Learning device, learning method, and learning program

Abstract

[Problem] To provide a learning device, a learning method, and a learning program that appropriately perform model learning. [Solution] A learning device (1) is a device for learning and making an appropriate determination in a machine learning model (M) that calculates an evaluation value for target data and determines that the target data is a non-defective product (first label) or a defective product (second label). A learning device (1) is provided with a data acquisition unit (2), an addition processing unit (3), a control unit (4), an extraction unit (5), a learning unit (6), a display unit (7), and a tag changing unit (8). The processed data is teaching data (11) to which the first tag or the second tag is assigned, teaching candidate data (12) to be candidates for the teaching data (11), and altered teaching candidate data (13) to which the addition processing unit (3) executes the addition processing. A data acquisition unit (2) acquires data serving as a matrix (14), an addition processing unit (3) adds the data, and an extraction unit (5) extracts extracted data (15). Before the extraction processing is performed by the extraction unit (5), the data is added by the addition processing unit (3) and learning can be performed.

Description

Technical Field

[0001] The invention relates to a learning device, a learning method and a learning program. Background Art

[0002] In the past, in order to utilize the learning results obtained by deep learning using a neural network, image processing technology has been used to determine whether a product is good or bad. In addition, in order to improve the efficiency of learning performed by deep learning, Patent Document 1 proposes a learning support device for appropriately learning a model, a learning device, etc.

[0003] The learning support device of Patent Document 1 is based on a model that uses and learns teacher data to classify target data into either a first label or a second label, and teacher data having first data assigned a first label and second data assigned a second label, and derives a feature value of the teacher data for each teacher data.

[0004] In addition, the same device is provided with: an export unit that exports the feature value of the professor candidate data for each professor candidate data based on the model and at least one professor candidate data respectively assigned with either the first label or the second label; a calculation unit that calculates at least one of the distance between the professor candidate data and the first data and the distance between the professor candidate data and the second data for each professor candidate data; and a selection unit that selects data to be added as professor data from the professor candidate data based on the distance.

[0005] In the learning support device of Patent Document 1, professor candidate data whose distance is closest to professor data of a second label is selected from a plurality of professor candidate data classified as a first label, and professor candidate data whose distance is closest to professor data of a first label is selected from a plurality of professor candidate data classified as a second label.

[0006] Through this processing, it becomes possible to select the professor candidate data closest to the second label among the professor candidate data as the first label, that is, it becomes possible to select the professor candidate data that is difficult to identify as the first label or the second label. The same is true for the professor candidate data for the second label. In the learning support device of Patent Document 1, by converting such difficult-to-identify professor candidate data into professor data, the quality of the professor data can be improved, thereby enabling the model to be learned efficiently. Prior Art Literature Patent Literature

[0007] Patent document 1: Japanese Patent Publication No. 2021-103344 Summary of the invention Problems to be solved by the invention

[0008] The learning support device described in Patent Document 1 enables efficient learning. Incidentally, in general learning methods, it is known that if the teaching data is increased by a specified method, the learning effect can be improved. Therefore, in the learning support device described in Patent Document 1, it is also expected to improve the learning effect by increasing the teaching data.

[0009] However, in the learning support device described in Patent Document 1, it is known that when the selected teaching data is increased, even if the maximum number of epochs is reached, the end condition of learning is not satisfied, or the learning time becomes too long.

[0010] In view of the above problems, an object of the present invention is to provide a learning device, a learning method, and a learning program that can appropriately perform model learning. Means for solving problems

[0011] In order to achieve the above-mentioned purpose, the learning device of the present invention is characterized by a learning device that enables a model to learn, wherein the model uses teaching data for learning to calculate an evaluation value for target data, and then compares the evaluation value with a specified threshold value and classifies it as at least one of the first label or the second label, and the learning device comprises: a data acquisition unit that acquires professor candidate data that is classified as at least one of the first label or the second label; an addition processing unit that changes part of the content of the professor candidate data to generate one or more changed professor candidate data; an extraction unit that performs extraction processing, wherein the extraction processing is to convert the professor candidate data and the changed professor candidate data into a new one; and The teaching data is used as a matrix, and the evaluation value is calculated for the matrix using the model, and then the data whose evaluation value falls within the specified extraction range including the threshold is extracted; a learning unit causes the model to perform a learning process at a specified learning rate; and a control unit controls the extraction unit and the learning unit; the control unit appends the extraction data to the teaching data and performs the learning process and the extraction process when the number of extracted data extracted as a result of the extraction process performed by the extraction unit exceeds a specified stop number, and the learning process and the extraction process are performed until the number of new extracted data reaches the stop number.

[0012] The learning device of the present invention is provided with an addition processing unit, which changes part of the content of the professor candidate data to generate one or more changed professor candidate data; the matrix used to extract the professor candidate data includes not only the professor candidate data but also the changed professor candidate data. Since the addition of data performed by the addition processing unit is performed before the learning process and the extraction process, the number of extracted data is finally made to reach the specified stop number by repeating the learning process and the extraction process, thereby ending the learning. Therefore, compared with the previous learning device, the learning device of the present invention can improve the learning efficiency by using more matrices for the model.

[0013] In the learning device of the present invention, the control unit may also be configured to, when performing extraction processing using a model that has completed learning, create a new matrix by removing the extracted data from the matrix, and perform the extraction processing on the new matrix by the extraction unit. According to this structure, since the number of matrix data is reduced each time the learning processing and the extraction processing are repeated, the processing burden of the control unit can be reduced.

[0014] The learning device of the present invention may also perform a limiting process in one extraction process when the candidate professor data that is the basis for the change has a plurality of common candidate professor data. The limiting process is to extract only the candidate professor data or the candidate professor data for change that is a specified limit. According to this structure, since the number of candidate professor data or the candidate professor data for change can be suppressed when the learning process is performed, the learning process can be performed quickly.

[0015] In addition, the learning device of the present invention may further include: a display unit for displaying the extracted data, and a label changing unit for changing the label of the extracted data displayed on the display unit; the label can be changed by the label changing unit before the extracted data is added to the teaching data. According to this structure, since the user can confirm the judgment result of the label in the extraction unit through the label changing unit and the input unit, and can change it through the label changing unit when the judgment result is incorrect, the judgment accuracy of the extraction unit can be improved.

[0016] In addition, in the learning device of the present invention, the control unit may also, when the number of the extracted data exceeds the stop number and does not reach the specified benchmark data number, perform an expansion process to expand the amplitude of the extraction range and perform the extraction process, and when the number of the extracted data becomes greater than or equal to the benchmark data number, perform the learning process.

[0017] When the number of extracted data is less than the number of reference data, if the user changes the label through the label changing unit, the frequency of label changes increases, which places a burden on the user. In the present invention, since the learning process and the extraction process are performed only after the number of extracted data becomes greater than or equal to the number of reference data, the frequency of label selection by the user through the selection unit can be reduced.

[0018] In addition, in the learning device of the present invention, the control unit may also expand the amplitude of the extraction range and reduce the learning rate as the expansion process. In this way, by expanding the extraction range, the amount of extracted data can be increased. In addition, the learning rate is an adjustment value for the amount of weight or deviation in the neural network updated in one learning. In the expansion process of the present invention, the amount of extracted data is fine-tuned by reducing the learning rate when the extraction range is expanded.

[0019] In addition, the learning device of the present invention is a learning device that enables a model to learn, wherein the model uses teaching data for learning to calculate an evaluation value for target data, and then compares the evaluation value with a specified threshold value and classifies it into at least one of the first label or the second label. The learning device comprises: a data acquisition unit that acquires teaching candidate data that is classified as at least one of the first label or the second label; an extraction unit that performs extraction processing, wherein the extraction processing is to use the teaching candidate data as a matrix, calculate an evaluation value for the matrix using the model, and then extract data whose evaluation value falls within a specified extraction range that includes the threshold value. data; a learning unit, which makes the model perform learning processing at a specified learning rate; and a control unit, which controls the extraction unit and the learning unit; the control unit, when the number of extracted data extracted by the extraction unit exceeds the specified stop number and does not reach the specified benchmark data number, performs an expansion process to expand the amplitude of the extraction range and performs an extraction process, when the number of the extracted data becomes greater than or equal to the benchmark data number, appends the extracted data to the teaching data and performs the learning process and performs the extraction process, and performs the learning process and the extraction process until the number of new extracted data reaches the stop number.

[0020] According to the learning device constructed as above, when the number of extractions obtained by the extraction unit is less than the designated reference data number, the extraction range is expanded to increase the number of extractions. Since the number of extractions increases, the learning process can be performed efficiently.

[0021] The learning method of the present invention is a learning method for making a model learn, wherein the model uses teaching data for learning to calculate an evaluation value for target data, and then compares the evaluation value with a specified threshold value and classifies it into at least one of the first label or the second label. The learning method comprises the following steps: a data acquisition step, acquiring professor candidate data that is classified as at least one of the first label or the second label; an additional processing step, changing part of the content of the professor candidate data to generate one or more changed professor candidate data; an extraction step, using the professor candidate data and the changed professor candidate data as A matrix, and using the model to calculate an evaluation value for the matrix, and then extracting data whose evaluation value falls within a specified extraction range including the threshold; and a learning step, learning the model at a specified learning rate; wherein when the number of extracted data extracted by the extraction step exceeds a specified stop number, the extracted data is appended to the teaching data to perform the learning step, and a new matrix is ​​created by removing the extracted data from the matrix, and then the extraction step is performed on the new matrix, and the learning step and the extraction step are performed until the number of new extracted data reaches the stop number.

[0022] The learning program of the present invention is a program for causing a computer to function as each of the above-mentioned learning devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The explanatory diagram of FIG. 1 discloses the functional structure of a learning device as an example of an embodiment of the present invention. Figure 2 The flowchart of FIG. 1 discloses the learning method of the learning device of this embodiment. Figure 3 The explanatory diagram of FIG. 1 is a schematic diagram of the extraction process performed by the extraction unit of the learning device of this embodiment. Figure 4 The explanatory diagrams show the display screens displayed by the display unit of the learning device of this embodiment, where (A) shows the situation of good products and (B) shows the situation of defective products. Figure 5 The explanatory diagrams disclose the addition processing performed by the addition processing unit of the learning device of this embodiment, (A) discloses the flip processing, (B) discloses the shift processing, (C) discloses the micro-rotation processing, and (D) discloses the filter processing. Figure 6 The graph reveals the relationship between the learning time and the amount of data of the learning device of this embodiment. DETAILED DESCRIPTION

[0024] Next, refer to Figures 1 to 6 , describing the learning device, learning method and learning program of the implementation form of the present invention. Figure 1 The explanatory diagram of FIG. 1 discloses the functional structure of the learning device 1 of this embodiment.

[0025] The learning device 1 of the present embodiment is a device for learning a model M, wherein the model M calculates an evaluation value for target data and determines whether the target data is classified into at least a first label or a second label so that the model M can make an appropriate determination. Here, the so-called model refers to a structure in which a computer receives a certain input value that can be identified, makes a certain evaluation or determination, and then outputs an output value.

[0026] The model M has a structure including a neural network and parameters. The neural network has a structure combining a plurality of neurons. As an example, the neural network may be a hierarchical multilayer neural network that connects a plurality of neurons that are grouped.

[0027] Neural networks are defined by the number of neurons and their connection relationships. Parameters (weight coefficients, etc.) are used to define the connection strength between neurons or layers. In a neural network, target data is input, and the target data is evaluated and labeled based on the calculation results and parameters of multiple neurons.

[0028] The model M identifies the content of the target data and implements at least the determination of the first label or the second label. For example, when the target data is image data, the type of the subject (people, vehicles, animals, etc.) or the quality of the subject (good products, bad products, etc.) can be used as a label. The label is stored in association with the target data. In addition, since the structure of the model M is the same as that of Patent Document 1, a detailed description is omitted.

[0029] Next, refer to Figure 1 The functional structure of the learning device 1 of this embodiment is described below. As a functional structure, the learning device 1 of this embodiment includes: a data acquisition unit 2, an addition processing unit 3, a control unit 4, an extraction unit 5, a learning unit 6, a display unit 7, and a label change unit 8.

[0030] The target data used by the learning device 1 of this embodiment is the professor data 11 to which either the first label or the second label is assigned, the professor candidate data 12 which is a candidate for the professor data 11, and the changed professor candidate data 13 which has been added by the addition processing unit 3. Like the professor data 11, the professor candidate data 12 and the changed professor candidate data 13 are also assigned either the first label or the second label.

[0031] Here, the teaching data refers to data of example questions and answers presented to the model M. Moreover, the teaching candidate data refers to data that can become a candidate of new teaching data given to the model M.

[0032] The data acquisition unit 2 is a functional unit that performs a data acquisition procedure for acquiring the teaching candidate data 12 so that the model M can be learned by the learning device 1 .

[0033] The addition processing unit 3 is a functional unit that changes part of the content of the professor candidate data 12 to generate one or more changed professor candidate data 13. For example, when the professor candidate data 12 is image data, the following processing can be cited as the addition processing.

[0034] For example, flip processing is performed to create data that reverses the original image data left-right, reverses the data up-down, or rotates the data by 180 degrees; or shift processing is performed to create data that slightly moves the coordinates of the original image data in the XY directions; or slight rotation processing is performed to rotate the original image data to the right and left at a slight angle; or filter processing such as a central value filter or a Gaussian filter is performed on the original image data.

[0035] The control unit 4 is a functional unit that controls the extraction unit 5 and the learning unit 6 and performs effective learning processing on the model M. The extraction unit 5 is a functional unit that performs extraction processing, and the extraction processing is to use the teaching candidate data 12 and the changed teaching candidate data 13 as the matrix 14, and use the model M to calculate the evaluation value for the matrix 14, and then extract data whose evaluation value falls within the specified extraction range including the threshold value as the first label and the second label. The learning unit 6 is a functional unit that performs learning processing on the model M at a specified learning rate. The learning processing is a process of adjusting so that the parameters of the model M are close to the optimal value.

[0036] The display unit 7 is a functional unit that causes a display device such as a display screen to display data so that a user who determines the target data can confirm the extracted data 15 extracted by the extracting unit 5. The label changing unit 8 is a functional unit that can change the label assigned to the extracted data 15 when the user confirms and needs to change the first label or the second label assigned to the extracted data 15.

[0037] The learning device 1 of this embodiment has a computer as the main hardware (not shown). The computer has a processor such as a CPU and a GPU, a storage device such as a RAM, a ROM, a hard disk or an SSD (solid state drive), and a communication unit connected to a network such as the Internet. In addition, the storage device of the computer stores a learning program for operating the learning device 1 of this embodiment. In addition, the computer includes a cloud computing system. The above-mentioned functional units are realized by the computer as hardware and the learning program as software.

[0038] Next, refer to Figure 2, taking the case where the target data is an image as an example, the learning method of the learning device 1 of this embodiment is described. First, the data acquisition unit 2 performs an acquisition step (STEP 1) of acquiring the professor candidate data 12. The professor candidate data 12 is assigned any one of the first label and the second label in advance. The professor candidate data 12 of the number required for the model M learning is prepared in advance.

[0039] Next, the addition processing unit 3 performs an addition processing step (STEP 2) of performing an addition processing on the professor candidate data 12 and generating the changed professor candidate data 13. As the addition processing, it can be performed by any one of the above-mentioned flip processing, shift processing, micro-rotation processing and filter processing or a combination of these processing. Which processing to perform can be appropriately selected according to the nature of the target data, the learning purpose of the model M, or the content determined by the model M.

[0040] Next, the control unit 4 confirms whether it is the first processing (STEP3). If it is the first processing (YES in STEP3), the learning rate is set to the initial value for the model M that has completed learning and is in the initial state (STEP5). The initial value of the learning rate can be set to 0.001, for example. In addition, the extraction range of the extraction unit 5 is set to the initial value (STEP5). The initial value of the extraction range can be set to 3, for example. In addition, the content of the extraction range is described later.

[0041] Next, the extraction unit 5 performs extraction processing (extraction step) (STEP 6). In the extraction processing, the professor candidate data 12 and the changed professor candidate data 13 are used as the matrix 14, the data of the matrix 14 is input into the model M, and the evaluation value is calculated in the model M. If the evaluation value is within the extraction range, it is extracted as the extracted data 15, and if the evaluation value is outside the extraction range, it is not extracted.

[0042] Figure 3 The following is a schematic diagram of the extraction process in the learning device 1 of this embodiment. In the extraction process of the extraction unit 5, the evaluation value is calculated for each data of the matrix 14 using the model M. Figure 3 In the figure, the evaluation value is higher as one moves to the right, and lower as one moves to the left. In this embodiment, the one with high evaluation is listed as the first label, and the one with low evaluation is listed as the second label.

[0043] exist Figure 3 In , the boundary line B between the first label and the second label is the threshold that separates the two labels, and the specified range including the threshold is used as the extraction range. Figure 3 As shown in FIG. 1 , if the evaluation value is located in the area close to the threshold, even if it is the first label, it is still very close to the second label, resulting in a small difference between good and bad products. Figure 3The value of the extraction range is set to 3, but the value may be set to 1 to 20. This value may be changed appropriately according to the extraction conditions or the object of learning.

[0044] In this way, if the model M is trained using data located in the designated area of ​​the first label and the second label, it is possible to train with data that is difficult to distinguish as the first label or the second label. The model M trained with such data can correctly judge subtle differences. In other words, such data is excellent data that enables the model M to learn efficiently.

[0045] The content of the extraction process in the extraction unit 5 is the same as the process described in Patent Document 1. Specifically, the professor candidate data 12 and the changed professor candidate data 13 are input into the model M, and based on the model M and the professor data 11, a feature quantity (vector) represented by a feature space of a predetermined dimension is calculated for each data. In addition, the distance between the professor data 11 and the professor candidate data 12 and the changed professor candidate data 13 in the feature space is calculated respectively. If the distance falls within the extraction range, the data is extracted, and if it falls outside the extraction range, the data is not extracted. In addition, since the method for calculating the distance in the feature space is the same as that in Patent Document 1, a detailed description is omitted.

[0046] In the learning device 1 of this embodiment, when the extraction processing is performed by the extraction unit 5, if the professor candidate data 12 that serves as the basis for changing the professor candidate data 13 in the matrix 14 has a plurality of common data, a limiting processing is performed to extract only a specified limit number of data in one extraction processing (STEP 6).

[0047] By performing this limiting process, when the model M is trained with new professor data 11 in STEP 5 described later, the teaching data 11 does not increase rapidly, so the learning process can be performed quickly. In this embodiment, the limit extraction number is set to 1. Therefore, in this embodiment, even if a plurality of changed professor candidate data 13 are generated from the same professor candidate data 12, only one professor candidate data 12 or changed professor candidate data 13 is extracted in one extraction process. This limit extraction number can be appropriately changed according to the state of the model M or the state of hardware, etc.

[0048] Next, the control unit 4 checks the number of extracted data 15 extracted by the extraction unit 5 (STEP 7). Specifically, as (1), it is checked whether the number x of extracted data 15 exceeds the specified stop number and is greater than or equal to the specified reference data number, or whether the extraction range is an expansion range (in Figure 2 In STEP 7, "reference data number" is indicated as "reference data number". Alternatively, in STEP 7, as (2), it is confirmed whether the extracted data 15 exceeds the stop number and does not reach the reference data number.

[0049] Here, the stop number is a number used as a reference for whether to stop the learning performed by the learning device 1, and can be set to, for example, 0. In this case, when the data extracted by the extraction unit 5 is 0, the learning is stopped.

[0050] The reference data number is a number used to reduce the frequency of the user's confirmation operation (STEP 8 to 9) described later, and for example, the first label can be set to 20 and the second label can be set to 20. The reference data number can be appropriately changed according to factors such as the type of data or the learning environment.

[0051] In this case, when the number of the extracted data 15 extracted by the extracting unit 5 is less than 20 for the first tag or less than 20 for the second tag, the user confirmation operation is not performed, but the extraction process is performed again, and the user confirmation operation is performed only when the number of the extracted data 15 becomes greater than or equal to 40. This process can reduce the frequency of the user's confirmation operation, thereby reducing the burden on the user.

[0052] When the number of extracted data 15 exceeds the stop number and is equal to or greater than the reference data number (1 in STEP 7), the extracted data 15 is displayed on the display unit 7 (STEP 8). Figure 4 The state of the extracted data 15 is displayed on the display unit 7. On the display screen 16, a change display 18 for confirming whether the label 17 given to the extracted data 15 by the label changing unit 8 is appropriate and for changing the label 17 is displayed.

[0053] exist Figure 4 In (A), a message "Is it really a good product?" is displayed at the top of the display screen 16, the extracted data 15 is displayed in the center of the display screen 16, a label 17 as "good product" is displayed on the left side of the extracted data 15, and three buttons "Yes", "No", and "Unknown" are displayed as a change display 18 at the bottom of the display screen 16. By the user clicking (tapping) the button of the change display 18, it can be confirmed whether the label 17 (in this case, "good product") assigned to the extracted data is appropriate, and if it is not appropriate, the label 17 is changed.

[0054] Figure 4 (B) is a case where the label 17 is "defective product", and an image determined to be a defective product in the extracted data 15 is displayed. In addition, when the user selects "unknown" in the change display 18, the user's determination is retained so that a determination can be made later.

[0055] After the user confirms the extracted data 15, the control unit 4 stores the extracted data 15 as the teaching data 11, and uses the new teaching data 11 for the next learning of the label 17. In addition, the control unit 4 removes the extracted data 15 from the matrix 14 and creates a new matrix 14 (STEP 10).

[0056] Next, it is confirmed again whether it is the first processing (STEP3). Since the first processing has been completed and it is the second or later processing (NO in STEP3), the control unit 4 uses the model M to which the new teaching data 11 has been added to perform learning processing (learning step) (STEP4). Since the newly added teaching data 11 is the extracted data 15 with high learning effect extracted by the extraction unit 5, the accuracy of the evaluation performed by the model M is improved. Thereafter, while setting the learning rate to the initial value, the extraction range in the extraction unit 5 is set to the initial value (STEP5).

[0057] The control unit 4 repeats the above extraction process and learning process until the number of the extracted data 15 reaches 0, which is the stop number. When the number of the extracted data 15 reaches 0, which is the stop number (3 in STEP 7), the process ends.

[0058] On the other hand, when the extracted data 15 exceeds the stop number but does not reach the reference data number (2 in STEP7), an expansion process is performed (STEP11) in which the width of the extraction range is expanded as an expansion range (for example, 3.7) and the learning rate is reduced (for example, 0.0002). If the extraction process (STEP6) is performed in this state, the width of the extraction range is wide, so that the number of data extracted by the extraction unit 5 increases. In addition, the learning rate is fine-tuned by this extraction process.

[0059] Next, the control unit 4 checks the number of extracted data (STEP 7). In this case, since the extraction range is changed to the expanded range (1 in STEP 7) by STEP 11, the extracted data extracted by the extraction process is displayed and checked (STEP 8 to 9). When the extraction range is the expanded range, the expansion process of STEP 11 has been executed, so even if the number of extracted data 15 does not reach the reference data number, it is expected that there is data close to the reference data number.

[0060] Next, refer to Figure 5 , a specific example of the addition processing performed by the addition processing unit 3 is described. Figure 5 (A) is an explanatory diagram showing the state of the flipping process, in which the image data of the top-down flipping, the image data of the left-right flipping, and the image data of the 180° rotation are generated from the original image of the candidate teaching data 12. By performing the flipping process, in addition to the candidate teaching data 12, three modified candidate teaching data 13 can be obtained.

[0061] Figure 5 (B) is an explanatory diagram showing the state of the shift processing. For the total of 4 data, namely the teaching candidate data 12 as the original image data and the 3 changed teaching candidate data 13 generated by the flip processing, the image is slightly moved in the form of (-1,-1), (-1,1)... in the XY direction. Through this shift processing, 32 image data can be obtained. If it is an image, the unit of the shift can be pixel (pixel), mm, or μm. In addition, the range of the shift can be appropriately changed according to the purpose of learning, etc.

[0062] Figure 5 (C) is an explanatory diagram showing a state where a micro-rotation process has been performed, where image data rotated 1° to the left and image data rotated 1° to the right are generated for image data without rotation. The micro-rotation angle can be appropriately changed according to the purpose of learning, etc.

[0063] Figure 5 (D) is an explanatory diagram showing the state of the filter processing, in which data of the image data without the filter is generated after the median filter is applied and data of the Gaussian filter is applied. The median filter is also called the median filter, and is a filter used to remove noise from the image. The Gaussian filter is a type of smoothing filter, and is a filter that blurs and smoothes the image according to the Gaussian distribution. In addition, other known filters can also be used as the filter processing.

[0064] If the addition processing performed by the addition processing unit 3 is only Figure 5 (A) is flipped, the number of data is quadrupled. Figure 5 The shift processing of (B) will become 32 times the data, and if the 4 data without shift processing are added, it will become 36 times the data. Figure 5 (C) The 32-fold data that has been shifted is further multiplied by 3 times to become 96 data, and if the initial 4 data are added, it becomes 100-fold data. Figure 5 The filter processing of (D) triples the 100-fold data to 300-fold.

[0065] As described above, in the learning device 1 of this embodiment, since a large amount of modified professor candidate data 13 similar to the professor candidate data 12 can be generated by the addition process, a large amount of data with high learning effect can be extracted by the extraction process.

[0066] Next, refer to Figure 6 The effects of the learning device 1 according to this embodiment will be described. Figure 6This is a graph showing the relationship between the number of target data and the learning time. In the graph, the data shown by the solid line is the learning device 1 of this embodiment, and the data shown by the dotted line is the existing learning device (comparative example). Figure 6 As shown in the graph of , the learning device 1 of this embodiment can also end the learning when the target data number is 300 times. On the other hand, in the comparative example, when the increase multiple of the target data number exceeds 8 times, the learning cannot be ended.

[0067] As described above, compared with the conventional learning apparatus, the learning apparatus 1 of the present embodiment can improve the quality of learning by increasing the target data, and can still finish the learning even when the target data increases.

[0068] Furthermore, in the above-described embodiment, when the number of extracted data 15 exceeds the stop number and is equal to or greater than the reference data number (1 in STEP 7), the extracted data 15 is displayed on the display unit 7 and the user performs a confirmation operation. However, the extracted data 15 may be added to the teaching data 11 without displaying the data on the display unit 7. For example, in the addition process, if the modified teaching candidate data 13 created based on the teaching candidate data 12 is displayed and the confirmation operation is performed, the display on the display unit 7 may be omitted.

[0069] In addition, in the above embodiment, although the target data is described as image data, the target data may also be sound data, chart data, or movie data. In addition, in the above embodiment, although the number of confirmed stops is set to 0 in STEP 7, it is not limited to this, and may be other numbers (for example, integers such as 1 or 10).

[0070] In addition, in the above embodiment, the example of model M determining whether the subject is a good product or a defective product and being labeled as the first label or the second label is used for explanation. However, the same embodiment as above can also be applied to model M for determining the type of the subject (people, vehicles, animals, etc.).

[0071] For example, it may be: in the case where the types of the photographed objects are classified into a model of label A, label B, and label C, in STEP 7, for label A, label B, and label C, when the number of extractions of each label exceeds the benchmark data number, STEP 8 to STEP 10 are processed, and when the number of extractions of each label exceeds the stop number and does not reach the benchmark data number, STEP 11 is processed.

[0072] In addition, in the above embodiment, although flip processing, shift processing, micro-rotation processing and filter processing are cited as the additional processing, it is not limited to this, and various processing such as adding impulse noise, contrast adjustment, brightness adjustment, expansion, reduction, partial masking, cropping, deformation, or color change can also be performed. Therefore, the amount of added data can be appropriately changed and is not limited to 300 times of the above embodiment.

[0073] In addition, in the above embodiment, an example of using a model M that has been learned to perform learning is disclosed, but this is not limited to this, and the above process may also be performed using a model M that has not been learned. Figure 2 As a pre-processing of STEP 1, the unlearned model M is trained using the teaching data for learning, and then the processing after STEP 1 can be performed.

[0074] In the above embodiment, when the extraction range is the expanded range, in STEP 7, the extraction data extracted by the extraction process is displayed and confirmed by moving to the side (1) (STEP 8 to 9), but the present invention is not limited to this. The extraction range may be set as the expanded range again (STEP 11) to move to the side (2) and the extraction process may be performed (STEP 6). In this case, the value of the expanded range may be changed, or the number of extraction processes (STEP 6) may be limited. Description of Reference Numerals

[0075] M…Model 1…Learning device 2…Data acquisition department 3…Add processing unit 4…Control Unit 5…Extraction 6…Learning Department 7…Display unit 8…Label Change Department 11…Teaching Data 12…Professor candidate data 13…Change professor candidate data 14…Mother 15…Extract data 16…Display screen 17… Tags 18…Change display

Claims

1. A learning device, characterized in that: The learning device learns in a model, the model uses teaching data to learn, so as to calculate an evaluation value for the target data, and then compares the evaluation value with a specified threshold value and classifies it into at least one of a first label or a second label, and the learning device has: A data acquisition unit that acquires professor candidate data classified into at least one of the first label and the second label; An additional processing unit is provided to change part of the content of the professor candidate data to generate one or more changed professor candidate data; an extraction unit that performs an extraction process, wherein the extraction process uses the professor candidate data and the changed professor candidate data as a matrix, calculates an evaluation value for the matrix using the model, and then extracts data whose evaluation value falls within a specified extraction range including the threshold value; A learning unit that causes the model to perform a learning process at a specified learning rate; as well as A control unit, controlling the extraction unit and the learning unit; The control unit is: When the number of extracted data as a result of the extraction process performed by the extraction unit exceeds a specified stop number, the extracted data is added to the teaching data to perform the learning process and the extraction process, and the learning process and the extraction process are performed until the number of new extracted data reaches the stop number.

2. The learning device according to claim 1, characterized in that When performing the extraction process using the learned model, the control unit creates a new matrix by removing the extracted data from the matrix, and performs the extraction process on the new matrix by the extraction unit.

3. The learning device according to claim 1, characterized in that: In the extraction process, when the professor candidate data to be changed has a plurality of common changed professor candidate data, a limitation process is performed in one extraction process to extract only the professor candidate data or the changed professor candidate data of a specified limit extraction number.

4. The learning device according to claim 1, characterized in that: Also available: a display unit for displaying the extracted data, and a label changing unit for changing the label of the extracted data displayed on the display unit; Before the extracted data is added to the teaching data, the label may be changed by the label changing unit.

5. The learning device according to claim 4, characterized in that: The control unit performs an expansion process to expand the width of the extraction range and performs the extraction process when the number of the extracted data exceeds the stop number and does not reach a specified reference data number. When the number of the extracted data becomes equal to or larger than the number of the reference data, the learning process is performed.

6. The learning device according to claim 5, characterized in that The control unit increases the width of the extraction range and decreases the learning rate as the expansion process.

7. A learning device for enabling a model to learn, characterized in that: The model is trained using teaching data to calculate an evaluation value for the target data, and then the evaluation value is compared with a specified threshold and classified into at least one of a first label or a second label, and the learning device comprises: A data acquisition unit that acquires professor candidate data classified into at least one of the first label and the second label; an extraction unit that performs an extraction process, wherein the extraction process is to use the professor candidate data as a matrix, calculate an evaluation value for the matrix using the model, and then extract data whose evaluation value falls within a specified extraction range including the threshold value; A learning unit that causes the model to perform a learning process at a specified learning rate; as well as A control unit, controlling the extraction unit and the learning unit; The control unit is: When the number of extracted data extracted by the extraction unit exceeds the specified stop number and does not reach the specified reference data number, an expansion process is performed to expand the width of the extraction range and perform extraction process. When the number of the extracted data becomes equal to or greater than the reference data number, the extracted data is added to the teaching data to perform the learning process and the extraction process, and the learning process and the extraction process are performed until the number of new extracted data reaches the stop number.

8. A learning method, characterized in that A learning method for enabling a model to learn, wherein the model uses teaching data to learn to calculate an evaluation value for target data, and then compares the evaluation value with a specified threshold and classifies it into at least one of a first label or a second label, wherein the learning method comprises the following steps: A data acquisition step of acquiring professor candidate data classified into at least one of the first label or the second label; Adding a processing step to change part of the content of the professor candidate data to generate one or more changed professor candidate data; an extraction step, taking the candidate professor data and the candidate professor change data as a matrix, calculating an evaluation value for the matrix using the model at a specified learning rate, and then extracting data whose evaluation value falls within a specified extraction range including the threshold value; and a learning step, allowing the model to learn; When the number of extracted data extracted by the extraction step exceeds the specified stop number, the extracted data is added to the teaching data to perform the learning step, and a new matrix is ​​created by removing the extracted data from the matrix, and then the extraction step is performed on the new matrix, and the learning step and the extraction step are performed until the number of new extracted data reaches the stop number.

9. A learning program, characterized in that A learning program for causing a computer to function as the learning device according to any one of claims 1 to 7.

Images