Information processing device, method for information processing, and program

Abstract

To appropriately present, to a user, contribution levels of respective prediction models with respect to target data.SOLUTION: The information processing device includes: a first acquisition unit for sequentially acquiring target data; a second acquisition unit for acquiring trained parameters for each of a plurality of target models; a calculation unit for calculating the weight for each of the plurality of target models when an objective variable contained in the target data is expressed as the weighted sum of the prediction values obtained by applying the trained parameters of the plurality of target models to explanatory variables contained in the target data; and a display unit for displaying transitions of the weight calculated by the calculation unit for each piece of target data acquired sequentially.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device, an information processing method, and a program. [Background technology]

[0002] There is known a technique for deriving a prediction result by applying a prediction model to target data. For example, Patent Literature 1 discloses a technique for predicting the next state of a target based on the state of the target using a prediction model. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-078529 Summary of the Invention [Problem to be solved by the invention]

[0004] On the other hand, a configuration that uses multiple prediction models to perform more flexible predictions is also possible. However, such a configuration has the problem that it is difficult for users to understand the contribution (weight) of each prediction model to the target data. Furthermore, in actual situations, the contribution (weight) of each prediction model to the target data that is sequentially acquired may also change.

[0005] The present disclosure has been made in consideration of the above-mentioned problems, and one exemplary purpose thereof is to provide a technology that can suitably present to a user the contribution (weight) of each prediction model to target data. [Means for solving the problem]

[0006] An information processing device according to an exemplary aspect of the present disclosure includes a first acquisition means for sequentially acquiring target data, a second acquisition means for acquiring learned parameters for each of a plurality of target models, a calculation means for calculating a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying the learned parameters for each of the plurality of target models to explanatory variables included in the target data, and a display means for displaying the progress of the weight calculated by the calculation means for each of the sequentially acquired target data.

[0007] An information processing method according to an exemplary aspect of the present disclosure includes sequentially acquiring target data, acquiring learned parameters for each of a plurality of target models, calculating a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying the learned parameters for each of the plurality of target models to explanatory variables included in the target data, and displaying a progress of the calculated weight for each of the sequentially acquired target data.

[0008] The information processing device according to each aspect of the present invention may be realized by a computer, in which case a program that causes the computer to operate as each part (software element) of the information processing device to realize the information processing device also falls within the scope of the present invention. [Effects of the Invention]

[0009] According to an exemplary aspect of the present disclosure, an exemplary effect is achieved in that the contribution (weight) of each prediction model to target data can be suitably presented to a user. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a block diagram illustrating a configuration of an information processing device according to the present disclosure. [Figure 2]FIG. 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Figure 3] 1 is a block diagram illustrating a configuration of an information processing device according to the present disclosure. [Figure 4] 10 shows an example of a graph displayed by an output unit via an input / output unit according to the present disclosure. [Figure 5] 10 shows an example of a graph displayed by an output unit via an input / output unit according to the present disclosure. [Figure 6] 10 shows an example of a graph displayed by an output unit via an input / output unit according to the present disclosure. [Figure 7] 10 shows an example of a graph displayed by an output unit via an input / output unit according to the present disclosure. [Figure 8] 1 is a block diagram illustrating a configuration of an information processing device according to the present disclosure. [Figure 9] FIG. 10 is a diagram schematically illustrating the outputs of each endpoint model in the LPV model according to the present disclosure and the interior division ratio parameters by which each output is multiplied. [Figure 10] FIG. 10 is a diagram illustrating an example of a processing flow in an information processing device according to the present disclosure. [Figure 11] 10 shows an example of a graph displayed by an output unit via an input / output unit according to the present disclosure. [Figure 12] 1 is a block diagram illustrating a configuration of an information processing device according to the present disclosure. [Figure 13] FIG. 10 is a diagram illustrating an example of a processing flow in an information processing device according to the present disclosure. [Figure 14] FIG. 1 is a block diagram illustrating a configuration of a computer that functions as an information processing device according to the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0011] The following are examples of embodiments of the present invention. However, the present invention is not limited to the exemplary embodiments shown below, and various modifications are possible within the scope of the claims. For example, embodiments obtained by appropriately combining the technical means employed in the exemplary embodiments shown below may also be included in the scope of the present invention. Furthermore, embodiments obtained by appropriately omitting some of the technical means employed in the exemplary embodiments shown below may also be included in the scope of the present invention. Furthermore, the effects mentioned in the exemplary embodiments shown below are examples of effects expected in the exemplary embodiments, and do not define the scope of the present invention. In other words, embodiments that do not exhibit the effects mentioned in the exemplary embodiments shown below may also be included in the scope of the present invention.

[0012] First Exemplary Embodiment A first exemplary embodiment, which is one example of an embodiment of the present invention, will be described in detail with reference to the drawings. This exemplary embodiment is the basic form of each exemplary embodiment described later. Note that the scope of application of each technical means employed in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technical means employed in this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure to the extent that no particular technical obstacles arise. Furthermore, each technical means shown in the drawings referred to in describing this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure to the extent that no particular technical obstacles arise.

[0013] (Configuration of information processing device 1) The configuration of the information processing device 1 will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the information processing device 1. As shown in Fig. 1, the information processing device 1 includes a first acquisition unit 21, a second acquisition unit 22, a calculation unit 23, and a display unit 24. In this exemplary embodiment, the first acquisition unit 21, the second acquisition unit 22, the calculation unit 23, and the display unit 24 respectively realize a first acquisition means, a second acquisition means, a calculation means, and a display means.

[0014] (First acquisition unit 21) The first acquisition unit 21 sequentially acquires target data. Here, the process of "sequentially acquiring target data" includes, as an example, a process of acquiring target data at multiple different timings. For example, the process includes a process of acquiring sensing data from a certain sensor provided in a production line of a target product, the sensing data at a certain time (timing) t1 as target data TD(t1), and a process of acquiring sensing data at another time (timing) t2 as target data TD(t2). Alternatively, the process includes a process of acquiring sales data from a certain store on a certain day d1 as target data TD(d1), and a process of acquiring sales data on another day d2 as target data TD(d2).

[0015] Furthermore, the target data may include, for example, an explanatory variable ~x (x with a tilde) and a target variable ~y (y with a tilde). If the timing is explicitly stated, the target data TD(t1) may be expressed, for example, as follows: {~x(t1),~y(t1)} The target data TD(t2) is expressed as follows, for example: {~x(t2),~y(t2)} Alternatively, a more general index k, not limited to time or date, can be used to express the target data (set) as follows: {~x k ,~y k} k=1 N It is sometimes written as follows. Note that explanatory variables are sometimes called features. Furthermore, when a response variable is the target of derivation, the response variable is sometimes called a predicted value. Furthermore, explanatory variables and response variables are sometimes called state variables. In the above formula, k is an index assigned to each state variable, and N represents the total number of state variables (same below).

[0016] Note that the above specific notation does not limit this exemplary embodiment. Also, in the above example, each of ~x and ~y for each time t or k is not limited to one-dimensional data, but may be multi-dimensional data.

[0017] As an example, the first acquisition unit 21 acquires the above-mentioned target data for each predetermined period. Here, the specific example of the predetermined period does not limit the present exemplary embodiment, and may be, for example, one day to several days, one hour to several hours, one minute to several minutes, or one second to several seconds.

[0018] (Second acquisition unit 22) The second acquisition unit 22 acquires trained parameters for each of the plurality of target models. As an example, the second acquisition unit 22 acquires trained parameters Pi of a model i designated by an index i (1≦i≦m: m is the total number of target models) among the plurality of target models. For example, if the target model i is a regression model, the trained parameters Pi include regression coefficients (regression coefficient vector, regression coefficient matrix) W (i) The acquisition unit 11 supplies the acquired learned parameters to the calculation unit 23.

[0019] (Calculation unit 23) The calculation unit 23 calculates the weight of each of the plurality of target models when the objective variable included in the target data is expressed as a weighted sum of each predicted value obtained by applying the learned parameters for each of the plurality of target models to the explanatory variables included in the target data.

[0020] As an example, the calculation unit 23 calculates, for each model i (i=1 to m) included in the plurality of target models (model 1, model 2, . . . model m), Explanatory variables included in the target data ~x k , the regression coefficient W (i) By applying k (i) (y without tildek )of y k (i) = W (i) ~x k Calculated by Each predicted value (y k (1) ,y k (1) ,···,y k (m) ) as the weighted sum of the objective variable y k (y with tilde k ), that is, ~y k = Σr (i) y k (i) (where Σ is the sum over index i, from i=1 to i=m) k (y with tilde k ) when expressing each weight r of the plurality of target models (i) (i=1~m) is calculated.

[0021] The calculation of the weight by the calculation unit 23 is performed for each of the target data sequentially acquired by the first acquisition unit 21. In other words, the calculation of the weight by the calculation unit 23 is performed for each index k. Then, each weight r (i) may depend on the index k. For example, when the index k is expressed as timing t, the calculation unit 23 calculates The weight r is calculated by the above calculation process using the target data TD(t1) acquired at the timing t1. (i) Calculate (t1), The weight r is calculated by the above calculation process using the target data TD(t2) acquired at the timing t2. (i) Calculate (t2) and... The following processing will be performed.

[0022] (Display section 24) The display unit 24 displays the transition of the weight calculated by the calculation unit 23 for each of the target data sequentially acquired by the first acquisition unit 21. As an example, the display unit 24 displays the transition of the weight as follows: The weight r calculated by the calculation unit 23 with reference to the target data TD(t1) acquired at the timing t1 (i) (t1), and The weight r calculated by the calculation unit 23 with reference to the target data TD(t2) acquired at the timing t2 (i) (t2), Display.

[0023] (Effects of information processing device 1) As described above, in the information processing device 1, -Sequentially acquire target data, Obtaining trained parameters for each of multiple target models, calculating a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying trained parameters for each of the plurality of target models to explanatory variables included in the target data; Displaying the transition of the weight calculated by the calculation means for each of the sequentially acquired target data The above configuration provides an effect that the user can easily recognize the transition of the weight of each of the plurality of target models. In other words, the above configuration provides an exemplary effect that the contribution (weight) of each prediction model (each target model) to the target data can be suitably presented to the user.

[0024] (Flow of information processing method S1) The flow of the information processing method S1 will be described with reference to Fig. 2. Fig. 2 is a flow diagram showing the flow of the information processing method S1. As shown in Fig. 2, the information processing method S1 includes a first acquisition process S21, a second acquisition process S22, a weight calculation process S23, and a display process S24.

[0025] (First acquisition process S21) In the first acquisition process S21, the first acquisition unit 21 sequentially acquires target data. The specific process performed by the first acquisition unit 21 has been described above, and therefore will not be described here.

[0026] (Second acquisition process S22) In the second acquisition process S22, the second acquisition unit 22 acquires trained parameters for each of the plurality of target models. The specific process by the second acquisition unit 22 has been described above, and therefore will not be described here.

[0027] (Weight calculation process S23) In the weight calculation process S23, the calculation unit 23 calculates the weight of each of the plurality of target models when the objective variable included in the target data is expressed as a weighted sum of each predicted value obtained by applying the learned parameters for each of the plurality of target models to the explanatory variables included in the target data acquired in the first acquisition process S21. The specific processing by the calculation unit 23 has been described above, and therefore will not be described here.

[0028] (Display process (step) S24) In the display process S24, the display unit 24 displays the transition of the weight calculated by the calculation unit 23 in the weight calculation process S23 for each of the target data sequentially acquired in the first acquisition process S21. (Effect of information processing method S1) As described above, in the information processing device 1, -Sequentially acquire target data, Obtaining trained parameters for each of multiple target models, calculating a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying trained parameters for each of the plurality of target models to explanatory variables included in the target data; Displaying the transition of the weight calculated by the calculation means for each of the sequentially acquired target data According to the above configuration, the same effects as those of the information processing device 1 are achieved.

[0029] Second Exemplary Embodiment A second exemplary embodiment, which is one example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same functions as those described in the above exemplary embodiment will be assigned the same reference numerals, and their description will be omitted as appropriate. The scope of application of each technical means employed in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technical means employed in this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise. Furthermore, each technical means shown in each drawing referenced to explain this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise.

[0030] (Configuration of information processing device 1A) The configuration of the information processing device 1A will be described with reference to Fig. 3. Fig. 3 is a block diagram showing the configuration of the information processing device 1A. As shown in Fig. 3, the information processing device 1A includes a control unit 10A, a storage unit 15A, a communication unit 16A, and an input / output unit 17A.

[0031] (Storage section 15A) First, various data (information) stored in the storage unit 15A will be described. Data referenced by the control unit 10A is stored in the storage unit 15A. Examples of the storage unit 15A include, but are not limited to, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination thereof.

[0032] Examples of data stored in the storage unit 15A include, but are not limited to, target data TD, correspondence relationship information CI, and learning results LR, as shown in FIG.

[0033] The target data TD may include an explanatory variable ~x (x with a tilde) and a target variable ~y (y with a tilde), as in the first exemplary embodiment. Explicitly expressing the timing, the target data TD(t1) may be expressed as follows, for example: {~x(t1),~y(t1)} The target data TD(t2) is expressed as follows, for example: {~x(t2),~y(t2)} Alternatively, a more general index k, not limited to time or date, can be used to express the target data (set) as follows: {~x k ,~y k} k=1 N As in the first exemplary embodiment, the explanatory variables may also be referred to as feature quantities. Furthermore, when a dependent variable is the target of derivation, the dependent variable may also be referred to as a predicted value. Furthermore, the explanatory variables and dependent variable may also be referred to as state variables. In the above formula, k is an index assigned to each state variable, and N represents the total number of state variables (same applies hereinafter).

[0034] Note that the above specific notation does not limit this exemplary embodiment. Also, in the above example, each of ~x and ~y for each time t or k is not limited to one-dimensional data, but may be multi-dimensional data.

[0035] The correspondence information CI is information about the correspondence between a plurality of target models and the learning data LD. An example of the information about the correspondence is as follows: Information indicating what kind of training data LD each of the multiple target models has been trained with (will be trained with) Here, the learning data LD may or may not be included in the above-mentioned target data TD. In other words, the learning data LD may be different from the above-mentioned target data TD. As an example, the information processing device 1A according to this exemplary embodiment: The plurality of target models are trained by a training process that refers to the training data LD (in other words, trained parameters for each of the plurality of target models are generated), Calculating the weight of each of the plurality of target models when the objective variable included in the target data TD is expressed as a weighted sum of each predicted value obtained by applying the trained parameters for each of the plurality of target models to the target data TD that is acquired sequentially. The processes are executed in the following order, but this is not a limitation of this exemplary embodiment.

[0036] An example of the correspondence information CI is, for example, the correspondence information CI relating to the target models a, b, and c and the learning data LD1, LD2, and LD3, as follows: Target model a: Training data LD1, LD2 Target model b: Training data LD3 Target model c: Training data LD1, LD2, LD3 Here, the first line in the above example indicates that the target model a has been trained (will be trained) using the training data LD1 and LD2. The same applies to the other lines.

[0037] In other words, the correspondence information CI is Information indicating whether each of the multiple target models used (or will use) each of the multiple learning data LD for learning. As an example, the correspondence relationship information CI regarding the target models a, b, and c and the learning data LD1, LD2, and LD3 may be expressed as follows: Target model a: Training data LD1 (○), LD2 (○), LD3 (×) Target model b: Training data TD1(×), TD2(×), TD3(○) Target model c: Training data TD1(○), TD2(○), TD3(○) Here, ○ indicates that it was used (will be used) in learning, and × indicates that it was not used (will not be used) in learning.

[0038] In the above example, the correspondence information CI may include information about the circumstances under which each piece of learning data was acquired. Learning data LD1 (e.g., sales volume): Situation A (e.g., sunny weather) Learning data LD2 (e.g., sales volume): Situation B (e.g., cloudy weather) Learning data LD3 (e.g., sales volume): Situation C (e.g., rainy weather) In addition, when the correspondence information CI includes the correspondence between each learning data and the situation in which the learning data was acquired in this way, the correspondence information CI may be expressed as including the relationship between each target model and information related to the target model. For example, in the above example, The correspondence information CI is Target model a: Situation A (e.g., sunny weather), Situation B (e.g., cloudy weather) Target model b: Situation C (e.g., rainy weather) Target model c: Situations A, B, and C The correspondence relationship information CI may include information indicating a correspondence such as the following. By including the above information in the correspondence relationship information CI, the information processing device 1A can identify what situation each target model is associated with, in other words, what situation each target model is preferably used in. Furthermore, the information processing device 1A can also present to the user what situation each target model is preferably used in.

[0039] Information about situations such as situations A, B, and C may be expressed as information about types. In other words, the correspondence information CI may include information indicating which types at least one of the target model and the training data is associated with.

[0040] In addition, the correspondence information CI is - Information indicating whether a certain learning data LD is similar to other learning data LD Here, the fact that the learning data LD and other learning data LD are similar refers to the case where the types of the learning data LD and other learning data LD are similar. For example, if the learning data LD is data on the number of sales items acquired in a situation A (e.g., sunny weather), the learning data LD', which is data on the number of sales items in a situation A' (e.g., sunny weather) similar to the situation A, is data similar to the learning data LD.

[0041] The learning result LR includes the learned parameters for each of the multiple target models. As an example, the learning result LR includes the learned parameters Pi of the target model i specified by the index i (1≦i≦m: m is the total number of target models). For example, if the target model i is a regression model, the learned parameters Pi include the regression coefficients (regression coefficient vector, regression coefficient matrix) W (i) may be included.

[0042] (Communication unit 16A) The communication unit 16A is an interface for transmitting and receiving data via a network. Examples of the communication unit 16A include, but are not limited to, communication chips for various communication standards such as Ethernet (registered trademark), Wi-Fi (Wireless Fidelity (registered trademark), and wireless communication standards for mobile data communication networks, and USB-compliant connectors.

[0043] (Input / output section 17A) The input / output unit 17A is an interface that receives input of data and outputs data. Examples of the input / output unit 17A include, but are not limited to, a microphone, a camera, a gaze input device, a keyboard, a touchpad, a speaker, and a liquid crystal display.

[0044] (Control unit 10A) The control unit 10A controls each component included in the information processing device 1A. As shown in Fig. 3, the control unit 10A also includes a first acquisition unit 21, a second acquisition unit 22, a calculation unit 23, a display unit 24, and an abnormality detection unit 19. In this exemplary embodiment, the first acquisition unit 21, the second acquisition unit 22, the calculation unit 23, the display unit 24, and the abnormality detection unit 25 function as a first acquisition unit, a second acquisition unit, a calculation unit, a display unit, and an abnormality detection unit, respectively.

[0045] (First acquisition unit 21) The first acquisition unit 21 sequentially acquires target data. Here, the process of "sequentially acquiring target data" includes, as in the first exemplary embodiment, a process of acquiring target data at a plurality of different timings. For example, it includes a process of acquiring sensing data from a certain sensor at a certain time (timing) t1 as target data TD(t1) and a process of acquiring sensing data at another time (timing) t2 as target data TD(t2). Or, it includes a process of acquiring sales data from a certain store on a certain day d1 as target data TD(d1) and a process of acquiring sales data on another day d2 as target data TD(d2).

[0046] As an example, the first acquisition unit 21 acquires the above-mentioned target data every predetermined period. Here, the specific example of the predetermined period does not limit this exemplary embodiment, but may be, for example, one day to several days, one hour to several hours, one minute to several minutes, or one second to several seconds. A more specific explanation of the target data has been given above, so a detailed explanation will be omitted here.

[0047] (Second acquisition unit 22) The second acquisition unit 22 acquires trained parameters for each of a plurality of target models. As an example, the second acquisition unit 22 acquires trained parameters Pi of a target model i, which is specified by an index i (1≦i≦m: m is the total number of target models) among the plurality of target models. Here, the trained parameters Pi are parameters trained by the target model i with reference to the above-mentioned training data LD, and the relationship between the model i and the training data LD is defined by the above-mentioned correspondence information CI (feature information), for example. For example, if the target model i is a regression model, the trained parameters Pi include regression coefficients (regression coefficient vector, regression coefficient matrix) W (i) Includes:

[0048] In this exemplary embodiment, the second acquisition unit 22 may be configured to further acquire feature information associated with the target model or the learned parameters. Here, the feature information refers to information that characterizes each of the multiple target models, and may be the above-mentioned correspondence relationship information CI, for example. More specifically, the feature information is Information indicating what kind of training data each of the multiple target models has been trained with (will be trained with), Information indicating the circumstances under which each piece of training data was acquired, and -What situations are associated with each of the multiple target models? The configuration may include at least one of the above.

[0049] (Calculation unit 23) As in exemplary embodiment 1, the calculation unit 23 calculates the weight of each of the multiple target models when the objective variable included in the target data is expressed as a weighted sum of each predicted value obtained by applying the learned parameters for each of the multiple target models to the explanatory variables included in the target data.

[0050] As an example, similar to the first exemplary embodiment, the calculation unit 23 calculates, for each model i (i=1 to m) included in the plurality of target models (model 1, model 2, . . . model m), Explanatory variables included in the target data ~x k , the regression coefficient W (i) By applying k (i) (y without tilde k )of y k (i) = W (i) ~x k Calculated by Each predicted value (y k (1) ,y k (1) ,···,y k (m) ) as the weighted sum of the objective variable y k (y with tilde k ), that is, ~y k = Σr (i) y k (i) (where Σ is the sum over index i, from i=1 to i=m) k (y with tilde k ) when expressing each weight r of the plurality of target models (i) (i=1~m) is calculated.

[0051] The calculation of the weight by the calculation unit 23 is performed for each of the target data sequentially acquired by the first acquisition unit 21. In other words, the calculation of the weight by the calculation unit 23 is performed for each index k. Then, each weight r (i) may depend on the index k. For example, when the index k is expressed as timing t, the calculation unit 23 calculates The weight r is calculated by the above calculation process using the target data TD(t1) acquired at the timing t1. (i) Calculate (t1), The weight r is calculated by the above calculation process using the target data TD(t2) acquired at the timing t2. (i) Calculate (t2) and... The following processing will be performed.

[0052] (Display section 24) The display unit 24 displays, via the input / output unit 17A, the transition of the weight calculated by the calculation unit 23 for each piece of target data sequentially acquired by the first acquisition unit 21. As an example, the display unit 24 displays the transition of the weight as follows: The weight r calculated by the calculation unit 23 with reference to the target data TD(t1) acquired at the timing t1 (i) (t1), and The weight r calculated by the calculation unit 23 with reference to the target data TD(t2) acquired at the timing t2 (i) (t2), Display.

[0053] The display unit 24 may also be configured to further display feature information for each of the plurality of models. As an example, the display unit 24 may display, as the feature information, information indicating the circumstances under which each piece of target data was acquired. The display unit 24 may also be configured to further display the results of an anomaly detection process performed by the anomaly detection unit 25, which will be described later. Specific examples of display by the display unit 24 will be described later.

[0054] (Abnormality detection unit 25) The anomaly detection unit 25 performs an anomaly detection process by referring to the transition of the weights. As an example, the anomaly detection unit 25 may perform the anomaly detection process by identifying one or more change amounts of the weights included in the transition of the weights and comparing the change amounts with a predetermined threshold. For example, the anomaly detection unit 25 may determine that there is an "anomaly" in the transition of the weights when any one or more change amounts of the weights included in the transition of the weights is equal to or greater than the threshold, and may determine that there is no anomaly when this is not the case.

[0055] (Display example 1) Below, examples of display by the information processing device 1A will be described with reference to different drawings. FIG. 4 shows Example 1 of display information displayed by the display unit 24. In FIG. 4, as an example, an example is given in which the horizontal axis represents the number of days elapsed and the vertical axis represents the demand y, but this does not limit the present exemplary embodiment. In addition, the example shown in FIG. 4 takes as an example a case in which the target data TD includes an explanatory variable x0 and a target variable y0.

[0056] In the example shown in FIG. 4, the display information displayed by the display unit 24 includes the following: For the target data TD, the trained parameters W for the target model 1 (1) The regression line L1 obtained by applying For the target data TD, the trained parameters W for the target model 2 (2) The regression line L2) obtained by applying The weight calculated by the calculation unit 23 for the target data TD (explanatory variable x0, objective variable y0) is the weight of model 1 (ratio r in FIG. 4). (1) ), The weight calculated by the calculation unit 23 for the target data TD (explanatory variable x0, objective variable y0) is the weight of model 2 (ratio r in FIG. 4). (2) ), - The regression line L1 or the trained parameters W associated with the target model 1 (1) Feature information associated with the situation ("L1::Model corresponding to situation A" in Figure 4) - Associated with the target model 2 (in other words, the regression line L2 or the trained parameters W (2) Feature information associated with situation B ("L2::Model corresponding to situation B" in Figure 4) Here, the display information includes, as shown in FIG. The explanatory variable x0 included in the target data TD is used to calculate the trained parameter W (1) The predicted value P1 obtained by applying The explanatory variable x0 included in the target data TD is used to calculate the trained parameter W (2) The predicted value P2 obtained by applying It may be configured to further include:

[0057] The y coordinate of PP0 shown in Fig. 4 is the objective variable y0 included in the target data TD. For example, PP0 is The weighted average of predicted values ​​P1 and P2, with weight r of Model 1 (1) and the weight r of Model 2 (2) The predicted value is the result of the weighted average with the weights of It may be expressed as, The explanatory variable x0 contained in the target data TD is used to calculate the trained parameters (W (1) , W (2) ) is a response variable expressed as a weighted sum of each predicted value (P1, P2) obtained by applying It can also be expressed as:

[0058] Note that the specific examples of situations A, B, etc. described above are not intended to limit the present example, but one example may be sales of drinking water at a store where the target data TD is located. Furthermore, a situation in which a campaign for the drinking water is being conducted may be situation A, and a situation in which the volume of the drinking water has been increased may be situation B. Note that examples of situations include, but are not limited to, a situation in which the appearance of the drinking water container has been changed, a situation in which a new store has opened near the store, a situation in which it is raining, and a situation that includes all of these situations.

[0059] (Display example 2) FIG. 5 shows Example 2 of the display information displayed by the display unit 24. The contents described with reference to FIG. 4 will not be described again. In the example shown in FIG. 5, the display information displayed by the display unit 24 is The weight calculated by the calculation unit 23 for the target data TD0 (explanatory variable x0, objective variable y0) at timing 0 is the weight of model 1 (ratio r0 in FIG. 4). (1) ), The weight calculated by the calculation unit 23 for the target data TD0 (explanatory variable x0, objective variable y0) at timing 0 is the weight of model 2 (ratio r0 in FIG. 4). (2) ), The weight calculated by the calculation unit 23 for the target data TD1 (explanatory variable x1, objective variable y1) at timing 1 is the weight of model 1 (ratio r1 in FIG. 4). (1) ), The weight calculated by the calculation unit 23 for the target data TD1 (explanatory variable x1, objective variable y1) at timing 1 is the weight of model 2 (ratio r1 in FIG. 4). (2) ) In other words, the display information displayed by the display unit 24 includes: The transition of the weight (r0 (1) From r1 (1) Transition to r0 (2) From r1 (2) This includes the transition to

[0060] As shown in this example, according to the information processing device 1A, the display information displayed by the display unit 24 includes the changes in the weights calculated by the calculation unit 23 for each of the target data acquired sequentially, thereby achieving an exemplary effect of suitably presenting to the user the contribution (weight) and changes in the contribution (weight) of each prediction model (each target model) to the target data.

[0061] (Display example 3) The display examples of the transition of weights by the display unit 24 are not limited to the above examples. Fig. 6 shows Example 3 of display information displayed by the display unit 24. In Fig. 6, as an example, the horizontal axis represents the number of days elapsed, and the vertical axis represents the ratio r indicating the weight of each target model.

[0062] As shown in FIG. 6, the display information displayed by the display unit 24 is The weight of the model 1 calculated by the calculation unit 23 for the target data TD (the ratio r (1) ) over time (transition) (1) The weight of the model 2 calculated by the calculation unit 23 for the target data TD (the ratio r (2) ) over time (transition) (2) More specifically, the display unit 24 may include a line graph G (1) As, The weight calculated by the calculation unit 23 for the target data TD0 (explanatory variable x0, objective variable y0) at timing 0 is the weight of model 1 (ratio r0 in FIG. 6). (1) ), The weight calculated by the calculation unit 23 for the target data TD1 (explanatory variable x1, objective variable y1) at timing 1 is the weight of model 1 (ratio r1 in FIG. 6). (1) ), The weight calculated by the calculation unit 23 for the target data TD2 (explanatory variable x2, objective variable y2) at timing 2 is the weight of model 1 (ratio r2 in FIG. 6). (1) ) The weights of Model 1 are displayed over time in a line graph G. (2) As, The weight calculated by the calculation unit 23 for the target data TD0 (explanatory variable x0, objective variable y0) at timing 0 is the weight of model 2 (ratio r0 in FIG. 6). (2) ), The weight calculated by the calculation unit 23 for the target data TD1 (explanatory variable x1, objective variable y1) at timing 1 is the weight of model 2 (ratio r1 in FIG. 6). (2) ), The weight calculated by the calculation unit 23 for the target data TD2 (explanatory variable x2, objective variable y2) at timing 2 is the weight of model 2 (ratio r2 in FIG. 6 (2) ) It is also possible to display the transition of the weights of model 2, including the above. With such a display method, the transition of each weight can be presented to the user in an appropriate manner.

[0063] (Display example 4) 7 shows Example 4 of the display information displayed by the display unit 24. In this example, the display information displayed by the display unit 24 includes the following: In a certain parameter space, the endpoints (V A ~V E ) and the convex hull defined by The transition of the weighted linear sum of the target model (P0 to P4 in the upper part of Figure 7) is shown.

[0064] Here, the weighted linear sum P0 of the target model shown in the upper part of FIG. 7 is, for example, The weights calculated by the calculation unit 23 for the target data TD0 (explanatory variable x0, objective variable y0) at timing 0 are the weights (r0 (A) ,r0 (B) ,r0 (C) ,r0 (D) ,r0 (E) ) as a linear sum of the endpoints of each model with weighting coefficients, P0=r0 (A) V A +r0 (B) V B +r0 (C) V C +r0 (D) V D +r0 (E) V E It is calculated as follows:

[0065] Similarly, the weighted linear sum P0 of the target model shown in the upper part of FIG. 7 is, for example, The weights calculated by the calculation unit 23 for the target data TD1 (explanatory variable x1, objective variable y1) at timing 1 are the weights (r1 (A) ,r1 (B) ,r1 (C) ,r1 (D) ,r1 (E) ) as a linear sum of the endpoints of each model with weighting coefficients, P0=r1 (A) V A +r1 (B) V B +r1 (C) V C +r1 (D) V D +r1 (E) V E The same applies to P2 to P4 shown in the upper part of Fig. 7. Even with this display method, the transition of each weight can be presented to the user in an appropriate manner.

[0066] (Example 5) The lower part of Fig. 7 shows Example 5 of display information displayed by the display unit 24. Explanation of the contents overlapping with Example 4 above will be omitted. In this example, the abnormality detection unit 25 executes an abnormality detection process with reference to the transition of the weight, and this abnormality detection process is included in the display information. More specifically, in this example, the abnormality detection unit 25 executes The weighted linear sum of the target model (P0 to P5 in the lower part of Figure 7) is the endpoint (V in the lower part of Figure 7) for each target model in the parameter space. A ~V E ) and sequentially determining whether the point is inside the convex hull defined by If the weighted linear sum of the target model is not inside the convex hull (corresponding to P5 in the lower part of Figure 7), it is determined that there is an "abnormality" in the transition of the weight. The display information displayed by the display unit 24 includes the information "An abnormality was detected" as the result of the determination. As an example, in a configuration in which sensing data from a certain sensor provided in a production line of a target product is sequentially acquired as the target data TD, the anomaly detection unit 25 executes the anomaly detection process as described above, thereby making it possible to suitably detect an anomaly that has occurred in the production line and present it to the user.

[0067] In this way, by using the above-described display method, the transition of each weight can be presented to the user in an appropriate manner, and the results of the anomaly detection process that references the transition of the weight can also be presented to the user in an appropriate manner.

[0068] Third Exemplary Embodiment A third exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same functions as those described in the above exemplary embodiment will be assigned the same reference numerals, and their description will be omitted as appropriate. The scope of application of each technical means employed in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technical means employed in this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise. Furthermore, each technical means shown in each drawing referenced to explain this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise.

[0069] The positioning of the algorithm of the processing by the information processing device 1B according to this exemplary embodiment will be described. The present inventors have been studying a Linear Parameter-Varying (LPV) model as a modeling of a system with fluctuations. In the LPV model, as an example, an internal state quantity (internal state variable) x k , and the output state quantity (output state variable) y k is updated and calculated using the following formulas (1A) and (1B).

number

[0070] Figure 9 shows the output of each end model in the LPV model (1st SS model to 5th SS model in Figure 9) and the internal division ratio parameter μ (i) k As shown in FIG. 9, the output of multiple endpoint models at the k-th step is (A (i) x k +B (i) u k ) (i=1~5) For each of the internal division ratio parameters μ (i) k (i=1~5) is multiplied, and x at the k+1th step is k+1 is calculated.

[0071] While such an LPV model is suitable for modeling systems with fluctuations, the internal ratio parameter μ (i) k However, there was a problem that it was difficult to apply to systems where the value of is unknown.

[0072] The present inventors The above internal division ratio parameter μ (i) kThe hidden variable (posterior probability) z k Treat it as - Applying hidden variable model learning methods in machine learning, The above internal division ratio parameter μ (i) k the hidden variable z k Calculate as the expected value of By doing so, the inventors have found that it is possible to realize the learning of an LPV model even if the interior ratio parameter μ (i) k The L2PV model (Latent Linear Parameter-Varying model) is defined by the following equations (2A) to (2C), which introduce the following as hidden variables:

number

number

[0073] Each process performed by the information processing device 1A described below is based on the above-mentioned formulation and is a process based on the unique viewpoint of the inventor.

[0074] (Configuration of information processing device 1B) The configuration of the information processing device 1B will be described with reference to Fig. 5. Fig. 8 is a block diagram showing the configuration of the information processing device 1B. As shown in Fig. 8, the information processing device 1B includes a control unit 10B, a storage unit 15B, a communication unit 16A, and an input / output unit 17A.

[0075] (Storage section 15B) First, various data (information) stored in the storage unit 15B will be described. Data referenced by the control unit 10B is stored in the storage unit 15B. Examples of the storage unit 15B include, but are not limited to, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination thereof.

[0076] Examples of data stored in memory unit 15B include, but are not limited to, target data TD, ratio parameter RP, regression coefficient RC, correspondence relationship information CI, distribution information DI, learning result LR, inference data PD, and prediction result PR, as shown in FIG. 5.

[0077] The target data TD is data used in at least one of the learning process and the weight transition calculation process in the information processing device 1B. k ) and state variables (~y k ) is expressed as the following equation (4). k ,~x k ,y k ,~y k are sometimes called features. Also, the state variable x k ,~x k are called explanatory variables, and the state variable y k ,~y k The objective variable is sometimes called a target variable. Furthermore, when the objective variable is the target of derivation, the objective variable is sometimes called a predicted value. These specific names do not limit the contents described in this specification.

number

[0078] The ratio parameter RP is a parameter that defines the relative weights of a plurality of state space models in an LPV model, and is also referred to as a scheduling parameter. The ratio parameter RP may be an internal ratio parameter or an external ratio parameter. In this exemplary embodiment, in order to describe the case where the ratio parameter RP is an internal ratio parameter, the ratio parameter RP is also referred to as an internal ratio parameter RP. The ratio parameter RP is also referred to as a weight parameter RP or a scheduling parameter RP. As an example, the ratio parameter RP is given by the following equation (5) corresponding to each of m models (model 1 to model m):

number

[0079] Also, the internal ratio parameter μ for a certain model j k (j) is the N-dimensional target data x k It can also be expressed as the components of an N-dimensional vector having components corresponding to each of the j-th internal division ratio parameters μ k (j) is the N-dimensional target data x k (k=1 to N) (j) , μ2 (j) , , μ N (j) ) are the components of an N-dimensional vector with

[0080] The ratio parameter RP includes a weight r calculated by a calculation unit 23 (to be described later), which is a weight for each of the plurality of target models. (i) (i=1 to m) are included. Here, the weight r (i) teeth, The weights of each of the plurality of target models when the objective variable included in the target data is expressed as a weighted sum of each predicted value obtained by applying the trained parameters for each of the plurality of target models to the explanatory variables included in the target data. and in this exemplary embodiment, the ratio parameter μ k (j) The learned parameters are distinguished from the learned parameters. For example, the learned parameters refer to parameters (regression coefficients, etc.) calculated by a ratio parameter calculation unit 14 (described later) and stored as the learned results LR.

[0081] The regression coefficient RC is a coefficient in the L2PV regression model. The regression coefficient RC is expressed as the following equation (6).

number

[0082] The distribution information DI includes a covariance matrix Φ of the prior distribution of the latent variables, a covariance parameter η of the prior distribution of the latent variables, and a covariance parameter Ψ of the posterior distribution of the latent variables.

[0083] Hidden variable z k The prior distribution p(z k ) is expressed as the following equation (7).

number

number

number

[0084] As will be described later, in the processing by the information processing device 1B, the hidden variable z k The posterior distribution p(z k |~y k ,~x k , W, η) is expressed under the constraints (constraints) of the following equation (10).

number

[0085] The learning result LR is data that is output by the output unit 15, which will be described later. The learning result LR includes the calculated ratio parameter RP and the calculated regression coefficient RC.

[0086] The inference data PD is an internal state quantity input to the L2PV regression model. The L2PV regression model makes predictions by using the inference data PD as input and applying the calculated regression coefficients RC to each of multiple target models.

[0087] The predicted result PR is a predicted result obtained by the L2PV regression model. Examples of the predicted result PR will be described later.

[0088] (Communication unit 16A) The communication unit 16A is an interface for transmitting and receiving data via a network. Examples of the communication unit 16A include, but are not limited to, communication chips for various communication standards such as Ethernet (registered trademark), Wi-Fi (Wireless Fidelity (registered trademark), and wireless communication standards for mobile data communication networks, and USB-compliant connectors.

[0089] (Input / output section 17A) The input / output unit 17A is an interface that receives input of data and outputs data. Examples of the input / output unit 17A include, but are not limited to, a microphone, a camera, a gaze input device, a keyboard, a touchpad, a speaker, and a liquid crystal display.

[0090] (control unit 10B) The control unit 10B controls each component included in the information processing device 1B. As shown in FIG. 8 , the control unit 10B includes an acquisition unit 11, a regression coefficient calculation unit 12, a covariance calculation unit 13, a ratio parameter calculation unit 14, an output unit 15, an initial value determination unit 16, a convergence determination unit 17, a prediction unit 18, a calculation unit 23, and an anomaly detection unit 25. In this exemplary embodiment, the acquisition unit 11, the regression coefficient calculation unit 12, the covariance calculation unit 13, the ratio parameter calculation unit 14, the initial value determination unit 16, the convergence determination unit 17, and the prediction unit 18 function as an acquisition unit, a regression coefficient calculation unit, a covariance calculation unit, a ratio parameter calculation unit, an initial value determination unit, a convergence determination unit, and a prediction unit, respectively. In this exemplary embodiment, the output unit 15 functions as a storage unit and an output unit.

[0091] Moreover, since the acquisition unit 11 also functions as the first acquisition unit 21 and the second acquisition unit 22 described in exemplary embodiment 2, the acquisition unit 11 will also be referred to as the acquisition unit 11 (21, 22). Furthermore, since the output unit 15 also functions as the output unit 24 described in exemplary embodiment 2, the output unit 15 will also be referred to as the output unit 15 (24). Specific examples of the processing executed by each unit will be described later with reference to different drawings.

[0092] The acquisition unit 11 acquires data via the communication unit 16A or the input / output unit 17A. Examples of the data acquired by the acquisition unit 11 include target data TD, the number of models of multiple target models, and correspondence relationship information CI. Furthermore, as an example, the acquisition unit 11 sequentially acquires target data, similar to the first acquisition unit 11 according to the second exemplary embodiment. Another example of data acquired by the acquisition unit 11 is the hidden variable z k , information on the prior distribution of the data PD. Inference data PD is another example of data acquired by the acquisition unit 11. The acquisition unit 11 stores the acquired data in the storage unit 15A.

[0093] The regression coefficient calculation unit 12 calculates a ratio parameter RP (μ k ) and the target data TD, and calculates the regression coefficient RC for each of the multiple target models. k ) is the initial value of the ratio parameter RP determined by the initial value determination unit 16, which will be described later. k ) is a ratio parameter RP calculated by a ratio parameter calculation unit 14, which will be described later. The regression coefficient calculation unit 12 stores the calculated regression coefficient RC in the storage unit 15A.

[0094] The covariance calculation unit 13 calculates the target data TD and the ratio parameter RP (μ k ), regression coefficient RC, and hidden variable z k By referring to the covariance matrix Φ of the prior distribution of the hidden variable z k The covariance parameter η of the prior distribution of and the hidden variable z k The covariance calculation unit 13 calculates the covariance matrix Ψ of the posterior distribution of the calculated hidden variable z k The covariance parameter η of the prior distribution of and the hidden variable z k and the covariance matrix Ψ of the posterior distribution of the above are stored in the storage unit 15A as distribution information DI.

[0095] The ratio parameter calculation unit 14 calculates the ratio parameter RP by referring to the target data TD and the regression coefficient RC. k Further referring to the covariance matrix Ψ of the posterior distribution of k The ratio parameter calculation unit 14 stores the calculated ratio parameter RP in the storage unit 15A.

[0096] Furthermore, the ratio parameter calculation unit 14 refers to the correspondence information CI and calculates the ratio parameter RP under a constraint condition according to the correspondence. As an example, the ratio parameter calculation unit 14 calculates the ratio parameter RP under a constraint condition on the sum of the ratio parameters RP according to the correspondence.

[0097] Furthermore, the ratio parameter calculation unit 14 calculates the ratio parameter RP using an update equation that includes a normalization term according to the correspondence relationship.

[0098] The output unit 15 outputs the regression coefficient RC and the ratio parameter RP (μ k ) in association with the corresponding relationship. k ) is supplied to the storage unit 15A in association with the correspondence relationship information CI.

[0099] The output unit 15 outputs the regression coefficient RC and the ratio parameter RP ((μ k ), and the learning result LR). As an example, the output unit 15 outputs the regression coefficient RC and the ratio parameter RP (μ k ) to the input / output unit 17A. In this configuration, the output unit 15 may display graphs defined by the regression coefficients R C for at least two of the multiple target models in a manner that allows them to be distinguished from one another. As another example, when the convergence determination unit 17 described later determines that the calculation related to the ratio parameter R P has converged, the output unit 15 outputs the regression coefficient R C and the ratio parameter R P calculated by the ratio parameter calculation unit 14.

[0100] Similarly to the exemplary embodiment 2, the output unit 15 (24) also functions as a display means for displaying, via the input / output unit 17A, the transition of the weight calculated by the calculation unit 23 for each of the target data TD sequentially acquired by the acquisition unit 11. Specific examples of display by the output unit 15 (24) are similar to those in the exemplary embodiment 2, and therefore will not be described here.

[0101] The initial value determination unit 16 determines the initial value of the ratio parameter RP to be referred to by the regression coefficient calculation unit 12. The initial value determination unit 16 determines the initial value of the determined ratio parameter RP (μ k ) is stored in the storage unit 15B.

[0102] The convergence determination unit 17 determines the ratio parameter RP(μ k The convergence determination unit 17 determines whether the calculation for ( ) has converged. The convergence determination unit 17 supplies the determination result to the output unit 15.

[0103] As in exemplary embodiment 2, the calculation unit 23 calculates the weight of each of the multiple target models when the objective variable included in the target data TD is expressed as a weighted sum of each predicted value obtained by applying the learned parameters for each of the multiple target models to the explanatory variables included in the target data TD.

[0104] As an example, similar to the second exemplary embodiment, the calculation unit 23 calculates, for each model i (i=1 to m) included in the plurality of target models (model 1, model 2, . . . model m), Explanatory variables ~x included in the target data TD k , the regression coefficient W (i) By applying k (i) (y without tilde k )of y k (i) = W (i) ~x k Calculated by Each predicted value (yk (1) ,y k (1) ,···,y k (m) ) as a weighted sum of the objective variable ∼y included in the target data TD k (y with tilde k ), that is, ~y k = Σr (i) y k (i) (where Σ is the sum over index i, from i=1 to i=m) k (y with tilde k ) when expressing each weight r of the plurality of target models (i) (i=1~m) is calculated.

[0105] The calculation of the weight by the calculation unit 23 is performed for each of the target data sequentially acquired by the acquisition unit 11. In other words, the calculation of the weight by the calculation unit 23 is performed for each index k. Then, each weight r (i) may depend on the index k. For example, when the index k is expressed as timing t, the calculation unit 23 calculates The weight r is calculated by the above calculation process using the target data TD(t1) acquired at the timing t1. (i) Calculate (t1), The weight r is calculated by the above calculation process using the target data TD(t2) acquired at the timing t2. (i) Calculate (t2) and... The specific processing by the calculation unit 23 is the same as in the second exemplary embodiment, and therefore will not be described here.

[0106] The anomaly detection unit 25 executes an anomaly detection process by referring to the transition of the weight calculated by the calculation unit 23. As an example, the anomaly detection unit 25 may execute the anomaly detection process by identifying one or more change amounts of the weights included in the transition of the weights and comparing the change amounts with a predetermined threshold. For example, the anomaly detection unit 25 may determine that there is an "anomaly" in the transition of the weights if any of the one or more change amounts of the weights included in the transition of the weights is equal to or greater than the threshold, and may determine that there is no anomaly if this is not the case. The specific process by the anomaly detection unit 25 is the same as in the second exemplary embodiment, and therefore a description thereof will be omitted here.

[0107] The prediction unit 18 derives a prediction result by applying the learned parameters for each of the plurality of target models and any of the weights calculated by the calculation unit 23 to the inference data PD acquired by the acquisition unit 11. As an example, the prediction unit 18 derives a prediction result by applying the learned parameters for each of the plurality of target models and any of the weights calculated by the calculation unit 23 to the inference data PD(T2) at a certain time point T2 to the objective variable ∼y(T2) included in the inference data PD(T2), a weight r(i) calculated by the calculation unit 23 with reference to target data TD(T1) at a time T1 prior to the time T2, the target data TD(T1) including an explanatory variable ~x(T1) and a target variable ~y(T1); Trained parameters for each of a plurality of target models trained by a training process that references target data TD(T0) at time T0 before time T1; The prediction result is derived by applying The prediction unit 18 may also be configured to derive the prediction result PR by applying, to the inference data PD, the regression coefficients RC for each of the plurality of target models and the ratio parameters RP calculated by the ratio parameter calculation unit 14, which are determined according to the type of the inference data PD. The prediction unit 18 stores the derived prediction result PR in the storage unit 15B.

[0108] (Example of processing flow in information processing device 1B) 10 is a diagram showing an example of a processing flow in the information processing device 1B according to this exemplary embodiment. Note that the processing example described below can also be regarded as a variational Bayes EM algorithm, but this does not limit this exemplary embodiment. Furthermore, the processing example described below can be regarded as a process of updating each parameter so as to maximize a variational lower bound (VLB) J obtained by the following equation (11).

number

number

[0109] In step S11, the acquisition unit 11 further acquires a parameter m indicating the number of models among the plurality of target models. Here, the number of models m is determined by the internal division ratio parameter vector μ k (i) It can also be expressed as the number of

[0110] In addition, in step S11, the acquisition unit 11 acquires correspondence relationship information CI relating to correspondence relationships between the target data TD and a plurality of target models.

[0111] In step S11, the acquisition unit 11 calculates the hidden variable z k As an example, the acquisition unit 11 acquires information about the prior distribution of the hidden variable z k The prior distribution p(z k ) covariance matrix Φ.k The prior distribution p(z k ) may further be obtained.

[0112] (Step S16: Initial value determination process) Subsequently, in step S16, the initial value determination unit 16 determines the ratio parameter RP (μ k ) is determined. As an example, the initial value determination unit 16 determines the initial value of the ratio parameter RP(μ k ) is determined as a random value. In this way, the initial value determination unit 16 determines the initial value of the ratio parameter RP(μ k By determining the initial value of the ratio parameter RP(μ), the regression coefficient RC can be suitably calculated in the regression coefficient calculation process S12 described later. k ) has been explained in detail, so the explanation will be omitted here.

[0113] (Step S12: Regression coefficient calculation process) Subsequently, in step S12, the regression coefficient calculation unit 12 calculates the ratio parameter (internal division ratio parameter vector) RP expressed as the following equation (13):

number

number

number

number

number

[0114] (Step S13: Covariance calculation process) Subsequently, in step S13, the covariance calculation unit 13 calculates the target data TD expressed as the following equation (18):

number

number

number

number

number

number

[0115] (Step S14: Ratio parameter calculation process) Subsequently, in step S14, the ratio parameter calculation unit 14 calculates the target data TD expressed as the following equation (24):

number

number

number

number

number

number

number

[0116] In addition, in the above constraint equation (29B), M i denotes the set of target data TD corresponding to the i-th endpoint model. In other words, M i indicates a set of target data TD that the correspondence relationship information CI indicates will be used as training data for the i-th end point model. Therefore, the above constraint condition equation (29B) indicates that the ratio parameter μ calculated using the j-th target data TD from the set of target data TD that indicates will be used as training data for the i-th end point model. jThe sum of these ratio parameters RP is 1. That is, the ratio parameter calculation unit 14 calculates the ratio parameters under the constraint condition according to the correspondence relationship, which is a constraint condition on the sum of the ratio parameters RP, by using the above constraint condition formula (29B). On the other hand, the above constraint condition formula (29C) calculates the ratio parameters μ in the end point model that does not use the j-th target data TD as training data. h indicates that ∇ ...

[0117] For example, consider a model that predicts sales of drinking water in a store. As an example, the endpoint models are (1) Model 1, which predicts sales when a measure is implemented, (2) Model 2, which predicts sales when the content volume is changed, (3) Model 3, which predicts sales when the appearance is changed, (4) Model 4, which predicts sales when the external environment changes, and Model 5, which predicts sales under all of the situations (1) to (5).

[0118] As an example of this case, if the target data TD is data on sales when a measure is implemented, the correspondence relationship information CI indicates that the target data TD is used as learning data for Model 1 and Model 5.

[0119] As described above, the ratio parameter calculation unit 14 may calculate the ratio parameter RP using an update equation including a normalization term according to the correspondence relationship. For example, the ratio parameter calculation unit 14 calculates the ratio parameter RP using the following equations (31A) and (31B).

number

number

[0120] (Step S17: Convergence determination process) Subsequently, in step S17, the convergence determination unit 17 determines whether or not the series of processes in the above-described steps S12, S13, and S14 have converged. This can be expressed as determining whether or not the above-described variational Bayes EM algorithm has converged, or determining whether or not the ratio parameter RP (μ k Alternatively, it may be expressed as judging whether or not the calculation for (VLB: Variational Lower Bound) J obtained by the following formula (33):

number

[0121] (Step S15: Output process) In step S17, if the convergence determination unit 17 determines that "convergence has occurred," in step S15, the output unit 15 outputs the regression coefficient RC calculated by the regression coefficient calculation unit 12 in step S12 and the ratio parameter RP (μ k , learning result LR). In this way, the convergence determination unit 17 determines the ratio parameter RP (μ k ) has "converged," the output unit 15 outputs the learning result LR, thereby making it possible to output a suitable learning result LR.

[0122] In step S15, the output unit 15 outputs the hidden variable z calculated by the covariance calculation unit 13 in step S13. k The covariance parameter η of the prior distribution of the hidden variable z k The output unit 15 may be configured to further output the covariance matrix Ψ of the posterior distribution of the hidden variable z k The covariance parameter η of the prior distribution of and the hidden variable z k The covariance matrix Ψ of the posterior distribution of σ can be presented to the user.

[0123] In step S15, the output unit 15 may be configured to display graphs defined by the regression coefficients RC for at least two of the plurality of target models in a manner that allows them to be distinguished from one another.

[0124] (Example of processing executed by the prediction unit 18) The prediction unit 18 calculates the trained parameters (regression coefficients RC) for each of the plurality of target models and any of the weights r calculated by the calculation unit 23 for the inference data PD. (i) The prediction result is derived by applying

[0125] The prediction unit 18 may be configured to derive multiple prediction results PR by applying, to the inference data PD, the regression coefficients RC for each of the multiple target models and the ratio parameters RP calculated by the ratio parameter calculation unit 14, which are determined depending on the type of the inference data PD. As an example, when the inference data PD includes a feature amount (explanatory variable) x0, the prediction unit 18 may apply, to the feature amount x0, the regression coefficients RC for each of the multiple target models and the ratio parameters RP (μ (i) or r (i) ) to derive the prediction result PR. In this case, an example of a graph including the prediction result RP that the output unit 15 displays via the input / output unit 17A is shown in FIG. 11. FIG. 11 shows another example of a graph that the output unit 15 displays via the input / output unit 17A. Here, in the example shown in FIG. 11, the ratio parameter RP is calculated by the ratio parameter calculation unit 14. (i) is used as an example, but this does not limit the present exemplary embodiment. (i) This exemplary embodiment also includes the case where the following is used:

[0126] (Effects of information processing device 1B) As described above, in the information processing device 1B, -Sequentially acquire target data TD, Obtain the trained parameters (such as the regression coefficients included in the training results LR) for each of the multiple target models, a calculation unit 23 calculates a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying trained parameters for each of the plurality of target models to explanatory variables included in the target data TD; The transition of the weight calculated by the calculation means for each of the sequentially acquired target data is displayed by the output unit 15 (24). The above configuration provides an effect that the user can easily recognize the transition of the weight of each of the plurality of target models. In other words, the above configuration provides an exemplary effect that the contribution (weight) of each prediction model (each target model) to the target data can be suitably presented to the user.

[0127] As described above, the trained parameters for each of the plurality of target models include: The regression coefficient W for the target model (i) and, A ratio parameter μ that defines the ratio of the target model (i) and Contains The regression coefficients and the ratio parameter μ (i) teeth, Ratio parameter μ (i) and a regression coefficient calculation process for calculating a regression coefficient for each of a plurality of target models by referring to the learning data (data included in the target data TD that is used for the learning process); and The ratio parameter μ (i) Ratio parameter calculation process to calculate are the regression coefficients and ratio parameters learned by a learning process including

[0128] As described above, the learning process for the regression coefficients and the ratio parameters is performed as follows: the target data TD, the ratio parameter μ (i) , the regression coefficient W (i) and a covariance calculation process for calculating the covariance parameters of the prior distribution of the latent variables and the covariance matrix of the posterior distribution of the latent variables by referring to information on the prior distribution of the latent variables. Further comprising: The ratio parameter calculation process is Calculate the ratio parameter by further referring to the covariance matrix of the posterior distribution of the hidden variables. This process includes: According to the above configuration, it is possible to preferably calculate the learned parameters for each of the plurality of target models.

[0129] Fourth Exemplary Embodiment A fourth exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same functions as those described in the above exemplary embodiment will be assigned the same reference numerals, and their description will be omitted as appropriate. The scope of application of each technical means employed in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technical means employed in this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise. Furthermore, each technical means shown in each drawing referenced to explain this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise.

[0130] (Configuration of information processing device 1C) The configuration of information processing device 1C will be described with reference to Fig. 12. Fig. 12 is a block diagram showing the configuration of information processing device 1C. As shown in Fig. 12, information processing device 1C includes a control unit 10C, a storage unit 15C, a communication unit 16A, and an input / output unit 17A. The communication unit 16A and the input / output unit 17A have been described above, and therefore description thereof will be omitted.

[0131] The memory unit 15C stores data referenced by the control unit 10C, similar to the memory unit 15B described above. Examples of data stored in the memory unit 15C include, but are not limited to, target data TD, ratio parameters RP, regression coefficients RC, correspondence information CI, learning results LR, inference data PD, and prediction results PR, as shown in Fig. 12. The target data TD, ratio parameters RP, regression coefficients RC, correspondence information CI, learning results LR, inference data PD, and prediction results PR are as described above, and therefore will not be described here.

[0132] (Control unit 10C) The control unit 10C controls each component included in the information processing device 1C, similar to the control unit 10B described above. As shown in Fig. 12, the control unit 10C includes an acquisition unit 11 (21, 22), a regression coefficient calculation unit 12, a ratio parameter calculation unit 14, an output unit 15 (24), an initial value determination unit 16, a convergence determination unit 17, a prediction unit 18, a calculation unit 23, and an abnormality detection unit 25. The acquisition unit 11 (21, 22), the output unit 15 (24), the initial value determination unit 16, the convergence determination unit 17, the prediction unit 18, the calculation unit 23, and the abnormality detection unit 25 are as described above, and therefore will not be described again.

[0133] The regression coefficient calculation unit 12 calculates a ratio parameter RP (μ k (i) ) and the target data TD, and calculates the regression coefficients RC for each of the plurality of target models. In this exemplary embodiment, the regression coefficient calculation unit 12 calculates the regression coefficients RC for each of the plurality of target models by referring to the hidden variables z k The covariance parameter η of the prior distribution of and the hidden variable z k The regression coefficients RC are calculated using the least squares method without updating the covariance matrices Φ and Φ of the prior distributions.

[0134] The ratio parameter calculation unit 14 calculates the ratio parameter RP (μ k (i) In this exemplary embodiment, the ratio parameter RP is not limited to an internal ratio parameter, but may also be an external ratio parameter.

[0135] (Example of processing flow in information processing device 1C) 13 is a diagram showing an example of the flow of processing in the information processing device 1C according to this exemplary embodiment. The acquisition processing S11, the initial value determination processing S16, and the output processing S15 are as described above, and therefore description thereof will be omitted.

[0136] (Step S12: Regression coefficient calculation process) In step S12, the regression coefficient calculation unit 12 calculates the ratio parameter RP expressed as the following equation (33):

number

number

number

number

[0137] (Step S14: Ratio parameter calculation process) Subsequently, in step S14, the ratio parameter calculation unit 14 calculates the target data TD expressed as the following equation (37):

number

number

number

number

number

number

[0138] As another example, the convergence determination unit 17 determines whether the value of the ratio parameter RP has converged. For example, in the nth convergence determination process of repeating a series of processes including the above-mentioned steps S12, S13, and S14, the convergence determination unit 17 compares the value of the ratio parameter RP at the (n-1)th time with the value of the ratio parameter RP at the nth time, and determines that the series of processes at the above-mentioned steps S12, S13, and S14 has converged if the absolute value of the difference between them is equal to or smaller than a predetermined threshold.

[0139] (Effects of information processing device 1C) As described above, the information processing device 1C calculates the regression coefficient R C using the least squares method. With this configuration, the information processing device 1C can also preferably calculate the trained parameters for each of the multiple target models, similar to the information processing device 1B.

[0140] Also, in the information processing device 1C, -Sequentially acquire target data TD, Obtain the trained parameters (such as the regression coefficients included in the training results LR) for each of the multiple target models, a calculation unit 23 calculates a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying trained parameters for each of the plurality of target models to explanatory variables included in the target data TD; The transition of the weight calculated by the calculation means for each of the sequentially acquired target data is displayed by the output unit 15 (24). Therefore, similar to the information processing device 1B, the information processing device 1C can also suitably present to the user the degree of contribution (weight) of each prediction model (each target model) to the target data.

[0141] [Software implementation example] Some or all of the functions of the information processing devices 1, 1A, 1B, and 1C (hereinafter also referred to as "each of the above devices") may be realized by hardware such as an integrated circuit (IC chip), or by software.

[0142] In the latter case, each of the above devices is realized by, for example, a computer that executes instructions of a program, which is software that realizes each function. An example of such a computer (hereinafter referred to as computer C) is shown in Figure 14. Figure 14 is a block diagram showing the hardware configuration of computer C that functions as each of the above devices.

[0143] The computer C includes at least one processor C1 and at least one memory C2. The memory C2 stores a program P for causing the computer C to operate as each of the above-mentioned devices. In the computer C, the processor C1 reads and executes the program P from the memory C2, thereby realizing the functions of each of the above-mentioned devices.

[0144] The processor C1 may be, for example, a central processing unit (CPU), a graphic processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a tensor processing unit (TPU), a quantum processor, a microcontroller, or a combination thereof. The memory C2 may be, for example, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination thereof.

[0145] The computer C may further include a RAM (Random Access Memory) for expanding the program P during execution and for temporarily storing various data. The computer C may also include a communication interface for transmitting and receiving data to and from other devices. The computer C may also include an input / output interface for connecting input / output devices such as a keyboard, mouse, display, and printer.

[0146] Furthermore, the program P can be recorded on a non-transitory tangible recording medium M that can be read by the computer C. Such a recording medium M can be, for example, a tape, a disk, a card, a semiconductor memory, or a programmable logic circuit. The computer C can acquire the program P via such a recording medium M. The program P can also be transmitted via a transmission medium. Such a transmission medium can be, for example, a communication network or broadcast waves. The computer C can also acquire the program P via such a transmission medium.

[0147] [Appendix A] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0148] (Appendix A1) a first acquisition means for sequentially acquiring target data; a second acquisition means for acquiring trained parameters for each of a plurality of target models; a calculation means for calculating a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying trained parameters for each of the plurality of target models to explanatory variables included in the target data; a display means for displaying the transition of the weight calculated by the calculation means for each of the sequentially acquired target data; An information processing device comprising:

[0149] (Appendix A2) the second acquisition means further acquires feature information associated with the target model or the trained parameters; The display means further displays feature information about each of the plurality of target models. 10. The information processing device according to claim 1,

[0150] (Appendix A3) The display means Further displaying each regression line obtained by applying the learned parameters for each of the plurality of target models to the target data. 10. The information processing device according to claim 9, wherein the information processing device is a

[0151] (Appendix A4) further comprising an anomaly detection means for executing an anomaly detection process by referring to the transition of the weight; The display means further displays the result of the abnormality detection process. An information processing device according to any one of appendices A1 to A3.

[0152] (Appendix A5) the first acquisition means further acquires inference data; The system further includes a prediction means for deriving a prediction result by applying the learned parameters for each of the plurality of target models and any of the weights calculated by the calculation means to the inference data. An information processing device according to any one of appendices A1 to A4.

[0153] (Appendix A6) The trained parameters for each of the plurality of target models include: The regression coefficients for the target model; A ratio parameter that defines the ratio of the target model; Contains The regression coefficients and the ratio parameters are a regression coefficient calculation process that calculates a regression coefficient for each of a plurality of target models by referring to the ratio parameters and the learning data; and A ratio parameter calculation process for calculating the ratio parameters by referring to the learning data and the regression coefficients. are the regression coefficients and ratio parameters learned by the learning process including An information processing device according to any one of appendices A1 to A5.

[0154] (Appendix A7) The learning process for the regression coefficients and the ratio parameters is a covariance calculation process for calculating covariance parameters of the prior distribution of the latent variables and a covariance matrix of the posterior distribution of the latent variables by referring to the target data, the ratio parameters, the regression coefficients, and information on the prior distribution of the latent variables; Further comprising: The ratio parameter calculation process is Calculate the ratio parameter by further referring to the covariance matrix of the posterior distribution of the hidden variables. 10. The information processing device according to claim 9, wherein the information processing device is a

[0155] [Appendix B] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0156] (Appendix B1) a first acquisition process in which at least one processor sequentially acquires target data; a second acquisition process in which the at least one processor acquires trained parameters for each of a plurality of target models; a calculation process in which the at least one processor calculates a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying trained parameters for each of the plurality of target models to explanatory variables included in the target data; a display process in which the at least one processor displays a transition of the weight calculated by the calculation process for each of the sequentially acquired target data; An information processing method comprising:

[0157] (Appendix B2) The second acquisition process further acquires feature information associated with the target model or the learned parameters, In the display process, the at least one processor further displays feature information about each of the plurality of target models. 1. The information processing method described in Appendix B1.

[0158] (Appendix B3) In the display processing, the at least one processor Further displaying each regression line obtained by applying the learned parameters for each of the plurality of target models to the target data. 1. The information processing method described in Appendix B2.

[0159] (Appendix B4) The at least one processor further includes an anomaly detection process that executes an anomaly detection process by referring to a transition of the weight, In the display process, the at least one processor further displays a result of the anomaly detection process. 1. An information processing method according to any one of appendices B1 to B3.

[0160] (Appendix B5) the first acquisition process further acquires inference data; The method further includes a prediction process in which the at least one processor applies, to the inference data, learned parameters for each of the plurality of target models and any of the weights calculated in the calculation process, to derive a prediction result. 1. An information processing method according to any one of appendices B1 to B4.

[0161] (Appendix B6) The trained parameters for each of the plurality of target models include: The regression coefficients for the target model; A ratio parameter that defines the ratio of the target model; Contains The regression coefficients and the ratio parameters are a regression coefficient calculation process that calculates a regression coefficient for each of a plurality of target models by referring to the ratio parameters and the learning data; and A ratio parameter calculation process for calculating the ratio parameters by referring to the learning data and the regression coefficients. are the regression coefficients and ratio parameters learned by the learning process including 1. An information processing method according to any one of Appendices B1 to B5.

[0162] (Appendix B7) The learning process for the regression coefficients and the ratio parameters is a covariance calculation process for calculating covariance parameters of the prior distribution of the latent variables and a covariance matrix of the posterior distribution of the latent variables by referring to the target data, the ratio parameters, the regression coefficients, and information on the prior distribution of the latent variables; Further comprising: In the ratio parameter calculation process, the at least one processor: Calculate the ratio parameter by further referring to the covariance matrix of the posterior distribution of the hidden variables. An information processing method as described in Appendix B6.

[0163] [Appendix C] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0164] (Appendix C1) A program that causes a computer to function as an information processing device, the program comprising: a first acquisition means for sequentially acquiring target data; a second acquisition means for acquiring trained parameters for each of a plurality of target models; a calculation means for calculating a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying trained parameters for each of the plurality of target models to explanatory variables included in the target data; a display means for displaying the transition of the weight calculated by the calculation means for each of the sequentially acquired target data; An information processing program that functions as a

[0165] (Appendix C2) the second acquisition means further acquires feature information associated with the target model or the trained parameters; The display means further displays feature information about each of the plurality of target models. An information processing program as described in Appendix C1.

[0166] (Appendix C3) The display means Further displaying each regression line obtained by applying the learned parameters for each of the plurality of target models to the target data. An information processing program as described in Appendix C2.

[0167] (Appendix C4) The computer further functioning as an anomaly detection means for executing an anomaly detection process by referring to the transition of the weight; The display means further displays the result of the abnormality detection process. An information processing program according to any one of appendices C1 to C3.

[0168] (Appendix C5) the first acquisition means further acquires inference data; The computer and further functioning as a prediction means for deriving a prediction result by applying the learned parameters for each of the plurality of target models and any of the weights calculated by the calculation means to the inference data. An information processing program according to any one of appendices C1 to C4.

[0169] (Appendix C6) The trained parameters for each of the plurality of target models include: The regression coefficients for the target model; A ratio parameter that defines the ratio of the target model; Contains The regression coefficients and the ratio parameters are a regression coefficient calculation process that calculates a regression coefficient for each of a plurality of target models by referring to the ratio parameters and the learning data; and A ratio parameter calculation process for calculating the ratio parameters by referring to the learning data and the regression coefficients. are the regression coefficients and ratio parameters learned by the learning process including An information processing program according to any one of appendices C1 to C5.

[0170] (Appendix C7) The learning process for the regression coefficients and the ratio parameters is a covariance calculation process for calculating covariance parameters of the prior distribution of the latent variables and a covariance matrix of the posterior distribution of the latent variables by referring to the target data, the ratio parameters, the regression coefficients, and information on the prior distribution of the latent variables; Further comprising: The ratio parameter calculation process is Calculate the ratio parameter by further referring to the covariance matrix of the posterior distribution of the hidden variables. An information processing program as described in Appendix C6.

[0171] [Appendix D] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0172] (Appendix D1) at least one processor, a first acquisition process for sequentially acquiring target data; a second acquisition process for acquiring trained parameters for each of the plurality of target models; a calculation process of calculating a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying trained parameters for each of the plurality of target models to explanatory variables included in the target data; a display process for displaying the transition of the weight calculated by the calculation process for each of the sequentially acquired target data; An information processing device that executes the above.

[0173] The information processing device may further include a memory, and the memory may store a program for causing the at least one processor to execute each of the processes.

[0174] (Appendix D2) The second acquisition process further acquires feature information associated with the target model or the learned parameters, In the display process, the at least one processor further displays feature information about each of the plurality of target models. 10. The information processing device according to claim 9, wherein the information processing device is an information processing device according to claim 1.

[0175] (Appendix D3) In the display processing, the at least one processor Further displaying each regression line obtained by applying the learned parameters for each of the plurality of target models to the target data. 10. The information processing device according to claim 9, wherein the information processing device is a

[0176] (Appendix D4) the at least one processor: further executing an anomaly detection process that executes an anomaly detection process by referring to the transition of the weight; In the display process, the at least one processor further displays a result of the anomaly detection process. An information processing device according to any one of appendices D1 to D3.

[0177] (Appendix D5) the first acquisition process further acquires inference data; the at least one processor: Further performing a prediction process to derive a prediction result by applying the learned parameters for each of the plurality of target models and any of the weights calculated in the calculation process to the inference data. An information processing device according to any one of appendices D1 to D4.

[0178] (Appendix D6) The trained parameters for each of the plurality of target models include: The regression coefficients for the target model; A ratio parameter that defines the ratio of the target model; Contains The regression coefficients and the ratio parameters are a regression coefficient calculation process that calculates a regression coefficient for each of a plurality of target models by referring to the ratio parameters and the learning data; and A ratio parameter calculation process for calculating the ratio parameters by referring to the learning data and the regression coefficients. are the regression coefficients and ratio parameters learned by the learning process An information processing device according to any one of appendices D1 to D5.

[0179] (Appendix D7) The learning process for the regression coefficients and the ratio parameters is a covariance calculation process for calculating covariance parameters of the prior distribution of the latent variables and a covariance matrix of the posterior distribution of the latent variables by referring to the target data, the ratio parameters, the regression coefficients, and information on the prior distribution of the latent variables; Further execute In the ratio parameter calculation process, the at least one processor: Calculate the ratio parameter by further referring to the covariance matrix of the posterior distribution of the hidden variables. 10. The information processing device according to claim 9, wherein the information processing device is an information processing device according to claim 8.

[0180] [Appendix E] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0181] (Appendix E1) A program that causes a computer to function as an information processing device, The computer, a first acquisition process for sequentially acquiring target data; a second acquisition process for acquiring trained parameters for each of the plurality of target models; a calculation process of calculating a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying trained parameters for each of the plurality of target models to explanatory variables included in the target data; a display process for displaying the transition of the weight calculated by the calculation process for each of the sequentially acquired target data; A non-transitory recording medium on which an information processing program for executing the above is recorded. [Explanation of symbols]

[0182] 1, 1A, 1B, 1C Information processing equipment 21 First Acquisition Section 22 Second Acquisition Section 23 Calculation section 24 Display section 25 Abnormality detection unit 11 Acquisition Department 12 Regression coefficient calculation section 13 Covariance calculation section 14 Internal division parameter calculation section 15 Output section 16 Initial value determination section 17 Convergence judgment section 18 Prediction Department

Claims

1. a first acquisition means for sequentially acquiring target data; a second acquisition means for acquiring learned parameters for each of a plurality of target models; a calculation means for calculating a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying trained parameters for each of the plurality of target models to explanatory variables included in the target data; a display means for displaying the transition of the weight calculated by the calculation means for each of the sequentially acquired target data; An information processing device comprising:

2. The second acquisition means further acquires feature information associated with the target model or the trained parameters, The display means further displays feature information about each of the plurality of target models. The information processing device according to claim 1 .

3. The display means Further displaying each regression line obtained by applying the learned parameters for each of the plurality of target models to the target data. The information processing device according to claim 2 .

4. further comprising an anomaly detection means for executing an anomaly detection process by referring to the transition of the weight; The display means further displays the result of the abnormality detection process. The information processing device according to claim 1 .

5. the first acquisition means further acquires inference data; The system further includes a prediction means for deriving a prediction result by applying the learned parameters for each of the plurality of target models and any of the weights calculated by the calculation means to the inference data. The information processing device according to claim 1 .

6. The trained parameters for each of the plurality of target models include: The regression coefficients for the target model; A ratio parameter that defines the ratio of the target model; Contains The regression coefficients and the ratio parameters are a regression coefficient calculation process that calculates a regression coefficient for each of a plurality of target models by referring to the ratio parameters and the learning data; and A ratio parameter calculation process for calculating the ratio parameters by referring to the learning data and the regression coefficients. are the regression coefficients and ratio parameters learned by the learning process including The information processing device according to claim 1 .

7. The learning process for the regression coefficients and the ratio parameters is a covariance calculation process for calculating covariance parameters of the prior distribution of the latent variables and a covariance matrix of the posterior distribution of the latent variables by referring to the target data, the ratio parameters, the regression coefficients, and information on the prior distribution of the latent variables; Further comprising: The ratio parameter calculation process is Calculate the ratio parameter by further referring to the covariance matrix of the posterior distribution of the hidden variables. The information processing device according to claim 6 .

8. sequentially acquiring target data; acquiring trained parameters for each of a plurality of target models; calculating a weight for each of the plurality of target models when a target variable included in the target data is expressed as a weighted sum of predicted values ​​obtained by applying trained parameters for each of the plurality of target models to explanatory variables included in the target data; Displaying the transition of the calculated weight for each of the sequentially acquired target data; An information processing method comprising:

9. 2. A program for causing a computer to function as the information processing device according to claim 1, the program causing a computer to function as the first acquisition means, the second acquisition means, the calculation means, and the display means.

Images