Estimation method, information processing apparatus, and program

Abstract

To improve the technique for estimating physical property information, and the like, of a sample, based on a spectrum measurement value related to the sample.SOLUTION: An estimation method to be executed by an information processing apparatus includes the steps of: acquiring a spectrum measurement value related to a sample; converting the spectrum measurement value into one-dimensional data; performing max pooling using multiple different window sizes to reduce the data size of the one-dimensional data to 1 / α(α is a natural number), performing expansion processing to multiply α by the processed one-dimensional data reduced to 1 / α, and generating two-dimensional data by combining the expanded multiple pieces of data; determining data on a predetermined region, out of the two-dimensional data, as input information; estimating information on the sample by inputting the data on the predetermined region, as input information, to a prediction model; and outputting the information on the sample.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present disclosure relates to an estimation method, an information processing apparatus, and a program.

Background Art

[0002] Conventionally, a technique for estimating information such as physical properties of a sample based on a spectrum measurement value related to the sample has been known. For example, Non-Patent Document 1 discloses the GAWLS method as a new variable selection method that enables high-accuracy modeling even in spectrum data.

Prior Art Documents

Patent Documents

[0003]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in actual measurement work, there are cases where signal points shift and signal intensities vary. For example, due to inevitable mechanical errors or the like, the peak position of the spectrum measurement value may shift. In this case, it has been difficult to estimate information related to the sample by the GAWLS method. Thus, there has been room for improvement in the technique for estimating the physical property information of a sample based on the spectrum measurement value.

[0005] In view of such circumstances, an object of the present disclosure is to improve the technique for estimating the physical property information of a sample based on the spectrum measurement value related to the sample.

Means for Solving the Problems

[0006] (1) The estimation method according to an embodiment of the present disclosure is an estimation method executed by an information processing apparatus, comprising: acquiring a spectral measurement value related to a sample; converting the spectral measurement value into one-dimensional data; performing max pooling processing with a plurality of different window sizes that make the data size of the one-dimensional data 1 / α times (α is a natural number), performing an expansion process that multiplies the plurality of processed one-dimensional data by α, and generating two-dimensional data by combining the plurality of expanded data; determining, as input information, data in a predetermined region among the two-dimensional data; inputting the data in the predetermined region as input information into a prediction model to estimate information of the sample; outputting the information of the sample; and including.

[0007] (2) The estimation method according to an embodiment of the present disclosure is the estimation method described in (1), wherein the spectral measurement value includes an IR spectrum.

[0008] (3) The estimation method according to an embodiment of the present disclosure is the estimation method described in (1) or (2), wherein the information of the sample includes at least any one of the composition information and physical property information of the sample.

[0009] (4) The estimation method according to an embodiment of the present disclosure is the estimation method described in any one of (1) to (3), wherein the step of generating the two-dimensional data is a step of generating a two-dimensional image, generating a one-dimensional image in which pixel information corresponding to the measurement values is arranged in the order of wave numbers based on the spectral measurement value; Perform max-pooling processing using the plurality of different window sizes, perform processing to multiply by α so that the processed image multiplied by 1 / α by the max-pooling processing has the same size as the size of the one-dimensional image, and generate a plurality of processed images from the one-dimensional image. Generate a two-dimensional image by combining the one-dimensional image and the plurality of processed images with the frequency as the X-axis and the window size as the Y-axis. including.

[0010] (5) The estimation method according to an embodiment of the present disclosure is the estimation method described in (4), The predetermined region is a plurality of rectangular regions in the two-dimensional image.

[0011] (6) The estimation method according to an embodiment of the present disclosure is the estimation method described in (5), In the step of determining, the plurality of rectangular regions are determined by a genetic algorithm, the chromosome information in the genetic algorithm includes a plurality of gene information corresponding to the plurality of rectangular regions, and each gene information includes X coordinate information and Y coordinate information indicating a rectangular region.

[0012] (7) An information processing apparatus according to an embodiment of the present disclosure An information processing apparatus including a control unit, The control unit Obtain the spectrum measurement value related to the sample Convert the spectrum measurement value into one-dimensional data, Perform max-pooling processing using a plurality of different window sizes that reduce the data size of the one-dimensional data to 1 / α (α is a natural number), perform an enlargement process of multiplying the plurality of processed one-dimensional data reduced to 1 / α by α, and generate two-dimensional data by combining the enlarged plurality of data. Determine the data in a predetermined region among the two-dimensional data as input information, Input the data in the predetermined region as input information into a prediction model to estimate the information of the sample, Output the information of the sample.

[0013] (8) A program according to an embodiment of the present disclosure is a program for estimation executed by an information processing apparatus, which causes a computer to acquire spectral measurement values related to a sample, convert the spectral measurement values into one-dimensional data, perform max-pooling processing with a plurality of different window sizes that make the data size of the one-dimensional data 1 / α times (α is a natural number), perform an enlargement process that multiplies the plurality of processed one-dimensional data that have been made 1 / α times by α, and generate two-dimensional data by combining the plurality of enlarged data, determine, as input information, data in a predetermined region among the two-dimensional data, input the data in the predetermined region as input information into a prediction model to estimate information of the sample, output the information of the sample, and execute.

Advantages of the Invention

[0014] According to the estimation method, information processing apparatus, and program according to an embodiment of the present disclosure, it is possible to improve the technology for estimating the physical properties information, etc. of a sample based on the spectral measurement values related to the sample.

Brief Description of the Drawings

[0015]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7A

Figure 7B

Figure 8A

Figure 8B

Figure 9

Figure 10

Mode for Carrying Out the Invention

[0016] Hereinafter, the estimation technique according to the embodiment of the present disclosure will be described with reference to the drawings.

[0017] In each figure, the same or corresponding parts are denoted by the same reference numerals. In the description of this embodiment, the description of the same or corresponding parts will be omitted or simplified as appropriate.

[0018] First, the outline of this embodiment will be described. The estimation technique according to the embodiment of the present disclosure is executed by the information processing apparatus 10. The information processing apparatus 10 first acquires the spectral measurement values related to the sample. Further, the information processing apparatus 10 converts the spectral measurement values into one-dimensional data. The information processing apparatus 10 performs max-pooling processing using a plurality of different window sizes that multiply the data size of the one-dimensional data by 1 / α. Here, α is a natural number. The information processing apparatus 10 performs an enlargement process of multiplying by α the plurality of one-dimensional data (hereinafter, also referred to as processed one-dimensional data) that have been reduced to 1 / α by the max-pooling process, and generates two-dimensional data by combining the enlarged plurality of data. The information processing apparatus 10 determines, as input information, the data in a predetermined area among the two-dimensional data. The information processing apparatus 10 inputs the data in the predetermined area as input information to the prediction model to estimate the information of the sample. Then, the information processing apparatus 10 outputs the information of the sample.

[0019] As described above, according to the estimation technique according to this embodiment, the information processing apparatus 10 estimates the information of the sample based on the two-dimensional data generated from the spectral measurement values. The two-dimensional data is data obtained by combining one-dimensional data generated by performing max-pooling processing using a plurality of different window sizes. Since max-pooling processing is performed using a plurality of different window sizes, such two-dimensional data includes local features and global features related to the spectral measurement values. Therefore, according to the estimation technique according to this embodiment, the technique of estimating the physical property information etc. of the sample based on the spectral measurement values related to the sample is improved in that it is possible to perform estimation considering the local features and global features related to the spectral measurement values.

[0020] (Configuration of Information Processing Apparatus) Next, each configuration of the information processing apparatus 10 will be described in detail. The information processing apparatus 10 is an arbitrary apparatus used by a user. For example, a personal computer, a server computer, a general-purpose electronic device, or a dedicated electronic device can be adopted as the information processing apparatus 10.

[0021] As shown in FIG. 1, the information processing apparatus 10 includes a control unit 11, a storage unit 12, an input unit 13, an output unit 14, and a communication unit 15.

[0022] The control unit 11 includes at least one processor, at least one dedicated circuit, or a combination thereof. The processor is a general-purpose processor such as a CPU (central processing unit) or a GPU (graphics processing unit), or a dedicated processor specialized for specific processing. The dedicated circuit is, for example, an FPGA (field-programmable gate array) or an ASIC (application specific integrated circuit). The control unit 11 executes processing related to the operation of the information processing apparatus 10 while controlling each unit of the information processing apparatus 10.

[0023] The storage unit 12 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or a combination of at least two of these. The semiconductor memory is, for example, a RAM (random access memory) or a ROM (read only memory). The RAM is, for example, an SRAM (static random access memory) or a DRAM (dynamic random access memory). The ROM is, for example, an EEPROM (electrically erasable programmable read only memory). The storage unit 12 functions as, for example, a main storage device, an auxiliary storage device, or a cache memory. The storage unit 12 stores data used for the operation of the information processing apparatus 10 and data obtained by the operation of the information processing apparatus 10.

[0024] The input unit 13 includes at least one input interface. The input interface is, for example, a physical key, a capacitive key, a pointing device, or a touch screen provided integrally with a display. The input interface may also be, for example, a sound sensor that receives voice input, or a camera that receives gesture input. The input unit 13 receives an operation for inputting data used for the operation of the information processing apparatus 10. Instead of being provided in the information processing apparatus 10, the input unit 13 may be connected to the information processing apparatus 10 as an external input device. As a connection method, for example, any method such as USB (Universal Serial Bus), HDMI (Registered Trademark) (High-Definition Multimedia Interface), or Bluetooth (Registered Trademark) can be used.

[0025] The output unit 14 includes at least one output interface. The output interface is, for example, a display that outputs information as video. The display is, for example, an LCD (liquid crystal display) or an organic EL (electro luminescence) display. The output unit 14 performs a display output of data obtained by the operation of the information processing apparatus 10. Instead of being provided in the information processing apparatus 10, the output unit 14 may be connected to the information processing apparatus 10 as an external output device. As a connection method, for example, any method such as USB, HDMI (Registered Trademark), or Bluetooth (Registered Trademark) can be used.

[0026] The communication unit 15 includes at least one interface for external communication. The communication interface may be either a wired communication interface or a wireless communication interface. In the case of wired communication, the communication interface may be an interface corresponding to a communication standard such as, for example, a LAN (Local Area Network) interface, a USB (Universal Serial Bus), an HDMI (registered trademark) (High-Definition Multimedia Interface), etc. In the case of wireless communication, the communication interface may be an interface corresponding to a mobile communication standard such as, for example, LTE (Long Term Evolution), 4G (4th generation), or 5G (5th generation), or an interface corresponding to a short-range wireless communication such as Bluetooth (registered trademark). The communication unit 15 receives data used for the operation of the information processing apparatus 10 and transmits data obtained by the operation of the information processing apparatus 10.

[0027] The functions of the information processing apparatus 10 are realized by executing the program according to this embodiment on a processor corresponding to the information processing apparatus 10. That is, the functions of the information processing apparatus 10 are realized by software. The program causes a computer to execute the operations of the information processing apparatus 10, thereby causing the computer to function as the information processing apparatus 10. That is, the computer functions as the information processing apparatus 10 by executing the operations of the information processing apparatus 10 according to the program.

[0028] In this embodiment, the program can be recorded on a computer-readable recording medium. The computer-readable recording medium includes non-transitory computer-readable media, such as a magnetic recording device, an optical disk, a magneto-optical recording medium, or a semiconductor memory. The distribution of the program can be performed, for example, by selling, transferring, or lending a portable recording medium such as a DVD (Digital Versatile Disc) or a CD-ROM (Compact Disc Read Only Memory) on which the program is recorded. The distribution of the program may also be performed by storing the program in the storage of an external server and transmitting the program from the external server to another computer. The program may also be provided as a program product.

[0029] Some or all of the functions of the information processing apparatus 10 may be realized by a dedicated circuit corresponding to the control unit 11. That is, some or all of the functions of the information processing apparatus 10 may be realized by hardware.

[0030] With reference to the flowchart of FIG. 2, an estimation method according to an embodiment of the present disclosure is shown.

[0031] Step S101: The control unit 11 of the information processing apparatus 10 acquires a spectrum measurement value. The spectrum measurement value may be, for example, an IR spectrum, and in this embodiment, the spectrum measurement value is described as being an IR spectrum. Any method can be adopted to acquire the spectrum measurement value. For example, the control unit 11 may acquire the spectrum measurement value by receiving the spectrum measurement value collected by an external device via the communication unit 15.

[0032] Step S102: The control unit 11 converts the spectral measurement values into one-dimensional data. Here, the one-dimensional data may be, for example, a one-dimensional image in which the pixel information of 1 pixel corresponding to the measurement values is arranged in the order of signal points such as frequency. In the present embodiment, the one-dimensional data will be described as such a one-dimensional image. In the one-dimensional image, the X-axis (horizontal axis) represents the frequency, and the Y-axis (vertical axis) direction is represented by the image information of 1 pixel. In other words, in the present embodiment, the one-dimensional image is a rectangular image with a short side of 1 pixel. Fig. 3 shows a schematic diagram of the process of converting the spectral measurement values into one-dimensional data. In Fig. 3, the spectral measurement values 200 are converted into corresponding one-dimensional data 300 (corresponding one-dimensional image 300). The spectral measurement values 200 have the X-axis as the frequency and the Y-axis as the measurement values. In the conversion process, the measurement values corresponding to each frequency are converted into corresponding pixel information. For example, the pixel information may be luminance, and may be luminance values from 0.0 (black) to 1.0 (white). For example, in the conversion process, the control unit 11 normalizes the measurement values corresponding to each frequency and uses the normalized values as the luminance values. The control unit 11 generates a one-dimensional image 300 by arranging the pixel information of each 1 pixel corresponding to the measurement values in the X-axis direction in the order of frequency. Note that the one-dimensional data is not limited to this, and may be vector data in which the measurement values corresponding to each frequency are arranged in the order of frequency, etc. If the one-dimensional data is vector data, the two-dimensional data may be matrix data.

[0033] Step S103: The control unit 11 generates two-dimensional data based on the one-dimensional data. In the present embodiment, since the one-dimensional data is a one-dimensional image, the two-dimensional data is a two-dimensional image. Specifically, the control unit 11 performs max-pooling processing with a plurality of different window sizes that reduce the data size of the one-dimensional data by 1 / α times, performs a process of enlarging the plurality of processed one-dimensional data reduced by 1 / α times by α times, and generates two-dimensional data by combining the enlarged data. As described above, α is an arbitrary natural number.

[0034] Here, the data size of one-dimensional data is the number of numerical values included in the one-dimensional data. When the one-dimensional data is represented by a one-dimensional image, the data size of such one-dimensional data corresponds to the image size. In other words, in the present embodiment, the data size of the one-dimensional image is represented by the number of pixels in the X-axis direction. Reducing the data size of the one-dimensional data by a factor of 1 / α corresponds to reducing the number of numerical values included in the one-dimensional data by a factor of 1 / α. Reducing the data size by a factor of 1 / α is equivalent to performing max pooling processing on the one-dimensional data with a window size that reduces the data size by a factor of 1 / α. Specifically, for example, when the one-dimensional data is "0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0" and α is 2, the data becomes the data in which only the maximum value of two adjacent data in the one-dimensional data remains. Specifically, the processed data is "0.2, 0.4, 0.6, 0.8, 1.0". In this way, by the max pooling processing, the number of one-dimensional data changes from 10 to 5, which is 1 / α times (here, 1 / 2 times) the original number.

[0035] Further, the control unit 11 performs a process of expanding the data size of the one-dimensional data (processed one-dimensional data) reduced by a factor of 1 / α by a factor of α. Specifically, when the above "0.2, 0.4, 0.6, 0.8, 1.0" is the processed one-dimensional data, the expansion process is performed by arranging α consecutive numerical values for each. For example, when α is 2, two consecutive data are arranged for each. Therefore, the expanded one-dimensional data becomes "0.2, 0.2, 0.4, 0.4, 0.6, 0.6, 0.8, 0.8, 1.0, 1.0". In this way, in the present embodiment, by expanding the data size of the data after the max pooling processing, the data size is made the same as that of the data before the max pooling processing.

[0036] FIG. 4 is a schematic diagram of a process for generating two-dimensional data based on one-dimensional data. Here, four different window sizes, α = 2, 4, 8, and 16, are used. In the example shown in FIG. 4, first, the control unit 11 performs max-pooling processing with a window size that halves the one-dimensional image 300. In such max-pooling processing, α = 2. The control unit 11 generates a processed image 501 after max-pooling processing based on the maximum value among each window 401 in such processing. The size of the processed image 501 in the X-axis direction is 1 / α times that of the original image. Therefore, the control unit 11 performs a process of expanding the processed image 501 α times in the X-axis direction so that it has the same size as the size of the original image 300 in the X-axis direction. That is, here, the control unit 11 performs a process of doubling the size in the X-axis direction so that the processed image 501 has the same size as the image size of the one-dimensional image 300. Through such processing, the control unit 11 generates a processed image 601.

[0037] Subsequently, in the example shown in FIG. 4, the control unit 11 performs max-pooling processing with a window size that quarters the processed image 601. In such max-pooling processing, α = 4. The control unit 11 generates a processed image 502 after max-pooling processing based on the maximum value among each window 402 in such processing. The size of the processed image 502 in the X-axis direction is 1 / α times that of the original image. Therefore, the control unit 11 performs a process of expanding the processed image 502 α times in the X-axis direction so that it has the same size as the size of the processed image 601 in the X-axis direction. That is, here, the control unit 11 performs a process of quadrupling the size in the X-axis direction so that the processed image 502 has the same size as the image size of the one-dimensional image 300. Through such processing, the control unit 11 generates a processed image 602. Here, max-pooling processing is performed with a window size that quarters the processed image 601, but it is not limited to this. The control unit 11 may perform max-pooling processing with a window size that quarters the original one-dimensional image 300.

[0038] Similarly, in the example shown in FIG. 4, the control unit 11 performs max-pooling processing with a window size that reduces the processed image 602 (or the one-dimensional image 300) to 1 / 8 its original size, and performs an enlargement process on the processed one-dimensional data to generate a processed image 603. Further, the control unit 11 performs max-pooling processing with a window size that reduces the processed image 603 (or the one-dimensional image 300) to 1 / 16 its original size, and performs an enlargement process on the processed one-dimensional data to generate a processed image 604.

[0039] Subsequently, the control unit 11 combines the one-dimensional image 300, the processed images 601, 602, 603, and 604 to generate a two-dimensional image 700. Specifically, the control unit 11 generates the two-dimensional image 700 by connecting these images adjacent to each other in the Y-axis direction in ascending order of the window size of the max-pooling process. In other words, the control unit 11 combines the one-dimensional image and the plurality of processed images with the frequency on the X-axis and the window size on the Y-axis to generate a two-dimensional image.

[0040] Step S104: The control unit 11 determines, as input information, data in a predetermined region among the two-dimensional data. When the two-dimensional data is a two-dimensional image, the predetermined region is a plurality of rectangular regions in the two-dimensional image.

[0041] FIG. 5 shows a schematic view of a predetermined region. In FIG. 5, the predetermined region is the rectangular region 710, the rectangular region 720, the rectangular region 730, the rectangular region 740, and the rectangular region 750 among the two-dimensional data 700 (two-dimensional image 700). In other words, the data of these plurality of regions are the feature amounts of the two-dimensional data. In the present embodiment, the control unit 11 determines the predetermined region by a genetic algorithm. That is, the plurality of rectangular regions described above are determined by a genetic algorithm. The chromosome information in the genetic algorithm includes a plurality of gene information corresponding to the plurality of rectangular regions. Each gene information includes X coordinate information and Y coordinate information indicating the rectangular region. The X coordinate information and Y coordinate information indicating the rectangular region may be, for example, the coordinates of the vertices on the diagonal of the rectangular region. For example, the gene information corresponding to the rectangular region 710 in FIG. 5 may be the coordinates of the vertices on the diagonal of the rectangular region 710. Specifically, the gene information of the rectangular region 710 may be the X coordinate information and Y coordinate information of the vertices 711 and 712. Similarly, the gene information of the rectangular region 720 may be the X coordinate information and Y coordinate information of the vertices 721 and 722.

[0042] FIG. 6 shows a part of the chromosome information of the predetermined region (hereinafter, also referred to as 2D-chromosome). The 2D-chromosome includes a plurality of gene information corresponding to the plurality of rectangular regions. Specifically, as described above, the gene information includes X coordinate information and Y coordinate information indicating the rectangular region. In FIG. 6, only the first two gene information (here, the gene information corresponding to the rectangular region 710 and the rectangular region 720) are shown. As shown in FIG. 6, the gene information corresponding to the rectangular region 710 is the X coordinate information and Y coordinate information (125, 1) of the vertex 711, and the X coordinate information and Y coordinate information (130, 4) of the vertex 712. The gene information corresponding to the rectangular region 720 is the X coordinate information and Y coordinate information (492, 1) of the vertex 721, and the X coordinate information and Y coordinate information (502, 2) of the vertex 722.

[0043] The determination of a predetermined area by a genetic algorithm may be executed by any method. Specifically, the chromosome information of the initial population is randomly generated, and the fitness is calculated using an evaluation function to optimize the chromosome information. Specifically, selection and elimination related to the chromosome information and gene manipulation processing are performed, and the fitness is calculated using the evaluation function for the chromosome after the gene manipulation processing, and optimization is executed again. By executing such processing over multiple generations, optimal chromosome information is determined. There are two types of gene manipulation processing: crossover and mutation.

[0044] FIG. 7A shows a conceptual diagram of the crossover of 2D-chromosomes. FIG. 7B shows a conceptual diagram of the replacement of chromosome information in the crossover of 2D-chromosomes. For example, in the chromosomes of both parents, when they contain gene #α1, #β1, and #γ1, and gene #α2, #β2, and #γ2 respectively, the gene information of the chromosomes of the two children of the parents may be the result of swapping the parts of gene #γ1 and gene #γ2 of the parents.

[0045] FIG. 8A shows a conceptual diagram of the mutation of 2D-chromosomes. FIG. 8B shows a conceptual diagram of the change of chromosome information in the mutation of 2D-chromosomes. For example, when the chromosome before mutation contains gene #α1, #β1, and #γ1, the gene information of the chromosome after mutation may be the result of randomly changing some of the gene information. Specifically, for example, as shown in FIG. 8A, the gene information of the chromosome after mutation may be the result of changing gene #β1 to gene #β5.

[0046] Step S105: The control unit 11 inputs the input information determined in step S104 into the prediction model to estimate the information of the sample. The prediction model may be any learning model generated based on machine learning algorithms such as linear regression, logistic regression, support vector machine (SVM), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and other deep learning.

[0047] Step S106: The control unit 11 outputs the information of the estimated sample. For example, the control unit 11 outputs the information of the estimated sample by causing the output unit 14 to display and output a user interface screen including the information of the sample. The information of the sample includes at least one of the composition information of the sample and the physical property information.

[0048] As described above, according to the estimation technique according to this embodiment, the information processing apparatus 10 estimates the information of the sample based on the two-dimensional data generated from the spectrum measurement values. The two-dimensional data is data obtained by combining one-dimensional data generated by performing max-pooling processing with a plurality of different window sizes. Since the max-pooling processing is performed with a plurality of different window sizes, such two-dimensional data includes local features and global features related to the spectrum measurement values. Therefore, according to the estimation technique according to this embodiment, the technique of estimating the physical property information and the like of the sample based on the spectrum measurement values related to the sample is improved in that it is possible to perform estimation considering the local features and global features related to the spectrum measurement values.

[0049] (Example) Hereinafter, an example in which the information processing apparatus 10 according to this embodiment estimates a binary classification as to whether there is a monomer composition of a specified monomer composition % or more from the actually measured IR spectrum values of an acrylic copolymer will be described. In this example, a prediction model was trained using the actually measured IR spectrum values of 400 samples with known monomer ratios as teacher data. Here, styrene (St) with relatively little bias in the distribution of the monomer composition ratio in the copolymers of the 400 samples was used as the analysis target.

[0050] Figure 9 shows the estimation results of St to be analyzed for 35 pieces of test data. The horizontal axis represents the actually measured composition % (correct value), and the vertical axis represents the estimation results. In this example, the estimation result is the prediction score. The prediction score is the probability that the composition % of St is 10% or more. Here, in this example, the upper limit score threshold is set to 0.85. When the prediction score is equal to or higher than the upper limit score threshold, it is defined that styrene is present. On the other hand, when the prediction score is less than the upper limit score threshold, it is determined that styrene is not present. Table 1 shows the estimation results according to the examples.

[0051]

Table 1

[0052] From the results in Table 1, the accuracy rate of the estimation by the information processing apparatus 10 according to this embodiment is 0.914, and the recall rate is 0.857.

[0053] As a comparative example, Figure 10 shows the estimation results for 35 pieces of test data using the conventional one-dimensional data according to the method of Non-Patent Document 1 as teacher data. The horizontal axis represents the actually measured composition % (correct value), and the vertical axis represents the estimation results, which are the prediction scores here. The upper limit score threshold is set to 0.85. Table 2 shows the estimation results according to the comparative example.

[0054]

Table 2

[0055] From the results in Table 2, the accuracy rate according to the comparative example is 0.8, and the recall rate is 0.5. Comparing the example and the comparative example, it can be seen that both the accuracy rate and the precision rate of the estimation technique according to this embodiment are higher than those of the comparative example, indicating that the prediction accuracy has been improved.

[0056] Although the present disclosure has been described based on the drawings and examples, it should be noted that those skilled in the art can easily make various modifications and corrections based on the present disclosure. Therefore, it should be noted that these modifications and corrections are included in the scope of the present disclosure. For example, the functions included in each means or each step, etc. can be rearranged so as not to be logically contradictory, and a plurality of means or steps, etc. can be combined into one or divided.

Explanation of Signs

[0057] 10 Information processing apparatus 11 Control unit 12 Storage unit 13 Input unit 14 Output unit 15 Communication unit 200 Spectrum measurement value 300 One-dimensional data 401 Window 501, 502, Processed image 601, 602, 603, 604 Processed image 700 Two-dimensional image 710, 720, 730, 740, 750 Rectangular region 711, 712, 721, 722 Vertex

Claims

1. An estimation method executed by an information processing apparatus, comprising: a step of obtaining a spectral measurement value related to a sample; a step of converting the spectral measurement value into one-dimensional data; performing max-pooling processing with a plurality of different window sizes that make the data size of the one-dimensional data 1 / α times (α is a natural number), performing an enlargement processing that multiplies the plurality of processed one-dimensional data by α, and generating two-dimensional data by combining the plurality of enlarged data; a step of determining, as input information, data in a predetermined region among the two-dimensional data; a step of inputting the data in the predetermined region as input information into a prediction model and estimating the information of the sample; a step of outputting the information of the sample; An estimation method including the above.

2. The estimation method according to claim 1, wherein the spectral measurement value includes an IR spectrum.

3. The estimation method according to claim 1, wherein the information of the sample includes at least one of the composition information and the physical property information of the sample.

4. The estimation method according to claim 1, wherein the step of generating the two-dimensional data is a step of generating a two-dimensional image, and generating a one-dimensional image in which pixel information corresponding to the measurement value is arranged in the order of wave numbers based on the spectral measurement value; performing max-pooling processing with the plurality of different window sizes, performing a processing of multiplying the processed image reduced by 1 / α times by α so as to have the same size as the one-dimensional image, and generating a plurality of processed images from the one-dimensional image; combining the one-dimensional image and the plurality of processed images with the wave number as the X axis and the window size as the Y axis to generate a two-dimensional image; An estimation method including the above.

5. The estimation method according to claim 4, wherein the predetermined region is a plurality of rectangular regions in the two-dimensional image.

6. The estimation method according to claim 5, wherein in the determining step, the plurality of rectangular regions are determined by a genetic algorithm, the chromosome information in the genetic algorithm includes a plurality of gene information corresponding to the plurality of rectangular regions, and each gene information includes X coordinate information and Y coordinate information indicating the rectangular region.

7. An information processing apparatus including a control unit, wherein the control unit is Obtain the spectral measurement values related to the sample Convert the spectral measurement values into one-dimensional data, Perform max-pooling processing using a plurality of different window sizes that make the data size of the one-dimensional data 1 / α times (α is a natural number), perform an expansion process of multiplying the plurality of processed one-dimensional data that have been made 1 / α times by α, and generate two-dimensional data by combining the plurality of expanded data, Among the two-dimensional data, determine the data in a predetermined area as input information, Input the data in the predetermined area as input information into a prediction model to estimate the information of the sample, An information processing apparatus that outputs the information of the sample.

8. An estimation program executed by an information processing apparatus, which causes a computer to: Obtain the spectral measurement values related to the sample; Convert the spectral measurement values into one-dimensional data; Perform max-pooling processing using a plurality of different window sizes that make the data size of the one-dimensional data 1 / α times (α is a natural number), perform an expansion process of multiplying the plurality of processed one-dimensional data that have been made 1 / α times by α, and generate two-dimensional data by combining the plurality of expanded data; Among the two-dimensional data, determine the data in a predetermined area as input information; Input the data in the predetermined area as input information into a prediction model to estimate the information of the sample; Output the information of the sample; A program for causing the above to be executed.

Images