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

Abstract

An information processing system according to an embodiment includes at least one processor. The at least one processor obtains a numerical representation and a complex ratio for each of a plurality of component objects, performs machine learning based on the plurality of numerical representations, calculates a plurality of regression parameters corresponding to the plurality of component objects, applies the plurality of complex ratios to a regression model defined by the plurality of regression parameters, and calculates a predicted value representing a characteristic of a complex object obtained by compounding the plurality of component objects.

Description

Technical Field

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

[0002] This method utilizes machine learning to analyze composite objects obtained by combining multiple component objects. For example, Patent Document 1 describes a method for predicting the bonding between the stereostructure of a biopolymer and the stereostructure of a compound. This method includes: generating a predicted stereostructure of the biopolymer-compound composite based on the stereostructure of the biopolymer and the stereostructure of the compound; transforming the predicted stereostructure into a predicted stereostructure vector representing a check result with an interaction pattern; and predicting the bonding between the stereostructure of the biopolymer and the stereostructure of the compound by using a machine learning algorithm to determine the predicted stereostructure vector.

[0003] Previous technical documents

[0004] Patent documents

[0005] Patent Document 1: Japanese Patent Application Publication No. 2019-28879 Summary of the Invention

[0006] The technical problem to be solved by the invention

[0007] When there are diverse or multiple component objects, it may be impossible to prepare sufficient data for these component objects. As a result, the analytical accuracy of the composite object may not reach the expected level. Therefore, there is a need for a structure that can improve the analytical accuracy of composite objects even when sufficient data for each component object is unavailable.

[0008] means for solving technical problems

[0009] The information processing system according to one aspect of the present invention includes at least one processor. The at least one processor acquires numerical representations and composite ratios for each of a plurality of component objects, performs machine learning based on the plurality of numerical representations, calculates a plurality of regression parameters corresponding to the plurality of component objects, applies the plurality of composite ratios to a regression model defined by the plurality of regression parameters, and calculates predicted values ​​representing the characteristics of the composite object obtained by combining the plurality of component objects.

[0010] The information processing method according to one aspect of the present invention is executed by an information processing system having at least one processor. The information processing method includes the following steps: obtaining a numerical representation and a composite ratio for each of a plurality of component objects; performing machine learning based on the plurality of numerical representations to calculate a plurality of regression parameters corresponding to the plurality of component objects; and applying the plurality of composite ratios to a regression model defined by the plurality of regression parameters to calculate a predicted value representing the characteristics of the composite object obtained by combining the plurality of component objects.

[0011] The information processing program involved in one aspect of the present invention causes a computer to perform the following steps: obtaining a numerical representation and a composite ratio for each of a plurality of component objects; performing machine learning based on the plurality of numerical representations to calculate a plurality of regression parameters corresponding to the plurality of component objects; and applying the plurality of composite ratios to a regression model defined by the plurality of regression parameters to calculate a predicted value representing the characteristics of the composite object obtained by combining the plurality of component objects.

[0012] In this approach, machine learning is performed based on the data for each component object to calculate multiple regression parameters corresponding to the multiple component objects. The composite ratio is then applied to a regression model defined by these regression parameters to predict the characteristics of the composite object. By using machine learning and regression models, the accuracy of composite object analysis can be improved even when sufficient data for each component object is unavailable.

[0013] Invention Effects

[0014] According to one aspect of the present invention, the analytical accuracy of composite objects can be improved even when sufficient data cannot be prepared for the component objects. Attached Figure Description

[0015] Figure 1 This is a diagram illustrating an example of the hardware structure of a computer constituting the information processing system according to the implementation method.

[0016] Figure 2 This is a diagram illustrating an example of the functional structure of an information processing system involved in an implementation method.

[0017] Figure 3 This is a flowchart illustrating an example of the operation of an information processing system involved in an implementation method.

[0018] Figure 4 This is a diagram illustrating an example of the steps involved in calculating regression parameters.

[0019] Figure 5 This is another example of the steps involved in calculating regression parameters. Detailed Implementation

[0020] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements are labeled with the same symbols, and repeated descriptions are omitted.

[0021] [System Overview]

[0022] The information processing system 10 involved in the implementation is a computer system that performs analysis related to a composite object obtained by combining multiple component objects in a given compounding ratio. Component objects refer to tangible or intangible objects used to generate the composite object. The composite object can be either tangible or intangible. Examples of tangible objects include any substance or object. Examples of intangible objects include data and information. "Combining multiple component objects" refers to the process of treating multiple component objects as a single object, i.e., a composite object. The method of combination is not limited; for example, it can be coordination, blending, synthesis, combination, mixing, merging, combination, chemical bonding, or fusion, or other methods. Analysis related to the composite object refers to the processing used to obtain data representing certain characteristics of the composite object.

[0023] Multiple component objects can be any number of materials; in this case, the composite object is a multi-component substance generated from these materials. A material refers to any constituent element used to generate a multi-component substance. For example, multiple materials can be any number of molecules or atoms; in this case, the composite object is a multi-component substance obtained by combining these molecules or atoms in any way. For example, materials can be polymers or monomers; in contrast, the multi-component substance can be a polymer alloy. Materials can be monomers; in contrast, the multi-component substance can be a polymer. Materials can be pharmaceuticals, i.e., chemical substances with pharmacological effects; in contrast, the multi-component substance can be a pharmaceutical preparation.

[0024] Information processing system 10 performs machine learning for analysis related to complex objects. Machine learning refers to a method of autonomously discovering laws or rules by learning from given information. There are no limitations on the specific methods of machine learning. For example, information processing system 10 can perform machine learning using a machine learning model that is a computational model comprising a neural network. A neural network is an information processing model that simulates the structure of the human brain's nervous system. As a more specific example, information processing system 10 can also use at least one of graph neural networks (GNNs), convolutional neural networks (CNNs), recurrent neural networks (RNNs), attention RNNs, and multi-head attention to perform machine learning.

[0025] [System Structure]

[0026] The information processing system 10 consists of one or more computers. When multiple computers are used, these computers are connected via communication networks such as the Internet and intranets to logically construct an information processing system 10.

[0027] Figure 1 This diagram illustrates an example of the general hardware structure of a computer 100 constituting an information processing system 10. For example, the computer 100 includes a processor 101 such as a CPU that executes an operating system, application programs, etc., a main storage unit 102 composed of ROM and RAM, an auxiliary storage unit 103 composed of hard disk, flash memory, etc., a communication control unit 104 composed of a network card or wireless communication module, input devices 105 such as a keyboard and mouse, and output devices 106 such as a monitor.

[0028] Each functional component of the information processing system 10 is implemented by reading a pre-set program into the processor 101 or main storage unit 102 and executing the program. According to the program, the processor 101 activates the communication control unit 104, the input device 105, or the output device 106 to read and write data from the main storage unit 102 or the auxiliary storage unit 103. The data or database required for processing is stored in the main storage unit 102 or the auxiliary storage unit 103.

[0029] Figure 2 This diagram illustrates an example of the functional structure of an information processing system 10. The information processing system 10 includes an acquisition unit 11, a calculation unit 12, and a prediction unit 13 as functional components.

[0030] The acquisition unit 11 is a functional component for acquiring data related to multiple component objects. Specifically, the acquisition unit 11 acquires the numerical representation and composite ratio for each of the multiple component objects. The numerical representation of a component object refers to data that uses multiple numerical values ​​to represent any attribute of the component object. The attribute of a component object refers to the property or characteristic possessed by the component object. The numerical representation can be visualized through various methods, such as through numbers, English characters, text, molecular diagrams, vectors, images, time series data, etc., or through any combination of two or more of these methods. Each numerical value constituting the numerical representation can be represented in decimal, or in binary, hexadecimal, or other notation. The composite ratio of the component objects refers to the proportion between the multiple component objects. The specific type, unit, and representation method of the composite ratio are not limited and can be arbitrarily set according to the component objects or composite objects. For example, the composite ratio can be represented by percentage ratios, histograms, or absolute quantities of each component object.

[0031] The calculation unit 12 is a functional component that calculates regression parameters for a regression model used to predict the characteristics of composite objects. Specifically, the calculation unit 12 performs machine learning based on multiple numerical representations corresponding to multiple component objects to calculate regression parameters. A regression model is a formula used to find the value of a target variable y that is greater than or equal to 1, given a value of an explanatory variable x greater than or equal to 1. The regression model can be a linear regression model or a nonlinear regression model. An example of a regression model is the Scheffe polynomial. However, the regression model can also be other parametric models. Regression parameters are the numerical values ​​included in the regression model.

[0032] The prediction unit 13 is a functional component that predicts the characteristics of the composite object and outputs its predicted value. The characteristics of the composite object refer to the unique properties possessed by the composite object. Specifically, the prediction unit 13 applies the composite ratio to a regression model defined by the calculated regression parameters to calculate the predicted value. In other words, the prediction unit 13 substitutes multiple composite ratios into the regression model to calculate the predicted value.

[0033] In one example, the combination of the calculation unit 12 and the prediction unit 13 is implemented using a machine learning model. Alternatively, the calculation unit 12 can be implemented using a machine learning model, while the prediction unit 13 can be implemented using an algorithm that does not use a machine learning model.

[0034] In one example, each of the at least one machine learning model used in this embodiment is the learned model with the highest expected inference accuracy, and therefore can be called the "best machine learning model". However, it should be noted that the learned model is not limited to "the best in reality". The learned model is generated by a given computer processing teacher data containing multiple combinations of input vectors and labels. The given computer inputs the input vectors into the machine learning model to calculate the output value and find the error between the output value and the label represented by the teacher data. The output value is, for example, a predicted value. The error between the output value and the label can be said to be the difference between the inference result and the correct answer. The computer updates the given parameters within the machine learning model based on this error. The computer generates the learned model by repeating such learning. The computer that generates the learned model is not limited, and can be, for example, information processing system 10 or other computer systems. The process of generating the learned model can be called the learning phase, and the process of using the learned model can be called the application phase.

[0035] In one example, the machine learning model used in this embodiment can be described by a function that is independent of the input order. This structure eliminates the influence of the order of multiple vectors in machine learning.

[0036] [System Actions]

[0037] refer to Figure 3 The operation of the information processing system 10 will be explained, and the information processing method involved in this embodiment will also be explained. Figure 3 This is a flowchart representing an example of the actions of the information processing system 10 as processing flow S1. Processing flow S1 is equivalent to the application phase.

[0038] In step S11, the acquisition unit 11 acquires the numerical representation and composite ratio for each of the multiple component objects. For example, if information related to two component objects Ea and Eb is input, the acquisition unit 11 acquires, for instance, the numerical representation of component object Ea {1, 1, 2, 3, 4, 3, 3, 5, 6, 7, 5, 4}, the numerical representation of component object Eb {1, 1, 5, 6, 4, 3, 3, 5, 1, 7, 0, 0}, and the composite ratio of component objects Ea and Eb {0.7, 0.3}. In this example, each numerical representation is represented by a vector. The composite ratio {0.7, 0.3} refers to obtaining a composite object using component objects Ea and Eb in a 7:3 ratio.

[0039] The acquisition unit 11 can acquire data from each of the multiple component objects using any method. For example, the acquisition unit 11 can access a given database to read data, receive data from other computers or computer systems, or accept data input by the user of the information processing system 10. Alternatively, the acquisition unit 11 can acquire data using any two or more of these methods.

[0040] In step S12, the calculation unit 12 calculates a feature vector for each of the plurality of component objects based on its numerical representation. A feature vector is a vector that represents the characteristics of a component object. The characteristics of a component object refer to any element that distinguishes it from other objects. A vector is an n-dimensional quantity with n numerical values, which can be represented as a one-dimensional arrangement.

[0041] In step S13, the calculation unit 12 calculates multiple regression parameters corresponding to multiple component objects based on the calculated multiple feature vectors.

[0042] In step S14, the prediction unit 13 uses a regression model defined by multiple calculated regression parameters to calculate predicted values ​​representing the characteristics of the composite object. In short, a regression model defined by regression parameters refers to a regression model that uses specific numerical values ​​as regression parameters. The prediction unit 13 applies multiple composite ratios to the regression model to calculate predicted values.

[0043] In step S15, the prediction unit 13 outputs the predicted value. The method of outputting the predicted value is not limited. For example, the prediction unit 13 may store the predicted value in a given database, send it to another computer or computer system, or display it on a display device. Alternatively, the prediction unit 13 may output the predicted value to other functional components for subsequent processing in the information processing system 10.

[0044] refer to Figure 4 and Figure 5 The treatments related to the regression model will be explained in more detail. Figure 4 and Figure 5 These are all diagrams illustrating examples of the steps involved in calculating regression parameters. In any example, the component objects represent three materials (polymers): polystyrene, polyacrylic acid, and polybutyl methacrylate. For each of these materials, any form of numerical representation can be prepared.

[0045] The above Scheffe polynomial is frequently used in problems related to material fit. Therefore, in Figure 4 and Figure 5 In the example, we assume the regression model is a Scheffe multinomial.

[0046] right Figure 4 The following example illustrates this. In step S121, which is part of step S12, the computation unit 12 uses a machine learning model with an embedding function for calculating the features of the vector to calculate the feature vector for each of the multiple component objects based on its numerical representation. This machine learning model is a learned model. In the embedding function, the input vector and the output vector have a one-to-one relationship. In this example, the input vector is a numerical representation, and the output vector is the feature vector Z. The computation unit 12 inputs multiple numerical representations corresponding to the multiple component objects into the embedding function model and calculates the feature vector Z for each of the multiple component objects. In one example, the computation unit 12 inputs the numerical representation corresponding to that component object into the embedding function model for each of the multiple component objects and calculates the feature vector Z for that component object. In one example, the embedding function model can generate a feature vector Z as a fixed-length vector based on the numerical representation as non-variable data. Non-variable data refers to data that is not represented by a fixed-length vector. Figure 4 In the example, the calculation unit 12 calculates the characteristic vector Z1 corresponding to polystyrene, the characteristic vector Z2 corresponding to polyacrylic acid, and the characteristic vector Z3 corresponding to polybutyl methacrylate.

[0047] The embedding function is not limited by the machine learning model used; it can be determined arbitrarily by considering factors such as the types of component objects and composite objects. For example, the computing unit 12 can use a graph neural network (GNN), a convolutional neural network (CNN), or a recurrent neural network (RNN) to execute the embedding function.

[0048] In step S122, which is part of step S12, the computation unit 12 uses a machine learning model with an interaction function for interacting multiple vectors to calculate other feature vectors M for multiple component objects based on feature vector Z. This machine learning model is a learned model. In the interaction function, the input vector and the output vector have a one-to-one relationship. In this example, the input vector is the feature vector Z, and the output vector is the feature vector M. In one example, the computation unit 12 inputs a set of multiple feature vectors Z corresponding to multiple component objects into the interaction function model and calculates the feature vector M for each of the multiple component objects. Figure 4 In the example, the calculation unit 12 calculates the characteristic vector M1 corresponding to polystyrene, the characteristic vector M2 corresponding to polyacrylic acid, and the characteristic vector M3 corresponding to polybutyl methacrylate.

[0049] The interaction function can be determined arbitrarily, taking into account factors such as the types of component objects and composite objects. For example, the computation unit 12 can use an attention RNN or multi-head attention to perform machine learning on the interaction function. In another example, the computation unit 12 can also use an interaction function without learning parameters to compute the feature vector M.

[0050] exist Figure 4 In step S13, the calculation unit 12 calculates the regression parameter 'a' of the first-order term of the linear regression model for each of the multiple component objects, based on the feature vector M. In one example, the calculation unit 12 calculates the regression parameter using a machine learning model. This machine learning model is a fully learned model. In the function that calculates the regression parameter of the first-order term, the input vector and the output value have a one-to-one relationship. In this example, the input vector is the feature vector M, and the output value is the regression parameter 'a'. In one example, the calculation unit 12 inputs a set of multiple feature vectors M corresponding to multiple component objects into the machine learning model and calculates the regression parameter 'a' for each of the multiple component objects. Figure 4 In the example, the calculation unit 12 calculates the regression parameter a1 corresponding to polystyrene, the regression parameter a2 corresponding to polyacrylic acid, and the regression parameter a3 corresponding to polybutyl methacrylate.

[0051] There are no limitations on the machine learning model used to calculate the regression parameters; it can be determined arbitrarily, taking into account factors such as the types of component objects and composite objects. For example, the calculation unit 12 can use a fully connected neural network (FCNN) to calculate the regression parameters.

[0052] exist Figure 4 In step S14, the prediction unit 13 calculates the predicted value E using the following Scheffe polynomial (1) defined by three regression parameters a1, a2, and a3. The regression parameter a can also be considered the regression coefficient of the first-order term in equation (1). The predicted value E represents the properties of the multi-component material (polymer alloy) obtained from polystyrene, polyacrylic acid, and polybutyl methacrylate. The variable r in equation (1) refers to the composite ratio. The composite ratios of polystyrene, polyacrylic acid, and polybutyl methacrylate are represented as r1, r2, and r3, respectively.

[0053] [Formula 1]

[0054]

[0055] right Figure 5 Let's illustrate this with an example. Figure 5 In the example, step S12 includes steps S121 and S122. Figure 4 The examples are the same, steps S13 and S14 are the same. Figure 4 The examples are different.

[0056] exist Figure 5 In step S13, the calculation unit 12 calculates the regression parameters of the linear regression model for each of the multiple component objects based on the feature vector M. Specifically, the calculation unit 12 calculates the regression parameter a for the first-order term and the regression parameter b for the second-order term. In one example, the calculation unit 12 calculates the regression parameters using machine learning such as FCNN. Machine learning models are prepared for the first-order and second-order terms of the linear regression model respectively.

[0057] and Figure 4 Similarly, in the function that computes the regression parameters of a linear term, the input vector and output value have a one-to-one relationship. In this example, the input vector is the feature vector M, and the output value is the regression parameter a. In one example, the computation unit 12 inputs a set of multiple feature vectors M corresponding to multiple component objects into the machine learning model, and computes the regression parameter a for each of the multiple component objects. Figure 5 Similarly, in the example, the calculation unit 12 calculates the regression parameter a1 corresponding to polystyrene, the regression parameter a2 corresponding to polyacrylic acid, and the regression parameter a3 corresponding to polybutyl methacrylate.

[0058] In the function that calculates the regression parameters for the quadratic term, the individual input vectors are obtained by synthesizing two feature vectors. This function calculates a regression parameter based on the two vectors. In this example, two feature vectors M are synthesized. Figure 5 In the example, the calculation unit 12 synthesizes two feature vectors M1 and M2 to generate the first input vector, synthesizes two feature vectors M1 and M3 to generate the second input vector, and synthesizes two feature vectors M2 and M3 to generate the third input vector. Therefore, the first input vector corresponds to polystyrene and polyacrylic acid, the second input vector corresponds to polystyrene and polybutyl methacrylate, and the third input vector corresponds to polyacrylic acid and polybutyl methacrylate. In the function that calculates the regression parameters of the quadratic term, the input vector and the output value also have a one-to-one relationship. In this example, the input vector is the synthesis of two feature vectors M, and the output value is the regression parameter b. In one example, the calculation unit 12 inputs all combinations of input vectors into the machine learning model and calculates the regression parameter b for each combination. Figure 5 In the example, the calculation unit 12 calculates the regression parameter b corresponding to the combination of polystyrene and polyacrylic acid. 12 The regression parameter b corresponding to the combination of polystyrene and polybutyl methacrylate 13 and the regression parameter b corresponding to the combination of polyacrylic acid and polybutyl methacrylate. 23 .

[0059] exist Figure 5 In step S14 shown, the prediction unit 13 uses six regression parameters a1, a2, a3, b... 12 b 13 b 23 The predicted value E is calculated using the following Scheffe polynomial (2). In equation (2), the regression parameter a can also be said to be the regression coefficient of the first-order term, and the regression parameter b can also be said to be the regression coefficient of the second-order term. The variable r in equation (2) has the same meaning as in equation (1), and is a composite ratio.

[0060] [Formula 2]

[0061]

[0062] exist Figure 4 and Figure 5 Three component objects are shown, but the number of component objects is of course unlimited, and the information processing system 10 can process any number of component objects.

[0063] Similarly, for regression models containing terms of degree three or more, or other parameters, the information processing system 10 outputs each regression parameter based on the feature vectors of all relevant component objects. In cases where regression parameters are calculated independently of specific explanatory variables, such as the intercept in linear regression, the information processing system 10 can output a regression parameter based on the feature vectors of all component objects.

[0064] exist Figure 4 and Figure 5 In the example, the computation unit 12 executes both the embedding function and the interaction function, but one of these functions can be omitted. For example, the computation unit 12 can also calculate the regression parameters based on the feature vector Z obtained by the machine learning model through the embedding function. In any case, the computation unit 12 executes machine learning to calculate the regression parameters.

[0065] In one example, the machine learning model for the embedding function, the machine learning model for the interaction function, the machine learning model for the regression parameters, and the regression model can be constructed from a single neural network or from a collection of multiple neural networks. Alternatively, the machine learning model for the embedding function, the machine learning model for the interaction function, and the machine learning model for the regression parameters can be constructed from a single neural network or from a collection of multiple neural networks.

[0066] [program]

[0067] The information processing program for enabling a computer or computer system to function as an information processing system 10 includes program code for enabling the computer system to function as an acquisition unit 11, a calculation unit 12, and a prediction unit 13. This information processing program can be provided on a tangible recording medium such as a CD-ROM, DVD-ROM, or semiconductor memory, without being temporarily stored. Alternatively, the information processing program can 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 an auxiliary storage unit 103. The processor 101 reads the information processing program from the auxiliary storage unit 103 and executes it, thereby realizing the aforementioned functional requirements.

[0068] [Effect]

[0069] As described above, the information processing system according to one aspect of the present invention includes at least one processor. The at least one processor acquires numerical representations and composite ratios for each of the plurality of component objects, performs machine learning based on the plurality of numerical representations, calculates a plurality of regression parameters corresponding to the plurality of component objects, applies the plurality of composite ratios to a regression model defined by the plurality of regression parameters, and calculates predicted values ​​representing the characteristics of the composite object obtained by combining the plurality of component objects.

[0070] The information processing method according to one aspect of the present invention is executed by an information processing system having at least one processor. The information processing method includes the following steps: obtaining a numerical representation and a composite ratio for each of a plurality of component objects; performing machine learning based on the plurality of numerical representations to calculate a plurality of regression parameters corresponding to the plurality of component objects; and applying the plurality of composite ratios to a regression model defined by the plurality of regression parameters to calculate a predicted value representing the characteristics of the composite object obtained by combining the plurality of component objects.

[0071] The information processing program involved in one aspect of the present invention causes a computer to perform the following steps: obtaining a numerical representation and a composite ratio for each of a plurality of component objects; performing machine learning based on the plurality of numerical representations to calculate a plurality of regression parameters corresponding to the plurality of component objects; and applying the plurality of composite ratios to a regression model defined by the plurality of regression parameters to calculate a predicted value representing the characteristics of the composite object obtained by combining the plurality of component objects.

[0072] In this approach, machine learning is performed based on the data for each component object to calculate multiple regression parameters corresponding to the multiple component objects. The composite ratio is then applied to a regression model defined by these regression parameters to predict the characteristics of the composite object. By using machine learning and regression models, the accuracy of composite object analysis can be improved even when sufficient data for each component object is unavailable.

[0073] Once the regression parameters are obtained, the composite ratio can be changed, and the characteristics of the composite object can be instantly recalculated using the regression model. In other words, the calculated regression parameters can be reused. By employing a method of inferring regression parameters through machine learning, it is possible to rapidly perform the process of exploring the characteristics of the composite object while changing the composite ratio.

[0074] In another information processing system, at least one processor can input multiple numerical representations into a first machine learning model to calculate multiple feature vectors corresponding to multiple component objects, and then input these feature vectors into a second machine learning model to calculate multiple regression parameters. Through this series of steps, the analytical accuracy of composite objects can be further improved even when sufficient data for the component objects is unavailable.

[0075] In another information processing system, the first machine learning model may include a machine learning model for embedding functions and a machine learning model for interaction functions. At least one processor can input multiple numerical representations into the machine learning model for embedding functions to calculate multiple first feature vectors corresponding to multiple component objects, input these first feature vectors into the machine learning model for interaction functions to calculate multiple second feature vectors corresponding to the multiple component objects, and input these second feature vectors into a second machine learning model to calculate multiple regression parameters. By constructing the first machine learning model in this way, the analytical accuracy of composite objects can be further improved even when a sufficient amount of data for the component objects is unavailable.

[0076] In another approach to information processing systems, the machine learning model used for the embedding function can be a model that generates a first feature vector as a fixed-length vector based on the numerical representation of undefined data. By using this machine learning model, the feature vector can be obtained from a numerical representation that cannot be represented by a fixed-length vector.

[0077] In another type of information processing system, the regression model can be a Scheffe polynomial. At least one processor can compute multiple regression coefficients of the first-degree term of the Scheffe polynomial as multiple regression parameters. By using the Scheffe polynomial, which is frequently used in problems related to fusion, it is possible to analyze composite objects obtained by fusion of multiple component objects with high precision. Furthermore, based on the regression coefficients of the first-degree term, predicted values ​​considering the individual influence of each component object can be calculated.

[0078] In another type of information processing system, at least one processor can also calculate multiple regression coefficients of the quadratic term of the Scheffe polynomial as multiple regression parameters. In this case, it is possible to calculate a predicted value that further considers the combined influence of the two component objects based on the regression coefficients of the quadratic term.

[0079] In another approach to information processing systems, the component object can be a material, and the composite object can be a multi-component substance. In this case, even when sufficient data on the material cannot be prepared, the analytical accuracy of multi-component substances can be improved.

[0080] In another approach to information processing systems, the material can be a polymer or monomer, and multi-component substances can be polymer alloys. In this case, even when sufficient data for the polymer or monomer is unavailable, the analytical accuracy of the polymer alloy can be improved. Polymers and monomers are very diverse, and correspondingly, the variety of polymer alloys is also vast. For such polymers, monomers, and polymer alloys, experiments are generally only conducted on a subset of desirable combinations, thus often resulting in insufficient data. According to this approach, even under such data limitations, polymer alloys can be analyzed with high precision.

[0081] [Variation Example]

[0082] The present invention has been described in detail above according to its embodiments. However, the present invention is not limited to the above embodiments. Various modifications can be made to the present invention without departing from its spirit.

[0083] The processing steps of the information processing method executed by at least one processor are not limited to the examples in the above embodiments. For example, some of the above steps or processes may be omitted, or the steps may be executed in a different order. Furthermore, any two or more of the above steps may be combined, or a part of a step may be modified or deleted. Alternatively, other steps may be performed in addition to the steps described above.

[0084] When comparing the magnitude of two values ​​within an information processing system, either the "above" or "greater than" benchmark can be used, or either the "below" or "less than" benchmark can be used. This choice of benchmark does not change the technical significance of processing the comparison of the magnitude of the two values.

[0085] In this invention, the statement "at least one processor executes the first process, executes the second process, ... executes the nth process," or its corresponding statement, represents the concept of a situation where the processor executing n processes from the first process to the nth process changes midway. That is, this statement represents the concept of two situations: one where all n processes are executed by the same processor, and the other where the processor changes in any direction among the n processes.

[0086] Symbol Explanation

[0087] 10-Information processing system, 11-Acquisition unit, 12-Computing unit, 13-Forecasting unit.

Claims

1. An information processing system comprising at least one processor, The at least one processor performs the following processing: For each of a plurality of polymers or monomers, obtain a numerical representation of the properties of that polymer or monomer and a composite ratio as a proportion among the plurality of polymers or monomers. The multiple numerical representations are input into the first machine learning model to calculate multiple feature vectors corresponding to the multiple polymers or monomers. The multiple feature vectors are input into the second machine learning model to calculate multiple regression parameters. The multiple composite ratios are applied to a regression model defined by the multiple regression parameters to calculate predicted values ​​representing the properties of the polymer alloy obtained by composited with the multiple polymers or monomers. The calculated regression parameters are reused to perform a process that explores the properties of the polymer alloy while varying the multiple composite ratios.

2. The information processing system according to claim 1, wherein, The first machine learning model includes a machine learning model for embedding functions and a machine learning model for interaction functions. The at least one processor performs the following processing: The multiple numerical representations are input into the embedding function using a machine learning model to calculate multiple first feature vectors corresponding to the multiple polymers or monomers. The plurality of first feature vectors are input into the interaction function using a machine learning model to calculate the plurality of second feature vectors corresponding to the plurality of polymers or monomers. The plurality of second feature vectors are input into the second machine learning model to calculate the plurality of regression parameters.

3. The information processing system according to claim 2, wherein, The embedding function is a machine learning model that generates the first feature vector as a fixed-length vector based on the numerical representation of the undefined data.

4. The information processing system according to any one of claims 1 to 3, wherein, The regression model is a Scheffe multinomial. The at least one processor calculates multiple regression coefficients of the first-degree term of the Scheffe polynomial as the multiple regression parameters.

5. The information processing system according to claim 4, wherein, The at least one processor also calculates multiple regression coefficients of the quadratic term of the Scheffe polynomial as the multiple regression parameters.

6. An information processing method, executed by an information processing system having at least one processor, the information processing method comprising the following steps: For each of a plurality of polymers or monomers, obtain a numerical representation of the properties of that polymer or monomer and a composite ratio as a proportion among the plurality of polymers or monomers; The multiple numerical representations are input into the first machine learning model to calculate multiple feature vectors corresponding to the multiple polymers or monomers; The multiple feature vectors are input into the second machine learning model to calculate multiple regression parameters; The multiple composite ratios are applied to a regression model defined by the multiple regression parameters to calculate predicted values ​​representing the properties of the polymer alloy obtained by composite of the multiple polymers or monomers. and The calculated regression parameters are reused to perform a process that explores the properties of the polymer alloy while varying the multiple composite ratios.

7. An information processing program that causes a computer to perform the following steps: For each of a plurality of polymers or monomers, obtain a numerical representation of the properties of that polymer or monomer and a composite ratio as a proportion among the plurality of polymers or monomers; The multiple numerical representations are input into the first machine learning model to calculate multiple feature vectors corresponding to the multiple polymers or monomers; The multiple feature vectors are input into the second machine learning model to calculate multiple regression parameters; The multiple composite ratios are applied to a regression model defined by the multiple regression parameters to calculate predicted values ​​representing the properties of the polymer alloy obtained by composite of the multiple polymers or monomers. and The calculated regression parameters are reused to perform a process that explores the properties of the polymer alloy while varying the multiple composite ratios.

Images