Information processing device, information processing method, and recording medium

Abstract

An information processing device (10) is provided with: an acquisition means (50) that sequentially acquires a plurality of elements included in series data; a likelihood ratio calculation means (100) that calculates a likelihood ratio representing the likelihood of a class to which the series data belongs on the basis of two or more adjacent elements each time an element is acquired, the likelihood ratio calculation means (100) using a state space model (110) based on a polynomial approximation of the series data when calculating the likelihood ratio; and a classification means (150) that classifies the series data into one of a plurality of classes on the basis of the likelihood ratio. Such an information processing device makes it possible to perform appropriate classification even when the series data includes a large number of elements.

Description

Information processing device, information processing method, and recording medium

[0001] The present disclosure relates to the technical fields of an information processing device, an information processing method, and a recording medium.

[0002] Known examples of this type of device include one that classifies sequential data using likelihood ratios. For example, Patent Document 1 discloses a device that sequentially acquires and analyzes multiple elements included in the sequential data to classify the sequential data into one of multiple predetermined classes.

[0003] International Publication No. 2020 / 194497

[0004] This disclosure aims to improve upon the related art discussed above.

[0005] One aspect of the information processing device disclosed herein includes an acquisition means for sequentially acquiring multiple elements included in sequence data, a likelihood ratio calculation means for calculating, each time an element is acquired, a likelihood ratio indicating the likelihood of the class to which the sequence data belongs based on two or more adjacent elements, the likelihood ratio calculation means using a state space model based on a polynomial approximation of the sequence data when calculating the likelihood ratio, and a classification means for classifying the sequence data into one of multiple classes based on the likelihood ratio.

[0006] One aspect of the information processing method disclosed herein sequentially acquires multiple elements contained in sequence data, and for each acquired element, calculates a likelihood ratio indicating the likelihood of the class to which the sequence data belongs based on two or more adjacent elements, and when calculating the likelihood ratio, uses a state space model based on a polynomial approximation of the sequence data, and classifies the sequence data into one of multiple classes based on the likelihood ratio.

[0007] One aspect of the recording medium of this disclosure has recorded thereon a computer program that causes a computer to execute an information processing method, which includes sequentially acquiring multiple elements included in sequence data, calculating, for each acquired element, a likelihood ratio indicating the likelihood of a class to which the sequence data belongs based on two or more adjacent elements, and classifying the sequence data into one of multiple classes based on the likelihood ratio using a state space model based on a polynomial approximation of the sequence data when calculating the likelihood ratio.

[0008] 1 is a block diagram showing the hardware configuration of a first information processing device; FIG. 2 is a block diagram showing the functional configuration of the first information processing device; FIG. 3 is a flowchart showing the flow of classification operations in the first information processing device; FIG. 4 is a block diagram showing the network structure of a state space model used in the first information processing device; FIG. 5 is a graph showing an example of a likelihood ratio calculated in a second information processing device together with a comparative example; FIG. 6 is a graph showing an example of an activity function used in a third information processing device; and FIG. 7 is a graph showing an example of an activity function used in a fourth information processing device.

[0009] Hereinafter, embodiments of an information processing device, an information processing method, and a recording medium will be described with reference to the drawings.

[0010] First Embodiment A first information processing apparatus will be described with reference to FIGS. 1 to 4. FIG.

[0011] (Hardware Configuration) First, the hardware configuration of the first information processing apparatus will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the hardware configuration of the first information processing apparatus.

[0012] 1, a first information processing device 10 includes a processor 11, a RAM (Random Access Memory) 12, a ROM (Read Only Memory) 13, and a storage device 14. The information processing device 10 may further include an input device 15 and an output device 16. The processor 11, RAM 12, ROM 13, storage device 14, input device 15, and output device 16 are connected to each other via a data bus 17. The data bus 17 may be an interface other than a data bus (for example, a LAN, a USB, etc.).

[0013] The processor 11 loads a computer program. For example, the processor 11 is configured to load a computer program stored in at least one of the RAM 12, the ROM 13, and the storage device 14. Alternatively, the processor 11 may load a computer program stored in a computer-readable storage medium using a storage medium reading device (not shown). The processor 11 may acquire (i.e., load) the computer program from a device (not shown) located outside the information processing device 10 via a network interface. The processor 11 controls the RAM 12, the storage device 14, the input device 15, and the output device 16 by executing the loaded computer program. In particular, in this embodiment, when the processor 11 executes the loaded computer program, functional blocks that execute various processes for classifying sequence data are realized within the processor 11. In other words, the processor 11 may function as a controller that executes each control in the information processing device 10.

[0014] The processor 11 may be configured as, for example, a central processing unit (CPU), a graphics processing unit (GPU), a field-programmable gate array (FPGA), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or a quantum processor. The processor 11 may be configured as one of these, or may be configured to use multiple processors in parallel.

[0015] The RAM 12 temporarily stores computer programs executed by the processor 11. The RAM 12 temporarily stores data that the processor 11 temporarily uses while it is executing the computer programs. The RAM 12 may be, for example, a dynamic random access memory (D-RAM) or a static random access memory (SRAM). Alternatively, other types of volatile memory may be used instead of the RAM 12.

[0016] The ROM 13 stores computer programs executed by the processor 11. The ROM 13 may also store fixed data. The ROM 13 may be, for example, a programmable read-only memory (PROM) or an erasable read-only memory (EPROM). Alternatively, other types of non-volatile memory may be used instead of the ROM 13.

[0017] The storage device 14 stores data that is to be saved long-term by the information processing device 10. The storage device 14 may operate as a temporary storage device for the processor 11. The storage device 14 may include, for example, at least one of a hard disk device, a magneto-optical disk device, an SSD (Solid State Drive), and a disk array device.

[0018] The input device 15 is a device that receives input instructions from a user of the information processing device 10. The input device 15 may include, for example, at least one of a keyboard, a mouse, and a touch panel. The input device 15 may be configured as a mobile terminal such as a smartphone or a tablet. The input device 15 may also be, for example, a device that includes a microphone and is capable of voice input.

[0019] The output device 16 is a device that outputs information related to the information processing device 10 to the outside. For example, the output device 16 may be a display device (e.g., a display) that can display information related to the information processing device 10. The output device 16 may also be a speaker or the like that can output information related to the information processing device 10 as audio. The output device 16 may be configured as a mobile terminal such as a smartphone or a tablet.

[0020] 1 shows an example of an information processing device 10 including multiple devices, but all or some of the functions may be realized by a single device. Such an information processing device may be configured to include only the above-mentioned processor 11, RAM 12, and ROM 13, and the other components (i.e., the storage device 14, input device 15, output device 16, etc.) may be provided by an external device connected to the information processing device 10. Furthermore, some of the calculation functions of the information processing device 10 may be realized by an external device (e.g., an external server, a cloud, etc.).

[0021] (Functional Configuration) Next, the functional configuration of the first information processing device 10 will be described with reference to Fig. 2. Fig. 2 is a block diagram showing the functional configuration of the first information processing device.

[0022] 2, the first information processing device 10 is configured as a device for classifying sequential data. For example, the first information processing device 10 may be configured as a device that acquires images of time-series data and classifies the types of objects included in the images. The first information processing device 10 is configured to include, as components for realizing its functions, an acquisition unit 50, a likelihood ratio calculation unit 100, and a classification unit 150. Each of the acquisition unit 50, the likelihood ratio calculation unit 100, and the classification unit 150 may be a processing block realized by, for example, the above-mentioned processor 11 (see FIG. 1).

[0023] The acquisition unit 50 is configured to acquire sequence data. Sequence data here refers to data including multiple elements arranged in a predetermined order, such as time-series data. More specific examples of sequence data include, but are not limited to, video data, audio data, or subdivided image data. The acquisition unit 50 is configured to sequentially acquire multiple elements included in the sequence data. For example, the acquisition unit 50 may be configured to acquire elements included in the sequence data one by one. The acquisition unit 50 may acquire data directly from any data acquisition device (e.g., a camera, a microphone, etc.), or may read data previously acquired by a data acquisition device and stored in storage, etc. When acquiring data from a camera, the acquisition unit 50 may be configured to acquire data from each of multiple cameras.

[0024] The likelihood ratio calculation unit 100 is configured to calculate a likelihood ratio from the sequence data acquired by the acquisition unit 50. The "likelihood ratio" here is an index indicating to which of multiple candidate classes the sequence data belongs (more specifically, an index indicating the likelihood of the class to which the sequence data belongs). When the acquisition unit 50 sequentially acquires elements, the likelihood ratio calculation unit 100 calculates a likelihood ratio each time an element is acquired. That is, the likelihood ratio calculation unit 100 calculates a likelihood ratio when the acquisition unit 50 acquires an element, and calculates a new likelihood ratio when the acquisition unit 50 acquires another element. The likelihood ratio calculation unit 100 calculates a likelihood ratio based on two or more consecutive elements among the elements sequentially acquired by the acquisition unit 50. For example, the likelihood ratio calculation unit 100 may calculate a likelihood ratio using a newly acquired element and a previously acquired element or a previously calculated likelihood ratio. In this case, the likelihood ratio calculation unit 100 may include a storage unit that stores previously acquired elements or previously calculated likelihood ratios. The likelihood ratio calculation unit 100 according to this embodiment is configured to calculate likelihood ratios using a state space model 110. This state space model 110 is a model based on polynomial approximation of sequence data. Details of the state space model 110 will be described later.

[0025] The classification unit 150 classifies the sequence data into one of a plurality of classes based on the likelihood ratio calculated by the likelihood ratio calculation unit 100. Specifically, the classification unit 150 classifies the sequence data into one of a plurality of classes by comparing the likelihood ratio calculated by the likelihood ratio calculation unit 100 with a preset threshold. For example, the classification unit 150 classifies the sequence data into class A when the likelihood ratio exceeds a threshold corresponding to class A. Furthermore, the classification unit 150 classifies the sequence data into class B when the likelihood ratio exceeds a threshold corresponding to class B. The classification unit 150 may classify, for example, whether a person's face in an image is a real face or a fake face (e.g., impersonation using a photograph or a 3D mask). The plurality of classes may also be three or more. In this case, the classification unit 150 may perform classification using a threshold set for each class.

[0026] (Classification Operation) Next, the classification operation (i.e., the operation of classifying sequence data into one of a plurality of classes) in the first information processing device 10 will be described with reference to Fig. 3. Fig. 3 is a flowchart showing the flow of the classification operation in the first information processing device.

[0027] 3 , when the classification operation in the first information processing device 10 starts, the acquisition unit 50 first acquires one element included in the sequence data (step S101). Then, the likelihood ratio calculation unit 100 calculates a likelihood ratio based on the element acquired by the acquisition unit 50 (step S102). At this time, the likelihood ratio calculation unit 100 calculates the likelihood ratio using the state space model 110.

[0028] Next, the classification unit 150 determines whether there is a class for which the likelihood ratio calculated by the likelihood ratio calculation unit 100 exceeds a threshold value (step S103). That is, the classification unit 150 compares the likelihood ratio calculated by the likelihood ratio calculation unit 100 with the threshold value corresponding to each class that is a classification candidate, and determines whether the likelihood ratio exceeds the threshold value corresponding to any class.

[0029] If there is a class whose likelihood ratio exceeds the threshold (step S103: YES), the classifier 150 classifies the sequence data into the class whose likelihood ratio exceeds the threshold (step S104). On the other hand, if there is no class whose likelihood ratio exceeds the threshold (step S103: NO), the process is repeated from step S101. That is, the acquirer 50 acquires the next element included in the sequence data and repeats the same process as described above.

[0030] The likelihood ratio calculated by the likelihood ratio calculation unit 100 tends to approach the threshold value corresponding to the class to be classified each time an element is acquired. Therefore, by repeating the process while acquiring elements one by one as described above, sequential data can be classified into appropriate classes. In this case, for sequential data that is easy to classify, the threshold value is exceeded at an early stage, allowing the sequential data to be classified early. On the other hand, for sequential data that is difficult to classify, elements continue to be acquired until the threshold value is exceeded, allowing the sequential data to be classified accurately over a period of time.

[0031] (State Space Model) Next, a state space model used when the first information processing device 10 calculates likelihood ratios will be described with reference to Fig. 4. Fig. 4 is a block diagram showing the network structure of the state space model used in the first information processing device.

[0032] As shown in FIG. 4, the state space model 110 used in the first information processing device 10 is configured as a model that outputs an output y each time an input u is sequentially input. ｋ－１ When input is given, the state space model 110 outputs y ｋ－１ At the next time, u ｋ When input is given, the state space model 110 outputs y ｋ At the next time, u ｋ+１ When input is given, the state space model 110 outputs y ｋ+１ where x is a state vector indicating the internal state. The state space model can be expressed by the following equations (1) and (2), where u is the input, y is the output, and x is the internal state.

[0033]

[0034] The likelihood ratio calculation unit 100 inputs multiple elements included in the sequence data to the state space model as input u. Then, a posterior probability indicating the probability that the input elements belong to a certain class is output as output y. The likelihood ratio calculation unit 100 calculates a likelihood ratio λ from the posterior probability that is the output of the state space model 100 using the following equation (3).

[0035]

[0036] Note that various extensions have been made to the state space model (see, for example, References 1 to 4 below). The first information processing device 10 may calculate the likelihood ratio using such an extended state space model.

[0037] [Reference 1] HiPPO: Recurrent Memory with Optimal Polynomial Projections [Reference 2] On the Parametrization and Initialization of Diagonal State Space Models [Reference 3] Efficiently Modeling Long Sequences with Structured State Spaces [Reference 4] Combining Recurrent, Convolutional, and Continuous-time Models with Linear State-Space Layers

[0038] (Technical Effects) Next, technical effects obtained by the first information processing device 10 will be described.

[0039] 1 to 4, the first information processing device 10 calculates likelihood ratios using a state space model 100 based on polynomial approximation of sequence data. This makes it possible to appropriately calculate likelihood ratios even when the sequence data contains a large number of elements. Specifically, by approximating a time series with a linear sum of orthogonal polynomials, accurate long-distance time series information can be incorporated.

[0040] Furthermore, with LSTM (Long Short Term Memory), a type of RNN (Recurrent Neural Network), gradient vanishing may occur when attempting to learn a long sequence, potentially preventing learning from progressing. However, gradient vanishing can be avoided by using the state space model 100. On the other hand, the state space model 100 allows for recursive specifications similar to RNN and LSTM, and therefore, compared to, for example, a Transformer, the amount of calculation is less likely to increase even if the number of data items increases. Furthermore, since the state space model 100 does not depend on the data sampling rate, for example, when acquiring time-series image data from a camera, it is possible to prevent performance from changing even if the camera's FPS is changed.

[0041] An example of a situation where the number of elements in the sequential data is large is a situation corresponding to a video with a long period of occlusion. For example, when re-matching a person who enters and exits a changing room, it is necessary to use images of the person before and after they enter the changing room, so the number of elements (i.e., images) included in the sequential data will be large. Furthermore, a situation where a judgment is made taking into account a long past history is also conceivable. For example, when trying to predict the current behavior of a criminal based on his or her experiences from childhood, it is necessary to handle data spanning several decades, so the number of elements included in the sequential data can be extremely large. The technical effects of the first information processing device 10 described above are particularly evident when the number of elements included in the sequential data is large.

[0042] Second Embodiment A second information processing apparatus 10 will be described with reference to Fig. 5. Note that the second embodiment differs from the first embodiment described above only in some configurations and operations, and other parts may be the same as those described in the first embodiment. Therefore, the following will describe in detail the parts that differ from the first embodiment already described, and will omit explanations of other overlapping parts as appropriate.

[0043] (Likelihood Ratio Calculation Using Activation Function) The second information processing device 10 is configured to be able to calculate likelihood ratios using an activation function. This activation function has at least the following three characteristics.

[0044] The first feature is that the value range is wider than [-1, 1]. For example, the tanh function and sigmoid function, which are sometimes used as general activation functions, have a value range of [-1, 1], but the activation function used by the second information processing device 10 has a wider value range than these functions. Therefore, the activation function used by the second information processing device 10 has higher expressive power compared to functions with a narrower value range. The value range of the activation function used by the second information processing device 10 may be, for example, [-∞, ∞].

[0045] The second feature is that the output value changes more slowly than the input value. For example, if the input value is X and the output value is Y, the slope of the activation function used by the second information processing device 10 is gentler than Y = X. Therefore, the activation function used by the second information processing device is less likely to diverge even when the input value becomes large.

[0046] The third feature is that the gradient at the origin does not diverge. That is, the activation function used by the second information processing device 10 is differentiated so that the value does not diverge when the input is set to 0.

[0047] (Adding a Linear Term) A linear term may be added to the activation function described above. Adding a linear term to the activation function can prevent the value of the differential coefficient from becoming zero (i.e., the gradient used during learning from disappearing).

[0048] For example, (1 + x) １／２ When differentiated, it becomes 1 / (1+x) １／２ x is included in the denominator, as in the above. When using such a function to calculate likelihood ratios over extremely long distances, x becomes extremely large, and there is a risk that the derivative value will approach zero. However, if a linear term such as ax (where a is a constant) is added, a value remains even after differentiation, so even if x becomes large, the derivative value will not become zero, preventing gradient vanishing. Therefore, learning can proceed appropriately even when the number of elements contained in the sequence data becomes large.

[0049] (Technical Effects) Next, technical effects obtained by the second information processing device 10 will be described with reference to Fig. 5. Fig. 5 is a graph showing an example of likelihood ratios calculated by the second information processing device together with a comparative example.

[0050] In Figure 5, it is preferable that the likelihood ratio used for class classification gradually changes in a predetermined direction as the number of samples (i.e., input elements) increases. However, in a comparative example using an activation function different from that of this embodiment, the likelihood ratio reaches a plateau (i.e., the likelihood ratio stops changing) even as the number of samples increases. One cause of the likelihood ratio reaching a plateau is, for example, a narrow threshold value for the activation function. While this type of plateau in the likelihood ratio does not occur in all cases, if it does occur, there is a risk that appropriate class classification will not be possible.

[0051] One possible measure to prevent the likelihood ratio from reaching a plateau is to not use an activation function. However, since the activation function contributes to stable learning, not using an activation function may result in unstable learning.

[0052] However, in the second information processing device 10, likelihood ratios are calculated using an activation function whose range is wider than [-1, 1], whose output values ​​change more gradually than input values, and whose gradient at the origin does not diverge. In this way, likelihood ratios can be calculated while suppressing the above-mentioned plateauing by taking advantage of the high expressive power provided by the wide range of the activation function. That is, likelihood ratios can be calculated with high accuracy for various sequence data. Furthermore, the use of an activation function also enables stable learning. That is, the second information processing device 10 can achieve both highly accurate likelihood ratio estimation and stable learning.

[0053] Third Embodiment A third information processing device 10 will be described with reference to Fig. 6. Note that the third embodiment is intended to describe an example of the activation function described above, and the device configuration and operational flow may be the same as those of the first and second embodiments already described. Therefore, the following will describe in detail the differences from the already described embodiments, and will omit a description of other overlapping parts as appropriate.

[0054] (Specific Example of Activation Function) First, a specific example of the activation function used in the third information processing device 10 will be described with reference to Fig. 6. Fig. 6 is a graph showing an example of the activation function used in the third information processing device.

[0055] In Fig. 6, the third information processing apparatus 10 uses an activation function that combines square root functions so as to be point-symmetric with respect to the origin. Hereinafter, this activation function will be referred to as a back-to-back-square-root function (B2Bsqrt function). The B2Bsqrt function satisfies the three characteristics already explained (i.e., the range is wider than [-1, 1], the change in output value is more gradual than the change in input value, and the gradient at the origin does not diverge). As an example of the B2Bsqrt function, for example, the function f shown in the following formula (4) is used. １ Examples include:

[0056]

[0057] Here, "x" is an input value, and "α" is a predetermined coefficient (hyperparameter). As shown in FIG. 6, the shape of the function changes by changing the value of α. A linear term (e.g., "ax") described in the second embodiment may be added to the above formula (4). Note that the B2Bsqrt function in the above formula (4) is merely an example, and can be modified as appropriate as long as it satisfies the three characteristics required for the activation function according to this embodiment.

[0058] (Technical Effects) Next, technical effects obtained by the third information processing apparatus 10 will be described.

[0059] The third information processing device 10 uses a B2Bsqrt function, which is a combination of a square root function, as an activation function. This B2Bsqrt function has the characteristics that its range is wider than [-1, 1], the change in output value is more gradual than the change in input value, and the gradient at the origin does not diverge. Therefore, as described in the second embodiment, it is possible to achieve both highly accurate likelihood ratio estimation and stable learning.

[0060] Fourth Embodiment A fourth information processing device 10 will be described with reference to Fig. 7. The fourth embodiment, like the third embodiment, describes an example of an activation function, and the device configuration and operational flow may be similar to those of the first and second embodiments. Therefore, the following will describe in detail only the differences from the previously described embodiments, and will omit a description of other overlapping parts as appropriate.

[0061] (Specific Example of Activation Function) First, a specific example of the activation function used in the fourth information processing device 10 will be described with reference to Fig. 7. Fig. 7 is a graph showing an example of the activation function used in the third information processing device.

[0062] As shown in Fig. 7, the fourth information processing apparatus 10 uses an activation function that combines log functions so as to be point-symmetric with respect to the origin. Hereinafter, this activation function will be referred to as a logistic-activation function. The logistic-activation function satisfies the three characteristics already explained (i.e., the range is wider than [-1, 1], the change in output value is more gradual than the change in input value, and the gradient at the origin does not diverge). As an example of the logistic-activation function, for example, the function f shown in the following formula (5) is used. １ Examples include:

[0063]

[0064] Here, "x" is an input value, and "α" is a predetermined coefficient (hyperparameter). A linear term (e.g., "ax") described in the second embodiment may be added to the above formula (5). Note that the logistic-activation function in the above formula (5) is merely an example, and can be modified as appropriate as long as it satisfies the three characteristics required for the activation function according to this embodiment.

[0065] (Technical Effects) Next, technical effects obtained by the fourth information processing apparatus 10 will be described.

[0066] In the fourth information processing device 10, a logistic-activation function combining a log function is used as the activation function. This logistic-activation function has the characteristics that its range is wider than [-1, 1], the output value changes more slowly than the input value, and the gradient at the origin does not diverge. Therefore, as explained in the second embodiment, it is possible to achieve both highly accurate likelihood ratio estimation and stable learning.

[0067] The scope of each embodiment also includes a processing method in which a program that operates the configuration of each embodiment to realize the functions of the above-described embodiments is recorded on a recording medium, the program recorded on the recording medium is read as code, and the program is executed on a computer. In other words, a computer-readable recording medium is also included in the scope of each embodiment. Furthermore, each embodiment includes not only a recording medium on which the above-described program is recorded, but also the program itself.

[0068] Examples of recording media that can be used include floppy disks, hard disks, optical disks, magneto-optical disks, CD-ROMs, magnetic tapes, non-volatile memory cards, and ROMs. Furthermore, the scope of each embodiment is not limited to programs that execute processes by themselves, but also includes programs that execute processes by operating on an OS in conjunction with other software or expansion board functions. Furthermore, the program itself may be stored on a server, and part or all of the program may be downloadable from the server to a user terminal. The program may be provided to the user in, for example, a SaaS (Software as a Service) format.

[0069] <Supplementary Notes> The above-described embodiment may be further described as in the following supplementary notes, but is not limited to the following.

[0070] (Supplementary Note 1) The information processing device described in Supplementary Note 1 is an information processing device that includes: an acquisition means that sequentially acquires multiple elements included in sequence data; and, each time an element is acquired, a likelihood ratio that indicates the likelihood of a class to which the sequence data belongs based on two or more adjacent elements, wherein, when calculating the likelihood ratio, a likelihood ratio calculation means uses a state space model based on a polynomial approximation of the sequence data; and a classification means that classifies the sequence data into one of multiple classes based on the likelihood ratio.

[0071] (Supplementary Note 2) The information processing device described in Supplementary Note 2 is the information processing device described in Supplementary Note 1, wherein the likelihood ratio calculation means calculates the likelihood ratio by further using an activation function whose range is wider than [-1, 1], whose output value changes more slowly than whose input value changes, and whose gradient at the origin does not diverge.

[0072] (Supplementary Note 3) The information processing device according to Supplementary Note 3 is the information processing device according to Supplementary Note 2, wherein the likelihood ratio calculation means calculates the likelihood ratio using the activation function to which a linear term has been added.

[0073] (Supplementary Note 4) The information processing device according to Supplementary Note 4 is the information processing device according to Supplementary Note 2 or 3, wherein the activation function is a function obtained by combining square root functions so as to be point-symmetric with respect to the origin.

[0074] (Supplementary Note 5) In the information processing device according to Supplementary Note 5, the activation function is f=sign(x){(α+|x|) １／２ -α １／２} (x: input value, α: predetermined coefficient).

[0075] (Supplementary Note 6) The information processing device according to Supplementary Note 6 is the information processing device according to Supplementary Note 2 or 3, wherein the activation function is a function obtained by combining log functions so as to be point-symmetric with respect to the origin.

[0076] (Supplementary Note 7) The information processing device according to Supplementary Note 7 is the information processing device according to Supplementary Note 6, in which the activation function includes f = sign(x) {log(α + |x|)} (x: input value, α: predetermined coefficient).

[0077] (Supplementary Note 8) The information processing method described in Supplementary Note 8 is an information processing method that sequentially acquires multiple elements included in sequence data, and for each acquired element, calculates a likelihood ratio indicating the likelihood of a class to which the sequence data belongs based on two or more adjacent elements, and when calculating the likelihood ratio, uses a state space model based on polynomial approximation of the sequence data, and classifies the sequence data into one of multiple classes based on the likelihood ratio.

[0078] (Supplementary Note 9) The recording medium described in Supplementary Note 9 is a recording medium having recorded thereon a computer program that causes a computer to execute an information processing method, which includes sequentially acquiring multiple elements included in sequence data, calculating, for each acquired element, a likelihood ratio indicating the likelihood of a class to which the sequence data belongs based on two or more adjacent elements, and classifying the sequence data into one of multiple classes based on the likelihood ratio using a state space model based on a polynomial approximation of the sequence data when calculating the likelihood ratio.

[0079] (Supplementary Note 10) The computer program described in Supplementary Note 10 causes a computer to execute an information processing method that sequentially acquires multiple elements included in sequence data, and for each acquired element, calculates a likelihood ratio indicating the likelihood of a class to which the sequence data belongs based on two or more adjacent elements, and when calculating the likelihood ratio, uses a state space model based on polynomial approximation of the sequence data, and classifies the sequence data into one of multiple classes based on the likelihood ratio.

[0080] This disclosure may be modified as appropriate within the scope that does not contradict the gist or idea of ​​the invention that can be read from the claims and the entire specification, and information processing devices, information processing methods, and recording media that involve such modifications are also included in the technical idea of ​​this disclosure.

[0081] REFERENCE SIGNS LIST 10 Information processing device 11 Processor 50 Acquisition unit 100 Likelihood ratio calculation unit 110 State space model 150 Classification unit

Claims

1. A means for sequentially acquiring multiple elements contained in sequential data, Each time an element is acquired, a likelihood ratio is calculated based on two or more adjacent elements to indicate the likelihood of the class to which the sequence data belongs, and the likelihood ratio calculation means uses a state-space model based on a polynomial approximation of the sequence data when calculating the likelihood ratio. A classification means for classifying the series data into one of several classes based on the likelihood ratio, An information processing device equipped with the following features.

2. The likelihood ratio calculation means further calculates the likelihood ratio using an activation function whose range is wider than [-1, 1], whose change in output value is slower than the change in input value, and whose gradient at the origin does not diverge. The information processing apparatus according to claim 1.

3. The likelihood ratio calculation means calculates the likelihood ratio using the activation function with a linear term added to it. The information processing apparatus according to claim 2.

4. The aforementioned activation function is a function obtained by combining the square root function in a way that is point-symmetric about the origin. The information processing apparatus according to claim 2 or 3.

5. The aforementioned activation function is, f=sign(x){(α+|x|) １／２ -α １／２ } (x: input value, α: predetermined coefficient) The information processing apparatus according to claim 4.

6. The aforementioned activation function is a function obtained by combining log functions in a way that is point-symmetric about the origin. The information processing apparatus according to claim 2 or 3.

7. The aforementioned activation function is, f=sign(x) {log(α+|x|)} (x: input value, α: predetermined coefficient) The information processing apparatus according to claim 6.

8. Multiple elements included in the series data are acquired sequentially, Each time an element is obtained, a likelihood ratio indicating the likelihood of the class to which the sequence data belongs is calculated based on two or more adjacent elements. When calculating the likelihood ratio, a state-space model based on a polynomial approximation of the sequence data is used. Based on the likelihood ratio, the series data is classified into one of several classes. Information processing methods.

9. Multiple elements included in the series data are acquired sequentially, Each time an element is obtained, a likelihood ratio indicating the likelihood of the class to which the sequence data belongs is calculated based on two or more adjacent elements. When calculating the likelihood ratio, a state-space model based on a polynomial approximation of the sequence data is used. Based on the likelihood ratio, the series data is classified into one of several classes. A computer program that instructs a computer to execute information processing methods.