Learning device, learning method, and learning program

Abstract

A removal unit (133) removes a predetermined component from a frequency component obtained by transforming data of a predetermined region. A calculation unit (134) calculates a loss function on the basis of the result obtained by inputting, to a discriminator that constitutes an adversarial learning model, data in which the frequency component from which the predetermined component has been removed by the removal unit (133) are returned to the predetermined region. An update unit (135) updates a parameter of the adversarial learning model so that the loss function is optimized.

Description

[Technical field]

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

[0002] Conventionally, deep generative models are known that are based on deep learning technology and generate samples that are close to the real thing by learning the distribution of trained data. For example, GANs (Generative Adversarial Networks) are known as deep generative models (see, for example, Non-Patent Document 1).

[0003] Other deep generative models include VAEs (Variational Auto Encoders) (Reference 1: Kingma, Diederik P., and Max Welling. "Auto-encoding variational bayes." arXiv preprint arXiv:1312.6114 (2013). (ICLR 2014)). [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Goodfellow, Ian, et al. "Generative adversarial nets." Advances in neural information processing systems. 2014. (NIPS 2014) Summary of the Invention [Problem to be solved by the invention]

[0005] However, conventional techniques have the problem that overfitting may occur and the accuracy of the model may not improve. For example, samples generated by a trained GAN generator contain high-frequency components that are not included in the actual training data. As a result, the classifier may rely on high-frequency components to determine authenticity, which may cause overfitting. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems and achieve the objective, the learning device is characterized by having a removal unit that removes a specified component from a frequency component obtained by converting data in a specified region, a calculation unit that calculates a loss function based on the result obtained by inputting the data, from which the specified component has been removed by the removal unit, back into the specified region, into a classifier that constitutes an adversarial learning model, and an update unit that updates parameters of the adversarial learning model so that the loss function is optimized. Effect of the Invention

[0007] According to the present invention, it is possible to prevent overlearning and improve the accuracy of the model. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram illustrating a deep learning model according to the first embodiment. [Diagram 2] FIG. 2 is a diagram illustrating the influence of high-frequency components. [Diagram 3] FIG. 3 is a diagram illustrating an example of the configuration of the learning device according to the first embodiment. [Figure 4] FIG. 4 is a diagram for explaining a method for removing high frequency components. [Diagram 5] FIG. 5 is a diagram showing an example of components to be removed. [Figure 6] FIG. 6 is a flowchart showing a flow of processing of the learning device according to the first embodiment. [Figure 7] FIG. 7 is a flowchart showing a flow of processing of the learning device according to the second embodiment. [Figure 8] FIG. 8 shows the results of the experiment. [Figure 9] FIG. 9 shows the results of the experiment. [Figure 10] FIG. 10 shows the results of the experiment. [Figure 11] FIG. 11 is a diagram showing an application example of a filter for removing high frequency components. [Figure 12] FIG. 12 is a diagram illustrating an example of a computer that executes a learning program. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a learning device, a learning method, and a learning program according to the present application will be described in detail with reference to the accompanying drawings. Note that the present invention is not limited to the following embodiments.

[0010] GAN is a technology that learns a data distribution p_data(x) using two deep learning models: a generator G and a discriminator D. G learns to deceive D, and D learns to distinguish between G and the training data. A model in which multiple models are in an adversarial relationship is sometimes called an adversarial learning model.

[0011] Adversarial learning models such as GANs are used in the generation of images, text, and speech.

[0012] Here, GAN has a problem that D overfits the training samples as the learning progresses. As a result, each model cannot meaningfully update the data generation, and the generation quality by the generator deteriorates.

[0013] In addition, Reference 2 describes that the trained CNN output makes predictions depending on the high-frequency components of the input. Reference 2: Wang, Haohan, et al. "High-frequency Component Helps Explain the Generalization of Convolutional Neural Networks." Proceedings of the IEEE / CVF Conference on Computer Vision and Pattern Recognition. 2020.(CVPR 2020)

[0014] [First embodiment] Therefore, in the first embodiment, one object is to prevent overlearning and improve the accuracy of the model by removing high frequency components from data input to the classifier D. Fig. 1 is a diagram for explaining a deep learning model according to the first embodiment. Fig. 2 is a diagram for explaining the influence of high frequency components.

[0015] As shown in Figure 2, the CIFAR-10 (two-dimensional power spectrum) is different between real data (Real) and data generated by a generator (Fake). Reference 3 also shows that data generated by various GANs has an increased power spectrum at high frequencies compared to real data. Reference 3: Durall, Ricard, Margret Keuper, and Janis Keuper. "Watch your Up-Convolution: CNN Based Generative Deep Neural Networks are Failing to Reproduce Spectral Distributions." Proceedings of the IEEE / CVF Conference on Computer Vision and Pattern Recognition. 2020. (CVPR 2020)

[0016] Returning to FIG. 1, in the deep learning model of this embodiment, a discriminator D discriminates which data is Real (or Fake) between data (Real) contained in a real data set X and data (Fake) generated by a generator G from a random number z.

[0017] In GAN, the classifier D is optimized to improve the classification accuracy of the classifier D, i.e., to increase the probability that the classifier D classifies Real as Real. Also, the generator G is optimized to increase the ability of the generator G to deceive the generator G, i.e., to increase the probability that the classifier D classifies Real as Fake.

[0018] In this embodiment, in addition to the above optimization, the generator G is optimized so that the frequency components of Real and Fake match. Hereinafter, the configuration of the learning device of this embodiment and the details of the learning process of the deep learning model will be described.

[0019] [Configuration of the first embodiment] 3 is a diagram illustrating an example of the configuration of a learning device according to the first embodiment. The learning device 10 receives input of learning data and updates parameters of a deep learning model. The learning device 10 may also output the updated parameters. As illustrated in FIG. 3, the learning device 10 includes an input / output unit 11, a storage unit 12, and a control unit 13.

[0020] The input / output unit 11 is an interface for inputting and outputting data. For example, the input / output unit 11 may be a communication interface such as a network interface card (NIC) for performing data communication with other devices via a network. The input / output unit 11 may also be an interface for connecting an input device such as a mouse or keyboard, and an output device such as a display.

[0021] The storage unit 12 is a storage device such as a hard disk drive (HDD), a solid state drive (SSD), an optical disk, etc. The storage unit 12 may be a semiconductor memory in which data can be rewritten, such as a random access memory (RAM), a flash memory, or a non-volatile static random access memory (NVSRAM). The storage unit 12 stores an operating system (OS) and various programs executed by the learning device 10. The storage unit 12 also stores model information 121.

[0022] The model information 121 is information such as parameters for constructing a deep learning model, and is appropriately updated in the learning process. In addition, the updated model information 121 may be output to another device or the like via the input / output unit 11.

[0023] The control unit 13 controls the entire learning device 10. The control unit 13 is, for example, an electronic circuit such as a central processing unit (CPU), a micro processing unit (MPU), or a graphics processing unit (GPU), or an integrated circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA). The control unit 13 also has an internal memory for storing programs that define various processing procedures and control data, and executes each process using the internal memory. The control unit 13 also functions as various processing units by the operation of various programs. For example, the control unit 13 has a generation unit 131, a conversion unit 132, a removal unit 133, a calculation unit 134, and an update unit 135.

[0024] The generating unit 131 inputs a random number z to a generator G to generate data.

[0025] The conversion unit 132 converts the data input to the discriminator D into frequency components. The conversion unit 132 converts real data (Real) and data (Fake) generated by a generator into frequency components.

[0026] The removal unit 133 removes a predetermined component from the frequency components obtained by converting the data in the predetermined region. In this embodiment, the removal unit 133 removes the high frequency components.

[0027] Here, the process of removing high frequency components by the conversion unit 132 and the removal unit 133 will be described with reference to Fig. 4. Fig. 4 is a diagram for explaining a method of removing high frequency components.

[0028] As shown in FIG. 4, in the DCT Layer, the transform unit 132 transforms x real and x fake is converted to frequency components.

[0029] x real is real data, which we call the first data here. Also, x fake is data generated by the generator, and is referred to as second data here. Furthermore, the conversion unit 132 converts the first data into a first frequency component and converts the second data into a second frequency component.

[0030] Next, the removal unit 133 removes (filters, masks) high-frequency components in F-Drop by equation (1). x is the first data x real and the second data x fake F(·) is a function for performing frequency transformation using DFT and DCT.

[0031]

number

[0032] Here, each component of the function M is calculated using equation (2).

[0033]

number

[0034] Here, each data in the frequency space (frequency domain) is represented by coordinates on the u-axis and v-axis. Also, the right-hand side of the inequality in equation (2) is determined according to the data.

[0035] For example, if the first data and the second data are image data, H is the height of the image and W is the width of the image. The height H and width W are expressed, for example, in terms of the number of pixels. In this case, the image data before conversion is data in an RGB space in which each element is expressed by an RGB value.

[0036] If the size of the image data is 5×5, the components shown in FIG. 5 are removed according to equation (1). FIG. 5 is a diagram showing an example of components to be removed. Here, the parameter γ=0.5. In this case, the right-hand side of the inequality in equation (2), which corresponds to the threshold value, is 0.5×(5 2 +5 2 ) 1 / 2 Therefore, the square of the threshold is 12.5.

[0037] The numbers in each box in Figure 5 are the squares of the left side of the inequality in equation (2). For example, if (u,v)=(0,0), then ((u 2 +v 2 ) 1 / 2 ) 2 = 0, which is less than the threshold value of 12.5, so it is not removed. On the other hand, if (u, v) = (2, 3), then ((u 2 +v 2 ) 1 / 2 ) 2 = 13, which is greater than the threshold value of 12.5, so it is removed.

[0038] (u 2 +v 2 ) 1 / 2can be said to be the distance from the origin in the frequency space. Therefore, the removal unit 133 removes components whose distance from the origin in the frequency domain is equal to or greater than a threshold value from the first frequency component obtained by converting the first image data in the RGB space and the second frequency component obtained by converting the second image data in the RGB space generated by the generator constituting the adversarial learning model.

[0039] Furthermore, the transform unit 132 restores the data from which the components have been removed by the removal unit 133 to the space before the transformation. For example, when the transform unit 132 performs transformation using a discrete cosine transform (DCT), the transform unit 132 performs an inverse transformation using an inverse discrete cosine transform (IDCT).

[0040] When the original data is image data in the RGB space, the conversion unit 132 converts the data in the frequency space from which the higher frequency components of the equation (1) have been removed, into data in the RGB space by inverse conversion.

[0041] In this way, the removal unit 133 removes predetermined components from the first frequency component obtained by converting the first data and the second frequency component obtained by converting the second data generated by the generator that constitutes the adversarial learning model.

[0042] The calculation unit 134 inputs the data in which the frequency components from which the predetermined components have been removed by the removal unit 133 have been returned to a predetermined region into a classifier constituting an adversarial learning model, and calculates a loss function based on the results obtained.

[0043] The calculation unit 134 calculates a loss function that increases as the classification accuracy by the classifier decreases, for each of the first frequency component from which the predetermined component has been removed and the data in which the second frequency component has been returned to a predetermined domain.

[0044] The update unit 135 updates the parameters of the adversarial learning model so as to optimize the loss function. For example, the update unit 135 updates the parameters of the generator so as to optimize the loss function.

[0045] For example, the calculation unit 134 and the update unit 135 update the parameters using a loss function used in a known adversarial learning model (GAN).

[0046] [Processing of the first embodiment] Fig. 6 is a flowchart showing a process flow of the learning device according to the first embodiment. As shown in Fig. 6, first, the learning device 10 reads learning data (step S101). Here, the learning device 10 reads real data (Real) as the learning data.

[0047] Next, the learning device 10 samples a random number z from a normal distribution, and creates a sample (Fake) by G(z) (step S102).

[0048] Here, the learning device 10 calculates Drop(Real, γ) and Drop(Fake, γ) and inputs the results to the discriminator D (step S103). The function Drop(·) is as explained in equation (1).

[0049] Here, the learning device 10 calculates the GAN loss function of the generator G (step S104).

[0050] Furthermore, the learning device 10 updates the parameters of the generator G by backpropagation of the global loss (here, the GAN loss function) (step S105).

[0051] Furthermore, the learning device 10 performs learning of the classifier D (step S106).

[0052] At this time, if the maximum number of learning steps is greater than the number of learning steps (step S107, True), the learning device 10 returns to step S101 and repeats the process. On the other hand, if the maximum number of learning steps is not greater than the number of learning steps (step S107, False), the learning device 10 ends the process.

[0053] [Advantages of the First Embodiment] As described above, the removal unit 133 removes a predetermined component from the frequency components obtained by converting the data in the predetermined region. The calculation unit 134 inputs the data obtained by returning the frequency components from which the predetermined component has been removed by the removal unit 133 back to the predetermined region to a classifier constituting an adversarial learning model, and calculates a loss function based on the result. The update unit 135 updates the parameters of the adversarial learning model so as to optimize the loss function.

[0054] As mentioned above, overfitting can occur when the generator G and the discriminator D in the GAN are excessively concentrated on the high-frequency components of the data. For example, if the discriminator D relies on the high-frequency components to make an authenticity judgment, the generator G learns the high-frequency components to deceive the discriminator D. Then, the result of the authenticity judgment becomes dependent only on the high-frequency components, and effective updates to bring the data distribution closer together are no longer performed.

[0055] In response to this, the learning device 10 can perform GAN learning after removing high-frequency components (frequency dropping).

[0056] As a result, according to the present embodiment, it is possible to suppress the deviation (frequency gap) of the frequency components from the training data that occurs during GAN training. Furthermore, since the characteristics of the frequency components are closer, the quality of data generation by the generator G is also improved.

[0057] As described above, according to this embodiment, it is possible to prevent overlearning from occurring and improve the accuracy of the model.

[0058] The removal unit 133 removes a predetermined component from a first frequency component obtained by converting the first data and a second frequency component obtained by converting the second data generated by a generator constituting an adversarial learning model. The calculation unit 134 calculates a loss function that increases as the classification accuracy by the classifier decreases for each of the first frequency component from which the predetermined component has been removed and the data in which the second frequency component has been returned to a predetermined domain. The update unit 135 updates the parameters of the generator so as to optimize the loss function.

[0059] In this way, by removing high frequency components from both real and generated data in GAN, the accuracy of the model can be further improved.

[0060] The removal unit 133 removes components whose distance from the origin in the frequency domain is equal to or greater than a threshold value from a first frequency component obtained by converting first image data in RGB space and a second frequency component obtained by converting second image data in RGB space generated by a generator constituting an adversarial learning model.

[0061] As a result, according to the embodiment, high frequency components can be removed from image data.

[0062] [Second embodiment] The learning device 10 may include in the loss function the frequency component matching loss between the generator G and the discriminator D. In the second embodiment, the learning device 10 optimizes the frequency component matching loss during learning.

[0063] The processes of the generation unit 131 and the conversion unit 132 are similar to those in the first embodiment.

[0064] The calculation unit 134 further calculates an inter-data error between the first frequency component and the second frequency component. The calculation unit 134 can calculate the error by any method such as MSE (Mean Square Error), RMSE (Root Mean Square Error), L1, etc. Here, the calculation unit 134 calculates the error by using L D and L in equation (4) G The calculation unit 134 also calculates the inter-data error L freq (Frequency component matching loss) is calculated using equation (5).

[0065]

number

[0066]

number

[0067]

number

[0068] where X real and X fake are batches of Real and Fake, respectively. Also, |X real | and |X fake | is the batch size of each. Real is real data. And Fake is data generated by generator G.

[0069] Also, F(·) is a function that converts spatial domain data into frequency components. x real i and x fake j are X real The i-th data and X fake The j-th data is an example of the first data and the second data. real i ) corresponds to the first frequency component. fake j ) corresponds to the second frequency component.

[0070] In this way, the calculation unit 134 calculates the error between the batch average of the multiple first frequency components obtained by converting each of the multiple first data and the batch average of the multiple second frequency components obtained by converting each of the multiple second data. In other words, the error here corresponds to the error between the batch averages, not the error between individual data samples.

[0071] Furthermore, the calculation unit 134 calculates a loss function L that increases as the error between the first frequency component and the second frequency component increases and increases as the classification accuracy between the first data and the second data by the classifier constituting the adversarial learning model decreases. Gis calculated as shown in equation (4), where λ is a hyperparameter that functions as a weight.

[0072] G(·) is a function that outputs data (Fake) generated by generator G based on arguments. Also, D(·) is a function that outputs the probability that discriminator D will classify data input as an argument as Real.

[0073] The update unit 135 updates the parameters of the adversarial learning model so that both the loss function and the data error are optimized. Specifically, the update unit 135 updates the parameters of the loss function L G Update the parameters of the generator G so that

[0074] In addition, the update unit 135 updates the loss function L D The parameters of the classifier D are updated so that x is optimized. Here, x is real data (Real).

[0075] [Processing of the second embodiment] Fig. 7 is a flowchart showing a process flow of the learning device according to the second embodiment. As shown in Fig. 7, first, the learning device 10 reads learning data (step S201). Here, the learning device 10 reads real data (Real) as the learning data.

[0076] Next, the learning device 10 samples a random number z from a normal distribution and generates a sample (Fake) by G(z) (step S202). The learning device 10 also converts the Real and Fake into frequency components by DCT or DFT, and calculates the batch average of the frequency components (step S203).

[0077] Here, the learning device 10 calculates Drop(Real, γ) and Drop(Fake, γ) and inputs the results to the discriminator D (step S204). The function Drop(·) is as explained in equation (1).

[0078] The learning device 10 calculates the GAN loss function of the generator G (step S205). The GAN loss of the generator G corresponds to the first term on the right side of equation (4). Then, the learning device 10 calculates the frequency component matching loss from the batch average of the Real-Fake frequency components (step S206). The frequency component matching loss is calculated by L freq is equivalent to.

[0079] Furthermore, the learning device 10 calculates the sum of the GAN loss function for G and the frequency component matching loss as the overall loss (step S207). The overall loss is calculated by L G The learning device 10 may multiply the frequency component matching loss by a weight λ. The learning device 10 updates the parameters of the generator G by backpropagation of the total loss (step S208).

[0080] In addition, the learning device 10 performs learning of the classifier D (step S209). Specifically, the learning device 10 uses the loss function L D The parameters of the classifier D are updated using the backpropagation method.

[0081] At this time, if the maximum number of learning steps>the number of learning steps (step S210, True), the learning device 10 returns to step S101 and repeats the process. On the other hand, if the maximum number of learning steps>the number of learning steps is not satisfied (step S210, False), the learning device 10 ends the process.

[0082] [Effects of the second embodiment] The calculation unit 134 further calculates an inter-data error between the first frequency component and the second frequency component. The update unit 135 updates the parameters of the adversarial learning model so that both the loss function and the inter-data error are optimized.

[0083] This makes it possible to further reduce the influence of frequency components in training the adversarial learning model.

[0084] [experiment] An experiment was conducted by actually implementing the above embodiment, and the experimental setup is as follows. Experimental setup Datasets (Images): CIFAR-10, CIFAR-100, TinyImageNet, STL-10, CelebA, ImageNet CIFAR-10 / -100: 50,000 TinyImageNet, STL-10: 100,000 sheets CelebA: 200,000 ImageNet: 1,300,000 images Neural network architecture: ResNet-SNGAN Experimental Procedure 100,000 iterations of training data Measure generation quality FID every 1,000 iterations The model with the best FID score is the final model. A total of 10 sessions were conducted and the following indicators were used for evaluation: Frequency gap: Difference in frequency components between training data and generated data FID (Reference 4) / KID (Reference 5) / IS (Reference 6): Measures of the quality of generated images Experimental Pattern SNGAN: Baseline (normal GAN) (Reference 7) Binomial: Add a low-pass filter to the existing method 1 Generator (Reference 8) SR: Existing method 2 Minimize the frequency component difference between the generated image and the training image (using 1D DFT and Binary Cross-entropy) (Reference 3) SSD-GAN: Adding a frequency discriminator to the existing method 3 discriminator (Reference 9) F-Drop: First embodiment F-Match: Calculates match loss in the second embodiment and does not perform frequency drop F-Drop&Match: Second embodiment Reference 4: Heusel, Martin, et al. "Gans trained by a two time-scale update rule converge to a local nash equilibrium." Advances in neural information processing systems. 2017. (NeurIPS 2017) Reference 5: Binkowski, Mikolaj, et al. "Demystifying mmd gans." arXiv preprint arXiv:1801.01401 (ICLR 2018). Reference 6: Salimans, Tim, et al. "Improved techniques for training gans." arXiv preprint arXiv:1606.03498 (NeurIPS 2016). Reference 7: Miyato, Takeru, et al. "Spectral normalization for generative adversarial networks." arXiv preprint arXiv:1802.05957 (ICLR 2018). Reference 8: Frank, Joel, et al. "Leveraging frequency analysis for deep fake image recognition." International Conference on Machine Learning. PMLR, 2020. Reference 9: Chen, Yuanqi, et al. "SSD-GAN: Measuring the Realness in the Spatial and Spectral Domains." arXiv preprint arXiv:2012.05535 (AAAI 2021)

[0085] Figures 8 and 9 are diagrams showing the results of the experiment. In Figure 8, the average absolute error of the frequency components is . Figure 9 visualizes the DCT coefficients of each frequency component. As shown in Figures 8 and 9, it can be said that the frequency gap can be reduced by the first embodiment (F-Drop) and the second embodiment (F-Drop & Match). It can be seen from Figure 9 that the second embodiment in particular has frequency components close to the real data (Real).

[0086] Fig. 10 is a diagram showing the results of the experiment, which shows that the generation quality by the generator G is improved by the first and second embodiments.

[0087] Figure 11 shows an example of applying a filter to remove high frequency components. γ is a parameter of the argument of function M. As shown in Figure 9, removing frequencies does not affect the spatial domain, but does affect the frequency domain.

[0088] Thus, high frequency components do not affect the human perception of the image, because the natural image perceived by humans is concentrated in low frequency components.

[0089] [System configuration, etc.] In addition, each component of each device shown in the figure is a functional concept, and does not necessarily have to be physically configured as shown in the figure. In other words, the specific form of distribution and integration of each device is not limited to that shown in the figure, and all or a part of it can be functionally or physically distributed or integrated in any unit according to various loads, usage conditions, etc. Furthermore, each processing function performed by each device can be realized in whole or in any part by a CPU (Central Processing Unit) and a program analyzed and executed by the CPU, or can be realized as hardware by wired logic. Note that the program may be executed not only by the CPU but also by other processors such as a GPU.

[0090] Furthermore, among the processes described in this embodiment, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically by a known method. In addition, the information including the processing procedures, control procedures, specific names, various data and parameters shown in the above documents and drawings can be changed arbitrarily unless otherwise specified.

[0091] [program] In one embodiment, the learning device 10 can be implemented by installing a learning program that executes the above learning process as package software or online software on a desired computer. For example, the above learning program can be executed by an information processing device, causing the information processing device to function as the learning device 10. The information processing device here includes desktop and notebook personal computers. In addition, the information processing device also includes mobile communication terminals such as smartphones, mobile phones, and PHS (Personal Handyphone Systems), as well as slate terminals such as PDAs (Personal Digital Assistants).

[0092] The learning device 10 can also be implemented as a learning server device that provides a service related to the above learning process to a client, the client being a terminal device used by a user. For example, the learning server device is implemented as a server device that provides a learning service that receives learning data as input and outputs information on a trained model. In this case, the learning server device may be implemented as a web server or as a cloud that provides a service related to the above learning process by outsourcing.

[0093] 12 is a diagram showing an example of a computer that executes a learning program. The computer 1000 has, for example, a memory 1010 and a CPU 1020. The computer 1000 also has a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. Each of these components is connected by a bus 1080.

[0094] The memory 1010 includes a read only memory (ROM) 1011 and a random access memory (RAM) 1012. The ROM 1011 stores a boot program such as a basic input output system (BIOS). The hard disk drive interface 1030 is connected to a hard disk drive 1090. The disk drive interface 1040 is connected to a disk drive 1100. A removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1100. The serial port interface 1050 is connected to a mouse 1110 and a keyboard 1120, for example. The video adapter 1060 is connected to a display 1130, for example.

[0095] The hard disk drive 1090 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, a program that defines each process of the learning device 10 is implemented as a program module 1093 in which computer-executable code is written. The program module 1093 is stored, for example, in the hard disk drive 1090. For example, a program module 1093 for executing processes similar to those of the functional configuration of the learning device 10 is stored in the hard disk drive 1090. The hard disk drive 1090 may be replaced by an SSD (Solid State Drive).

[0096] Furthermore, setting data used in the processing of the above-described embodiment is stored as program data 1094, for example, in the memory 1010 or the hard disk drive 1090. Then, the CPU 1020 reads out the program module 1093 or the program data 1094 stored in the memory 1010 or the hard disk drive 1090 into the RAM 1012 as necessary, and executes the processing of the above-described embodiment.

[0097] Note that the program module 1093 and the program data 1094 are not limited to being stored in the hard disk drive 1090, but may be stored in, for example, a removable storage medium and read by the CPU 1020 via the disk drive 1100 or the like. Alternatively, the program module 1093 and the program data 1094 may be stored in another computer connected via a network (such as a local area network (LAN) or wide area network (WAN)). The program module 1093 and the program data 1094 may then be read by the CPU 1020 from the other computer via the network interface 1070. [Explanation of symbols]

[0098] 10 Learning Device 11 Input / output section 12 Storage section 121 Model Information 13 Control section 131 Generation part 132 Conversion unit 133 Removal section 134 Calculation Department 135 Update Department

Claims

1. a removal unit that removes a predetermined component from a frequency component obtained by converting image data of a predetermined region by using a threshold based on a size of the image data; a calculation unit that calculates a loss function based on a result obtained by inputting data obtained by returning the frequency components from which the predetermined components have been removed by the removal unit to a classifier constituting an adversarial learning model, and returning the data to the predetermined region; an update unit that updates parameters of the adversarial learning model so as to optimize the loss function; A learning device comprising:

2. The removal unit removes a predetermined component from a first frequency component obtained by converting first data and a second frequency component obtained by converting second data generated by a generator constituting the adversarial learning model; the calculation unit calculates a loss function that increases as the classification accuracy by the classifier decreases, for each of the first frequency component from which the predetermined component has been removed and the second frequency component returned to the predetermined domain; The learning device according to claim 1 , wherein the update unit updates parameters of the generator so as to optimize the loss function.

3. The learning device according to claim 2, characterized in that the removal unit removes components whose distance from an origin in the frequency domain is equal to or greater than a threshold value for the first frequency component obtained by converting first image data in RGB space and the second frequency component obtained by converting second image data in RGB space generated by a generator that constitutes the adversarial learning model.

4. The calculation unit further calculates an inter-data error between the first frequency component and the second frequency component, 4. The learning device according to claim 2, wherein the update unit updates parameters of the adversarial learning model so as to optimize both the loss function and the inter-data error.

5. A learning method performed by a learning device, comprising: a removing step of removing a predetermined component from a frequency component obtained by converting image data of a predetermined region by using a threshold based on the size of the image data; a calculation step of inputting data obtained by returning the frequency components from which the predetermined components have been removed by the removal step back into the predetermined region to a classifier constituting an adversarial learning model, and calculating a loss function based on the result obtained; updating parameters of the adversarial learning model such that the loss function is optimized; A learning method comprising:

6. A learning program for causing a computer to function as the learning device according to any one of claims 1 to 4.

Images