, and can be embodied in various other forms without departing from the spirit of the present invention. [Explanation of symbols]

[0076] 110: Small-scale 120: Large-scale systems 150: Surrogate model 300: Condition determination program 301: First parameter group input section 303: Second parameter group input unit 305: First parameter adjustment unit 307: Second parameter adjustment unit 309: Model generation unit 311: Parameter search unit 330: Database 400: Condition determination device 401: Control unit 403:RAM 405: Storage section 407: Communication interface (I / F) section 409: Input section 411: Output section 415: Bus Line D L :Large-scale dataset D L ' :Large-scale dataset D S : Small-scale dataset D S ' : Small-scale dataset W L :parameter W S :parameter

Claims

1. 1. A method for improving an output in a second system based on a first data set comprising a first set of adjustable parameters in a first system and a second data set comprising a second set of adjustable parameters in a second system different from the first system, the method comprising: the first parameter group and the second parameter group include one or more parameters of different types, The method comprises: a first process of generating a third data set and a fourth data set consisting of a third parameter group by supplementing and / or deleting the one or more parameters of different types from the first data set and the second data set; a second process for generating a model simulating input and output in the second system using the third data set and the fourth data set; a third process for determining a set of parameters for improving an output in the second system using the third data set, the fourth data set, and the model; and A method comprising:

2. the first processing includes processing of deleting at least some parameters from the one or more parameters of different types; The method of claim 1.

3. the second processing includes, for one or more types of parameters that are included in one of the first data set and the second data set and not included in the other, adding an average value of the parameter included in one of the first data set and the second data set to the other data set, among the one or more different types of parameters; The method of claim 1.

4. the first processing includes, for one or more types of parameters that are included in one of the first data set and the second data set and not included in the other, predicting parameters to be included in the other data set based on the parameters included in one of the one or more different types of parameters, and adding the predicted parameters to the other data set. The method of claim 1.

5. A program for improving an output in a second system based on a first data set consisting of a first group of parameters adjustable in a first system and a second data set consisting of a second group of parameters adjustable in a second system different from the first system, the program comprising: the first parameter group and the second parameter group include one or more parameters of different types, The program a first process of generating a third data set and a fourth data set consisting of a third parameter group by supplementing and / or deleting the one or more parameters of different types from the first data set and the second data set; a second process for generating a model simulating input and output in the second system using the third data set and the fourth data set; a third process for determining the third set of parameters for improving an output in the second system using the third data set, the fourth data set, and the model; and A program for causing an information processing system to execute the above.

6. 1. An information processing system for improving an output in a second system based on a first data set consisting of a first group of parameters adjustable in a first system and a second data set consisting of a second group of parameters adjustable in a second system different from the first system, comprising: the first parameter group and the second parameter group include one or more parameters of different types, The information processing system includes: a first means for generating a third data set and a fourth data set each including a third parameter group by supplementing and / or deleting one or more parameters of different types from the first data set and the second data set; second means for generating a model simulating input and output in the second system using the third data set and the fourth data set; a third means for determining the third set of parameters for improving an output in the second system using the third data set, the fourth data set, and the model; and An information processing system having:

Images