Information processing apparatus, information processing server, information processing method, and non-transitory computer-readable storage medium

Abstract

[Problem] To achieve both privacy protection and high inference accuracy. [Solution] An information processing apparatus is provided, including: a learning unit that clusters hierarchical data based on a plurality of inference models distributed from an information processing server, and performs learning using an inference model corresponding to each cluster; and a communication unit that transmits an intermediate result generated for each cluster in the learning of the learning unit to the information processing server, wherein the hierarchical data includes information specifying a main element and a log collected or generated in association with the main element.

Description

Technical Field

[0001] This disclosure relates to information processing apparatus, information processing server, information processing method, and non-transitory computer-readable storage medium. Background Technology

[0002] In recent years, models for performing certain inferences based on collected data have been developed. Furthermore, techniques for enhancing the accuracy of inferences, as described above, have been proposed. For example, Patent Document 1 discloses a technique for clustering data to be used for inference.

[0003] Reference List

[0004] Patent documents

[0005] Patent Document 1: Japanese Patent Application No. 2020-154825 Summary of the Invention

[0006] The problem to be solved by the present invention

[0007] However, depending on the structure of the data to be used for inference, the technology disclosed in Patent Document 1 may not provide sufficient effect.

[0008] Solution to the problem

[0009] According to one aspect of this disclosure, an information processing apparatus is provided, comprising: a learning unit that clusters hierarchical data based on a plurality of inference models distributed from an information processing server, and performs learning using an inference model corresponding to each cluster; and a communication unit that sends intermediate results generated by the learning unit for each cluster to the information processing server during learning. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements.

[0010] Furthermore, according to another aspect of this disclosure, an image processing method executed by a processor is provided, comprising: clustering hierarchical data based on multiple inference models distributed from an information processing server; performing learning using an inference model corresponding to each cluster; and sending intermediate results generated for each cluster during learning to the information processing server. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements.

[0011] Furthermore, according to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided, storing a program that enables a computer to be used as an information processing apparatus, the information processing apparatus comprising: a learning unit that clusters hierarchical data based on a plurality of inference models distributed from an information processing server, and performs learning using an inference model corresponding to each cluster; and a communication unit that sends intermediate results generated for each cluster during the learning process of the learning unit to the information processing server. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements.

[0012] Furthermore, according to another aspect of this disclosure, an information processing server is provided, comprising: a learning unit that generates multiple inference models corresponding to multiple clusters respectively; and a communication unit that transmits information about the multiple inference models generated by the learning unit to multiple information processing devices. The communication unit receives from the multiple information processing devices intermediate results generated by learning from hierarchical data clustered based on the multiple inference models and an inference model corresponding to each cluster. The learning unit updates the multiple inference models based on the multiple intermediate results. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements.

[0013] Furthermore, according to another aspect of this disclosure, an information processing method executed by a processor is provided, comprising: generating multiple inference models corresponding to multiple clusters respectively; sending information about the generated multiple inference models to multiple information processing devices; receiving intermediate results generated by learning from the multiple information processing devices through hierarchical data clustered based on the multiple inference models and inference models corresponding to each cluster; and updating the multiple inference models based on the multiple intermediate results. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements.

[0014] Furthermore, according to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided, storing a program that enables a computer to function as an information processing server. The information processing server includes: a learning unit that generates multiple inference models corresponding to multiple clusters; and a communication unit that transmits information about the multiple inference models generated by the learning unit to multiple information processing devices. The communication unit receives from the multiple information processing devices intermediate results generated by learning from hierarchical data clustered based on the multiple inference models and an inference model corresponding to each cluster. The learning unit updates the multiple inference models based on the multiple intermediate results.

[0015] Hierarchical data includes information used to specify the main elements, as well as logs collected or generated in association with the main elements. Attached Figure Description

[0016] Figure 1 This is a block diagram illustrating a configuration example of system 1 according to an embodiment of the present disclosure.

[0017] Figure 2 This is a block diagram illustrating an example configuration of the information processing apparatus 10 according to this embodiment.

[0018] Figure 3 This is a block diagram illustrating a configuration example of the information processing server 20 according to this embodiment.

[0019] Figure 4 This is a diagram illustrating an example of a model corresponding to layered data according to this embodiment.

[0020] Figure 5 This is a schematic diagram illustrating the process executed by System 1 according to this embodiment.

[0021] Figure 6 This is a flowchart illustrating the process executed by System 1 according to this embodiment.

[0022] Figure 7 This is a diagram illustrating an example of clustering of hierarchical data according to this embodiment.

[0023] Figure 8 This is a diagram illustrating the operation of system 1 when the main element according to this embodiment is access point 40.

[0024] Figure 9 This is a diagram illustrating the operation of System 1 when the main element according to this embodiment is the product category GC.

[0025] Figure 10 This is a diagram illustrating the operation of System 1 in a scenario where the primary element according to this embodiment is a person and label y is an indicator related to the person's health status.

[0026] Figure 11 This is a diagram illustrating the operation of system 1 in the case where the main element according to this embodiment is a human and the label y is a human expression.

[0027] Figure 12 This is a block diagram illustrating an example of the hardware configuration of an information processing apparatus 90 according to an embodiment of the present disclosure. Detailed Implementation

[0028] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings. Note that in this specification and the drawings, components having substantially the same functional configuration are assigned the same reference numerals, and their descriptions are not repeated.

[0029] Note that the descriptions will be presented in the following order.

[0030] 1. Implementation Method

[0031] 1.1. Overview

[0032] 1.2. System Configuration Example

[0033] 1.3. Configuration Example of Information Processing Device 10

[0034] 1.4. Configuration Example of Information Processing Server 20

[0035] 1.5. Functional Details

[0036] 1.6. Application Examples

[0037] 2. Hardware Configuration Example

[0038] 3. Conclusion

[0039] <1. Implementation Method>

[0040] <<1.1. Overview>>

[0041] First, an overview of the embodiments of this disclosure will be described.

[0042] As mentioned above, in recent years, models (inference models) have been developed for performing certain inferences based on collected data.

[0043] Based on the inference model, various inferences can be made with high accuracy using unknown data. Therefore, the generation and utilization of inference models can be actively implemented in different fields.

[0044] However, based on the generation of the inference model and the data structure used for inference, it may be difficult to generate inference models with high accuracy.

[0045] Examples of the data mentioned above include hierarchical data.

[0046] Hierarchical data can be defined, for example, as including information for specifying primary elements and log data collected or generated in association with the primary elements.

[0047] As an example, suppose device 80 communicates with multiple other devices and performs some inferences based on logs collected for each communication partner.

[0048] In this case, the primary element can be a device that acts as a communication partner of device 80.

[0049] Furthermore, in this case, the generated logs may have characteristics corresponding to communication partners, a pair of devices 80, and communication partners.

[0050] Therefore, when learning from usage logs without distinguishing communication partners, there is a possibility that it is not only difficult to generate inference models with high accuracy, but also that the time required for convergence, overlearning, etc., increases.

[0051] On the other hand, for example, as disclosed in Patent Document 1, there are also methods for clustering data and then performing learning on each data belonging to each cluster.

[0052] As mentioned above, the method of classifying data with different characteristic trends into multiple clusters and performing learning on each data point belonging to each cluster is also useful for hierarchical data.

[0053] However, in the method disclosed in Patent Document 1, it is difficult to generate a highly accurate inference model because the clustering results cannot be corrected based on the inference results.

[0054] Furthermore, privacy protection is also an issue, for example, when a server collects data from multiple devices and performs learning based on that data.

[0055] On the other hand, as a method to protect data privacy when performing learning using data collected from multiple devices, there is also a method called federated learning.

[0056] However, joint learning generally aims to generate a single inference model based on data received from multiple devices.

[0057] Therefore, when data with various trends, such as hierarchical data, are used for general joint learning, there is a possibility of increased time required for convergence, overlearning, and other issues.

[0058] To reduce the aforementioned possibilities, for example, a method could be conceived for clustering data based on the characteristics of the device that collected the data and performing learning on each data point belonging to each cluster.

[0059] However, in this case, all the data collected from a certain device is classified into the same cluster.

[0060] Therefore, when making inferences using logs generated for each of the aforementioned multiple communication partners, it may be difficult to reduce the possibility of increased time required for convergence, overlearning, etc.

[0061] The technical concept of the embodiments of this disclosure has focused on the above-mentioned points and achieves both privacy protection and high inference accuracy.

[0062] Thus, the information processing apparatus 10 according to an embodiment of the present disclosure includes a learning unit 120 that clusters hierarchical data based on a plurality of inference models distributed from the information processing server 20, and performs learning using an inference model corresponding to each cluster.

[0063] Furthermore, the information processing apparatus 10 according to the embodiments of the present disclosure includes a communication unit 130, which sends intermediate results generated for each cluster during the learning process of the learning unit 120 to the information processing server 20.

[0064] That is, the information processing apparatus 10 according to the embodiments of the present disclosure clusters the hierarchical data held by the information processing apparatus itself by using multiple inference models distributed from the information processing server 20, and sends intermediate results generated by learning for each cluster to the information processing server 20.

[0065] Here, intermediate results can be information that includes feature values ​​included in the hierarchical data and values ​​calculated based on labels, and can be information where feature values ​​and labels are difficult to recover.

[0066] On the other hand, the information processing server 20 according to an embodiment of the present disclosure includes: a learning unit 210 that generates multiple inference models corresponding to multiple clusters respectively; and a communication unit that sends information about the multiple inference models generated by the learning unit 210 to multiple information processing devices 10.

[0067] Here, one of the features of the communication unit 220 is that it receives intermediate results from multiple information processing devices 10, which are generated by learning from hierarchical data clustered based on multiple inference models and inference models corresponding to each cluster.

[0068] In addition, one of the features of learning unit 210 is to update multiple inference models based on multiple intermediate results.

[0069] That is, according to the embodiments of the present disclosure, the information processing server 20 updates the inference model corresponding to each cluster based on the intermediate results of each cluster received from the plurality of information processing devices 10, and distributes the updated information to each information processing device 10.

[0070] As described above, in system 1 according to an embodiment of the present disclosure, the clustering and transmission of intermediate results performed by the information processing device 10, as well as the updating of the inference model and the distribution of information related to the update performed by the server, can be repeatedly performed.

[0071] Based on the above processing, the convergence of the inference model can be guaranteed, and the clustering system and inference accuracy can be improved at the same time.

[0072] Furthermore, based on the above processing, privacy protection can be improved by using intermediate results that are difficult to recover from the original data.

[0073] The following sections will describe in detail a system configuration example for implementing the above description.

[0074] <<1.2. System Configuration Example>>

[0075] Figure 1 This is a block diagram illustrating a configuration example of system 1 according to an embodiment of the present disclosure.

[0076] like Figure 1 As shown, the system 1 according to this embodiment includes a plurality of information processing devices 10 and an information processing server 20.

[0077] Each information processing device 10 and information processing server 20 are connected to each other so that they can communicate via network 30.

[0078] It should be noted that Figure 1 The example shown illustrates a system 1 including information processing devices 10a and 10b, but the number of information processing devices 10 according to this embodiment is not limited to this example.

[0079] (Information processing device 10)

[0080] According to this embodiment, the information processing apparatus 10 performs clustering of hierarchical data by using an inference model distributed from the information processing server 20.

[0081] Furthermore, according to this embodiment, the information processing apparatus 10 uses data belonging to each cluster and an inference model corresponding to that cluster to perform learning, and sends intermediate results to the information processing server 20.

[0082] The information processing device 10 according to this embodiment may be, for example, a personal computer, a smartphone, a tablet computer, a game console, a wearable device, etc.

[0083] (Information Processing Server 20)

[0084] According to this embodiment, the information processing server 20 generates an inference model corresponding to the set clustering and distributes the inference model to multiple information processing devices 10.

[0085] Furthermore, according to this embodiment, the information processing server 20 receives intermediate results corresponding to each cluster from multiple information processing devices 10, and updates each inference model based on the intermediate results.

[0086] According to this embodiment, the information processing server 20 distributes information about the updates of each inference model to multiple information processing devices 10.

[0087] (Network 30)

[0088] According to this embodiment, the network 30 mediates the communication between the information processing device 10 and the information processing server 20.

[0089] <<1.3. Configuration Example of Information Processing Device 10>>

[0090] Next, a configuration example of the information processing apparatus 10 according to this embodiment will be described in detail.

[0091] Figure 2 This is a block diagram illustrating an example configuration of the information processing apparatus 10 according to this embodiment.

[0092] like Figure 2 As shown, the information processing apparatus 10 according to this embodiment may include a sensor unit 110, a learning unit 120, and a communication unit 130.

[0093] (Sensor Unit 110)

[0094] According to this embodiment, the sensor unit 110 collects various types of sensor information.

[0095] The sensor information collected by sensor unit 110 can be used as an element (feature value) of the hierarchical data.

[0096] Thus, the sensor unit 110 may include various sensors for collecting sensor information of elements that can be used as hierarchical data.

[0097] On the other hand, if the layered data does not include sensor information, the information processing device 10 may not include the sensor unit 110.

[0098] (Learning Unit 120)

[0099] In this embodiment, the learning unit 120 clusters hierarchical data based on multiple inference models distributed from the information processing server 20.

[0100] Furthermore, the learning unit 120 according to this embodiment performs learning using an inference model corresponding to each cluster.

[0101] As described above, the hierarchical data according to this embodiment may include information for specifying key elements and logs collected or generated in association with the key elements.

[0102] In addition, the learning unit 120 can cluster hierarchical data related to different principal elements based on multiple inference models.

[0103] The functions of the learning unit 120 according to this embodiment are implemented by various processors.

[0104] The details of the function of the learning unit 120 according to this embodiment will be described separately.

[0105] (Communication Unit 130)

[0106] According to this embodiment, the communication unit 130 communicates with the information processing server 20 via the network 30.

[0107] The communication unit 130 receives information from the information processing server 20, such as the inference model and updates to the inference model.

[0108] In addition, the communication unit 130 sends the intermediate results generated by the learning unit 120 to the information processing server 20.

[0109] It should be noted that the communication unit 130 according to this embodiment can communicate with another device different from the information processing server 20.

[0110] Furthermore, in this case, the communication unit 130 can generate and store logs related to communication with another device.

[0111] Logs of communications with another device can be used as part of hierarchical data.

[0112] The configuration example of the information processing apparatus 10 according to this embodiment has been described above. It should be noted that the above refers to... Figure 2 The configuration described is merely an example, and the configuration of the information processing apparatus 10 according to this embodiment is not limited to this example.

[0113] For example, the information processing apparatus 10 according to this embodiment may further include an input unit for receiving information input from a user, a display unit for displaying various types of information, etc.

[0114] Furthermore, when the information processing device 10 includes an input unit, the input information can be used as part of the hierarchical data.

[0115] The configuration of the information processing device 10 according to this embodiment can be flexibly modified according to specifications and operation.

[0116] <<1.4. Configuration Example of Information Processing Server 20>>

[0117] Next, a configuration example of the information processing server 20 according to this embodiment will be described.

[0118] Figure 3 This is a block diagram illustrating a configuration example of the information processing server 20 according to this embodiment.

[0119] like Figure 3 As shown, the information processing server 20 according to this embodiment may include a learning unit 210 and a communication unit 220.

[0120] (Learning Unit 210)

[0121] According to the learning unit 210 of this embodiment, multiple inference models corresponding to multiple clusters are generated respectively.

[0122] Furthermore, according to this embodiment, the learning unit 210 updates multiple inference models based on multiple intermediate results received by the communication unit 220.

[0123] The functions of the learning unit 210 according to this embodiment are implemented by various processors.

[0124] The details of the function of the learning unit 210 according to this embodiment will be described separately.

[0125] (Communication Unit 220)

[0126] According to this embodiment, the communication unit 220 communicates with multiple information processing devices 10 via network 30.

[0127] The communication unit 220 sends, for example, the inference model generated by the learning unit 210 and information about updates to the inference model to multiple information processing devices 10.

[0128] In addition, the communication unit 220 receives intermediate results from multiple information processing devices 10.

[0129] The configuration example of the information processing server 20 according to this embodiment has been described above. It should be noted that reference... Figure 3 The above configuration is merely an example, and the configuration of the information processing server 20 according to this embodiment is not limited to this example.

[0130] For example, the information processing server 20 according to this embodiment may further include an input unit for receiving user information input, a display unit for displaying various information, etc.

[0131] The configuration of the information processing server 20 according to this embodiment can be flexibly modified according to specifications and operation.

[0132] <<1.5. Functional Details>>

[0133] Next, the functions of each of the information processing apparatus 10 and the information processing server 20 according to this embodiment will be described in detail.

[0134] As described above, the information processing apparatus 10 according to this embodiment clusters the hierarchical data held by the information processing apparatus itself by using multiple inference models distributed from the information processing server 20, and sends intermediate results generated by learning for each cluster to the information processing server 20.

[0135] Furthermore, according to this embodiment, the information processing server 20 updates the inference model corresponding to each cluster based on the intermediate results received from the plurality of information processing devices 10 for each cluster, and distributes information about the update to each information processing device 10.

[0136] To achieve the above processing, it is necessary to share the model corresponding to the hierarchical data between the information processing device 10 and the information processing server 20.

[0137] Figure 4 This is a diagram illustrating an example of a model corresponding to layered data in this embodiment.

[0138] exist Figure 4 left side and Figure 4 The right side shows examples of graphical models corresponding to hierarchical data and examples of generative models corresponding to hierarchical data, respectively.

[0139] exist Figure 4 In each model shown, η, θ, and κ correspond to the set of information processing devices 10 (global), information processing devices 10, and principal element, respectively.

[0140] but, Figure 4 Each model shown is merely an example. The graphical and generative models in this implementation can be appropriately designed based on the characteristics of the hierarchical data and the characteristics of the inferred labels (target variables).

[0141] Next, we will refer to Figure 5 and Figure 6 The processing flow performed by System 1 according to this embodiment is described in detail.

[0142] Figure 5 This is a schematic diagram illustrating the process executed by System 1 according to this embodiment.

[0143] also, Figure 6 This is a flowchart illustrating the process performed by system 1 according to this embodiment.

[0144] It should be noted that, Figure 5 The processing of information processing devices 10a and 10b is shown in the figure, but as described above, the system 1 according to this embodiment may include three or more information processing devices 10.

[0145] also, Figure 5 The example shown illustrates a scenario where the information processing server 20 generates three inference models M1 to M3 corresponding to three clusters C1 to C3, but the number of clusters and inference models according to this embodiment is not limited to such an example.

[0146] Based on the number of clusters and the number of inference models in this implementation method, it is only necessary to make appropriate designs according to the characteristics of the hierarchical data and the characteristics of the inferred labels (target variables).

[0147] First, such as Figure 6 As shown, the information processing server 20 initializes the inference model (S100).

[0148] exist Figure 5 In the example shown, the information processing server 20 initializes the inference models M1 to M3.

[0149] Subsequently, the information processing server 20 distributes information related to the inference model (S101).

[0150] exist Figure 5 In the example shown, the information processing server 20 can send all the information that constitutes the inference models M1 to M3.

[0151] Next, each information processing device 10 uses an inference model to cluster the hierarchical data (S102).

[0152] exist Figure 5 In the example shown, the information processing device 10a classifies the hierarchical data D1 into any one of the clusters C1 to C3 corresponding to the inference models M1 to M3 by using the distributed inference models M1 to M3.

[0153] Similarly, the information processing device 10b classifies the hierarchical data D2 into any one of the clusters C1 to C3 corresponding to the inference models M1 to M3 by using the distributed inference models M1 to M3.

[0154] Subsequently, each information processing device performs learning on each cluster (S103).

[0155] exist Figure 5 In the example shown, the information processing device 10a performs learning on each of the clusters C1 to C3 and generates intermediate results w corresponding to clusters C1 to C3 respectively. 11 to w 31 .

[0156] Similarly, the information processing device 10b performs learning on each of the clusters C1 to C3 and generates intermediate results w corresponding to clusters C1 to C3 respectively. 12 to w 32 .

[0157] Subsequently, each information processing device 10 sends the intermediate results to the information processing server 20 (S104).

[0158] exist Figure 5 In the example shown, the information processing device 10a processes the intermediate result w 11 to w 31 Send to the information processing server.

[0159] Similarly, the information processing device 10b will process the intermediate results w 12 to w 32 Send to information processing server 20.

[0160] Subsequently, the information processing server 20 collects the received intermediate results for each cluster and updates the inference model corresponding to the cluster (S105).

[0161] exist Figure 5 In the example shown, information processing server 20 obtains intermediate results w 11 and w 12 Calculate w l And update the inference model M1.

[0162] Similarly, information processing server 20 obtains intermediate results w 21 and w 22 Calculate w2 and update the inference model M2.

[0163] Similarly, information processing server 20 obtains intermediate results w 31 and w 32 Calculate w3 and update the inference model M3.

[0164] Next, the information processing server 20 determines whether each inference model has converged (S106).

[0165] Once the information processing server 20 determines that each inference model has converged (S106: Yes), the system 1 terminates a series of processes.

[0166] On the other hand, if the information processing server 20 determines that each inference model has not converged (S106: Yes), the information processing server 20 returns to step S101 and distributes information related to the inference model.

[0167] For example, in Figure 5 In the example shown, information processing server 20 can...l w2 and w3 are sent to information processing devices 10a and 10b.

[0168] If the information processing server 20 returns to step S101, the information processing device 10 and the information processing server 20 repeat the following process.

[0169] The processing flow of System 1 according to this embodiment has been described in detail above.

[0170] Next, the information sent and received between the information processing apparatus 10 and the information processing server 20 according to this embodiment will be described in more detail.

[0171] Figure 7 This is a diagram illustrating an example of clustering of hierarchical data according to this embodiment.

[0172] exist Figure 7 In the example shown, the hierarchical data with primary element ID: ME1, primary element ID: ME2, and primary element ID: ME3 are classified into cluster C1.

[0173] In addition, the hierarchical data with primary element ID: ME4 and primary element ID: ME5 were classified into cluster C2.

[0174] In addition, the hierarchical data with primary element ID: ME6, primary element ID: ME7, and primary element ID: ME8 were classified into cluster C2.

[0175] Here, the primary element ID is an example of information used to specify the primary element.

[0176] In addition, the hierarchical data includes not only the primary element ID, but also logs collected or generated in association with the primary element.

[0177] Logs can include, for example, feature values ​​and labels (purpose variables).

[0178] Note that in Figure 7 In the example shown, the eigenvalues ​​include x. n1 To x n5 The five elements are merely an example, and the number of elements included in the feature value according to this embodiment is not limited to such an example.

[0179] Intermediate results according to this embodiment may be values ​​calculated based on feature values ​​and labels included in the hierarchical data.

[0180] In such Figure 7 When performing hierarchical data clustering as shown, the information processing device 10 can set the intermediate results w corresponding to clusters C1 to C3 as follows: 11 w21 and w 31 .

[0181] w 11 ={A1, b1}

[0182] w 21 ={A2, b2}

[0183] w 31 ={A3, b3}

[0184] Here, in the above description A k and b k These can be values ​​calculated based on the feature values ​​and labels belonging to cluster Ck, respectively.

[0185] In the following text, A will be described. k and b k Calculation example.

[0186] A1=A(x 11 ,x 12 ,x 13 ,x 14 ,x 15 ,y1,

[0187] x 21 ,x 22 ,x 23 ,x 24 ,x 25 ,y2,

[0188] x 31 ,x 32 ,x 33 ,x 34 ,x 35 ,y3)

[0189] b1=b(x 11 ,x 12 ,x 13 ,x 14 ,x 15 ,y1,

[0190] x 21 ,x 22 ,x 23 ,x 24 ,x 25 ,y2,

[0191] x 31 ,x 32 ,x 33 ,x 34 ,x 35 ,y3)

[0192] A2=A(x41 ,x 42 ,x 43 ,x 44 ,x 45 ,y4,

[0193] x 51 ,x 52 ,x 53 ,x 54 ,x 55 ,y5)

[0194] b2=b(x 41 ,x 42 ,x 43 ,x 44 ,x 45 ,y4,

[0195] x 51 ,x 52 ,x 53 ,x 54 ,x 55 ,y5)

[0196] A3=A(x 61 ,x 62 ,x 63 ,x 64 ,x 65 ,y6,

[0197] x 71 ,x 72 ,x 73 ,x 74 ,x 75 ,y7,

[0198] x 81 ,x 82 ,x 83 ,x 84 ,x 85 ,y8)

[0199] b3=b(x 61 ,x 62 ,x 63 ,x 64 ,x 65 ,y6,

[0200] x 71 ,x 72 ,x 73 ,x 74 ,x 75 ,y7,

[0201] x 81 ,x 82 ,x 83 ,x84 ,x 85 ,y8)

[0202] As described above, the information processing apparatus 10 according to this embodiment can calculate intermediate results without using information for specifying main elements.

[0203] Furthermore, based on the above calculations, as the number of hierarchical data belonging to cluster Ck increases, from A... k and b k Recover the original eigenvalues ​​x ij and tag y i It has become more difficult.

[0204] Therefore, the generation of intermediate results according to this embodiment can effectively enhance privacy protection performance.

[0205] Next, the information regarding the update of the inference model sent by the information processing server 20 to the information processing device 10 will be described in more detail.

[0206] Here, we assume that inference models M1 to M3 are generated, corresponding to clusters C1 to C3 respectively.

[0207] In this case, the information processing server 20 can, for example, calculate the updated information w about the inference models M1 to M3 respectively as follows: l Up to w3.

[0208] w1=(w 11 ,w 12 ,w 13 ,...,w 1n )

[0209] w2=(w 21 ,w 22 ,w 23 ,...,w 2n )

[0210] w3=(w 31 ,w 32 ,w 33 ,...,w 3n )

[0211] Based on the above calculations, as the number n of information processing devices for calculating intermediate results increases, it becomes more difficult to obtain results from w. k Recover the original eigenvalues ​​x ij and tag y i Furthermore, it can effectively enhance privacy protection.

[0212] On the other hand, the information processing device 10 can perform inference with higher accuracy by receiving updated information w1 to w3 about the inference models M1 to M3 respectively.

[0213] For example, when inferring the label y9 that belongs to cluster C3 with feature value x9, the information processing device 10 only needs to calculate f(w3,x9).

[0214] As described above, the learning unit 120 of the information processing apparatus 10 according to this embodiment can infer labels based on feature values ​​and inference models.

[0215] <<1.6. Application Examples>>

[0216] Next, the application of System 1 according to this embodiment will be described with specific examples.

[0217] For example, the main elements according to this embodiment may be various devices that communicate with the information processing device 10.

[0218] In this case, the information processing device 10 can use various logs collected through communication with the device as hierarchical data.

[0219] In the following text, an example will be described where the main element according to this embodiment is an access point that communicates with the information processing device 10.

[0220] Figure 8 This is a diagram illustrating the operation of system 1 when the main element according to this embodiment is access point 40.

[0221] It should be noted that Figure 8 The operation is shown when the information processing device 10 is a smartphone.

[0222] In this example, the information processing device 10 centrally maintains a communication log L40 for each access point 40.

[0223] Communication record L40 is associated with information used to specify access point 40 and is used as hierarchical data.

[0224] At this point, each communication log L40 includes feature values ​​x1 to x n And the tag y.

[0225] For example, received radio wave intensity, CCA busy time, etc., can be used as characteristic values ​​x1 to x2. n .

[0226] In addition, indicators representing the communication quality associated with access point 40 can be used as label y.

[0227] Each of the information processing devices 10 uses multiple inference models distributed from the information processing server 20 to process the aforementioned feature value x. l To x n Cluster the communication logs L40 with label y.

[0228] For example, in Figure 8 In the example shown, the information processing device 10a classifies the communication log L40a for the three records corresponding to access point 40a and the communication log L40b for the two records corresponding to access point 40b into cluster C1.

[0229] In addition, the information processing device 10a classifies the communication log L40c corresponding to the two records corresponding to the access point 40c into cluster C2.

[0230] Similarly, the information processing device 10b classifies the communication log L40a corresponding to the two records of access point 40a into cluster C1, and classifies the communication log L40c corresponding to the two records of access point 40c into cluster C2.

[0231] As described above, the information processing apparatus 10 according to this embodiment can perform clustering, such that hierarchical data related to the same principal element are classified into the same cluster.

[0232] This clustering enables the learning of features corresponding to the predetermined type of access point 40, the predetermined type of information processing device 10, and the pairing of the predetermined type of access point 40, as well as inferences with higher accuracy.

[0233] In addition, the information processing device 10 can perform clustering, so that hierarchical data related to different principal elements are classified into the same cluster.

[0234] This clustering method can suppress the number of clusters and improve learning efficiency.

[0235] After performing the above clustering, each of the information processing apparatus 10 according to this embodiment sends the intermediate results of the above calculation to the information processing server 20 and receives information about the update of the inference model.

[0236] Subsequently, each of the information processing apparatus 10 according to this embodiment infers the label y by using the collected feature value x and the inference model.

[0237] For example, the information processing device 10 can infer the communication quality when using a specific access point 40 by using an inference model from the received radio wave intensity from the access point 40.

[0238] Furthermore, in this case, the information processing device 10 can perform control, such as connecting to an access point 40 in which the inferred communication quality meets predetermined conditions, or connecting to the access point 40 among a plurality of access points 40 that has the highest inferred communication quality.

[0239] The operation has been described above in the case where the main element according to this embodiment is a device that communicates with the information processing device 10.

[0240] Next, an example will be described where the main element according to this embodiment is a product category.

[0241] The main element according to this embodiment is not necessarily a device.

[0242] Figure 9 This is a diagram illustrating the operation of System 1 when the main element according to this embodiment is the product category GC.

[0243] Notice, Figure 9 The operation is shown when the information processing device 10 is a game console and the product is a game.

[0244] In this example, the information processing device 10 maintains a purchase log Lgc uniformly for each category of game GC.

[0245] Purchase logs (Lgc) correspond to information used for the category GC of a specified game and are used as hierarchical data.

[0246] At this point, each purchase log Lgc includes feature values ​​x1 to x n And the tag y.

[0247] For example, game manufacturers, sales orders, etc., can be used as feature values ​​x1 to x2. n .

[0248] In addition, metrics related to the purchase of games belonging to category GC (e.g., whether the game was purchased or retained) can be used as label y.

[0249] Each information processing device 10 uses multiple inference models distributed from the information processing server 20 to analyze the aforementioned feature value x. l To x n Cluster the purchase logs Lgc of tag y.

[0250] For example, in Figure 9 In the example shown, the information processing device 10a classifies the purchase log Lgc1, which contains three records corresponding to the game category GC1, and the purchase log Lgc2, which contains two records corresponding to the game category GC2, into cluster C1.

[0251] In addition, the information processing device 10a classifies the purchase logs Lgc3 of the two records corresponding to the game category GC3 into cluster C2.

[0252] Similarly, the information processing device 10a classifies the purchase logs Lgcl corresponding to the game category GC1 into cluster C1, and the purchase logs Lgc3 corresponding to the game category GC3 into cluster C3.

[0253] After performing the above clustering, each of the information processing apparatus 10 according to this embodiment sends the intermediate results of the above calculation to the information processing server 20 and receives information about the update of the inference model.

[0254] Subsequently, each of the information processing apparatus 10 according to this embodiment infers the label y by using the collected feature value x and the inference model.

[0255] For example, the information processing device 10 can use an inference model to infer the likelihood of a user purchasing a specific game.

[0256] Furthermore, in this case, the information processing device 10 can perform controls, such as explicitly presenting the user with games whose purchase probability exceeds a threshold, or setting the games to be easily visible to the user in the online store.

[0257] Next, an example will be described where the main element according to this embodiment is a human being.

[0258] In this context, the label y can be, for example, an indicator representing a person's physical or mental condition.

[0259] Figure 10 This is a diagram illustrating the operation of System 1 in a scenario where the primary element according to this embodiment is a person and label y is an indicator related to the person's health status.

[0260] It should be noted that Figure 10 The operation of the information processing device 10 is shown when it is installed in a medical facility.

[0261] In this example, the information processing device 10 maintains a unified check log Lpe for each individual P.

[0262] Check the feature quantity x contained in the log LPE. l To x n Examples include various test results such as blood pressure, heart rate, and symptoms.

[0263] In addition, the label y can be a doctor's diagnosis.

[0264] Each information processing device 10 uses multiple inference models distributed from the information processing server 20 to analyze the aforementioned feature value x. l To x n Cluster the Lpe check logs with label y.

[0265] For example, in Figure 10 In the example shown, the information processing device 10a classifies the inspection log Lpe1 for the three records corresponding to person P1 and the inspection log Lpe2 for the two records corresponding to person P2 into cluster C1.

[0266] In addition, the information processing device 10a classifies the inspection log Lpe3, which is used to record the two records corresponding to person P3, into cluster C2.

[0267] Similarly, the information processing device 10a classifies the inspection log Lpe4 for the two records corresponding to person P4 into cluster C1, and classifies the inspection log Lpe5 for the two records corresponding to person P5 into cluster C3.

[0268] After performing the above clustering, each of the information processing apparatus 10 according to this embodiment sends the intermediate results of the above calculation to the information processing server 20 and receives information about the update of the inference model.

[0269] Subsequently, each of the information processing apparatus 10 according to this embodiment infers the label y by using the collected feature value x and the inference model.

[0270] For example, the information processing device 10 can infer a person's health status based on new examination results about the person using an inference model.

[0271] This inference allows for the temporary determination of a person's health condition without the need for an actual diagnosis by a doctor or other physician.

[0272] Next, we will describe the case where label y is an indicator of human emotion.

[0273] For example, indicators of human emotion include a user's facial expressions.

[0274] Figure 11 This is a diagram illustrating the operation of system 1 in the case where the main element according to this embodiment is a human and the label y is a human expression.

[0275] Notice, Figure 11 The operation is shown in the case where the information processing device 10 is a robot that communicates with the user.

[0276] In this example, the information processing device 10 maintains an imaging log Lpp for each individual P uniformly.

[0277] Feature values ​​x1 to x in the imaging log Lpp n Examples include imaging images, the location of various parts of the face, and the size of each part.

[0278] Furthermore, the label y can be various inferential expressions.

[0279] Each information processing device 10 uses multiple inference models distributed from the information processing server 20 to analyze the aforementioned feature value x. l To x n Cluster the imaging logs Lpp with label y.

[0280] For example, in Figure 11 In the example shown, the information processing device 10a classifies the imaging log Lpp1 for the three records corresponding to person P1 and the imaging log Lpp2 for the two records corresponding to person P2 into cluster C1.

[0281] In addition, the information processing device 10a classifies the imaging log Lpp3 for the three records corresponding to human P3 into cluster C2.

[0282] Similarly, the information processing device 10a classifies the imaging log Lpp4 for the three records corresponding to human P4 into cluster C1, and classifies the imaging log Lpp5 for the three records corresponding to human P5 into cluster C3.

[0283] After performing the above clustering, each of the information processing apparatus 10 according to this embodiment sends the intermediate results of the above calculation to the information processing server 20 and receives information about the update of the inference model.

[0284] Subsequently, each of the information processing apparatus 10 according to this embodiment infers the label y by using the collected feature value x and the inference model.

[0285] For example, the information processing device 10 can use an inference model to infer the expression of a person from an image of that person.

[0286] Furthermore, each of the information processing devices 10 according to this embodiment can perform controls such as changing the behavior of the user based on inferred facial expressions, etc.

[0287] <2. Hardware Configuration Example>

[0288] Next, an example of the hardware configuration common to the information processing apparatus 10 and the information processing server 20 according to embodiments of the present disclosure will be described.

[0289] Figure 12This is a block diagram illustrating an example hardware configuration of an information processing apparatus 90 according to an embodiment of the present disclosure.

[0290] The information processing device 90 may be a device with a hardware configuration equivalent to that of the information processing device 10 and the information processing server 20.

[0291] Information processing device 90 includes, for example, a processor 871, a read-only memory (ROM) 872, a random access memory (RAM) 873, a host bus 874, a bridge 875, an external bus 876, an interface 877, an input device 878, an output device 879, a memory 880, a driver 881, a connection port 882, and a communication device 883, as shown in FIG19. Note that the hardware configuration shown here is an example, and some components may be omitted. Furthermore, components other than those shown here may also be included.

[0292] (Processor 871)

[0293] For example, processor 871 is used as an arithmetic processing device or control device, and controls the overall operation of each component or part thereof based on various programs recorded in ROM 872, RAM 873, memory 880, or removable storage medium 901.

[0294] (ROM 872 and RAM 873)

[0295] ROM 872 is a unit that stores programs read by processor 871, data used for calculations, etc. RAM 873 temporarily or permanently stores, for example, programs read by processor 871, various parameters that change appropriately when the program is executed, etc.

[0296] (Host bus 874, bridge 875, external bus 876, and interface 877)

[0297] The processor 871, ROM 872, and RAM 873 are interconnected via a host bus 874, for example, capable of high-speed data transfer. Alternatively, the host bus 874 may be connected via a bridge 875 to an external bus 876, which has a relatively low data transfer speed. Furthermore, the external bus 876 is connected to various components via an interface 877.

[0298] (Input device 878)

[0299] As an input device 878, components such as a mouse, keyboard, touch panel, button, switch, and lever can be used, for example. Alternatively, a remote control (hereinafter referred to as a remote control) capable of transmitting control signals using infrared or other radio waves can be used as an input device 878. Furthermore, the input device 878 includes a voice input device such as a microphone.

[0300] (Output device 879)

[0301] Output device 879 is a device capable of visually or audibly notifying a user of acquired information, such as a display device (e.g., a cathode ray tube (CRT), LCD, or OLED), an audio output device (e.g., a speaker or headphones), a printer, a mobile phone, or a fax machine. Furthermore, output device 879 according to this disclosure includes various vibrating devices capable of outputting tactile stimuli.

[0302] (Memory 880)

[0303] The memory 880 is a device for storing various types of data. Examples of memory 880 include magnetic storage devices such as hard disk drives (HDDs), semiconductor storage devices, optical storage devices, magneto-optical storage devices, etc.

[0304] (Driver 881)

[0305] The drive 881 is, for example, a device for reading information recorded on or writing information to a removable storage medium 901 such as a disk, optical disk, magneto-optical disk, or semiconductor memory.

[0306] (Removable storage medium 901)

[0307] The removable storage medium 901 can be, for example, a DVD media, a Blu-ray (registered trademark) media, an HD DVD media, or various semiconductor storage media. Of course, the removable storage medium 901 can also be, for example, an IC card or electronic device with a contactless IC chip installed thereon.

[0308] (Connect to port 882)

[0309] Connection port 882 is a port for connecting external connection devices 902, such as a Universal Serial Bus (USB) port, an IEEE 1394 port, a Small Computer System Interface (SCSI) port, an RS-232C port, or an optical audio terminal.

[0310] (External connection device 902)

[0311] External connection device 902 is, for example, a printer, portable music player, digital camera, digital imaging machine, IC recorder, etc.

[0312] (communication device 883)

[0313] Communication device 883 is a communication device for connecting to a network, such as a wired or wireless LAN, Bluetooth (registered trademark) or a communication card for wireless USB (WUSB), a router for optical communication, a router for asymmetric digital subscriber line (ADSL) or a modem for various communications, etc.

[0314] <3. Conclusion>

[0315] As described above, the information processing apparatus 10 according to an embodiment of the present disclosure includes a learning unit 120, which clusters hierarchical data based on a plurality of inference models distributed from the information processing server 20, and performs learning using an inference model corresponding to each cluster.

[0316] Furthermore, the information processing apparatus 10 according to an embodiment of the present disclosure includes a communication unit 130, which sends intermediate results generated by the learning unit 120 for each cluster during learning to the information processing server 20.

[0317] With the above configuration, both privacy protection and high inference accuracy can be achieved.

[0318] Preferred embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings; however, the scope of the present disclosure is not limited to these examples. Obviously, those skilled in the art will be able to conceive of various modifications or alterations within the scope of the technical concept set forth in the claims, and these modifications or alterations also fall within the scope of the present invention.

[0319] Furthermore, each step associated with the processing described in this specification need not be processed sequentially in the order described in the flowchart or sequence diagram. For example, each step associated with the processing of each device may be processed in a different order than that described, or may be processed in parallel.

[0320] Furthermore, the series of processes performed by each device described in this specification can be implemented using any software, hardware, and combination of software and hardware. For example, the programs constituting the software are located internally or externally to each device and are pre-stored in a non-transitory computer-readable medium. Each program is then read into RAM when executed by a computer and executed by various processors. Storage media include, for example, magnetic disks, optical disks, magneto-optical disks, flash memory, etc. Additionally, computer programs can be distributed via, for example, a network without using storage media.

[0321] Furthermore, the effects described in this specification are merely exemplary or illustrative, and not restrictive. That is, in addition to or in lieu of the effects described above, the technology according to this disclosure may provide other effects that are obvious to those skilled in the art from the description in this specification.

[0322] It should be noted that the following configurations also fall within the technical scope of this disclosure.

[0323] (1) An information processing apparatus is provided, comprising: a learning unit that clusters hierarchical data based on multiple inference models distributed from an information processing server, and performs learning using an inference model corresponding to each cluster; and a communication unit that sends intermediate results generated for each cluster during the learning process of the learning unit to the information processing server. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements.

[0324] (2) In the information processing apparatus according to (1) above, the learning unit clusters hierarchical data related to different main elements based on multiple inference models.

[0325] (3) In the information processing apparatus according to (1) or (2) above, the learning unit performs clustering such that hierarchical data related to the same principal element are classified into the same cluster.

[0326] (4) In the information processing apparatus according to any one of (1) to (3) above, the learning unit performs clustering such that hierarchical data associated with different principal elements are classified into the same cluster.

[0327] (5) In the information processing apparatus according to any one of (1) to (4) above, the log includes feature values ​​and tags.

[0328] (6) In the information processing apparatus according to (5) above, the intermediate results include values ​​calculated from feature values ​​and tags.

[0329] (7) In the information processing apparatus according to (5) or (6) above, the learning unit infers the label based on the feature value and the inference model.

[0330] (8) In the information processing apparatus according to any one of (1) to (7) above, the communication unit receives information related to the inference model and transmits the information to the learning unit, and the inference model is updated based on intermediate results received by the information processing server from multiple devices.

[0331] (9) In any one of (5) to (7) above, the main element of the information processing apparatus includes a device for communicating with a communication unit.

[0332] (10) In the information processing apparatus according to (9) above, the main elements include an access point for communicating with a communication unit.

[0333] (11) In the information processing apparatus according to (10) above, the tag includes an indicator representing the communication quality associated with the access point.

[0334] (12) In the information processing apparatus according to any one of (5) to (7) above, the main element includes the category of the product.

[0335] (13) In the information processing apparatus according to (12) above, the label includes indicators related to the purchase of products belonging to a category.

[0336] (14) In any one of (5) to (7) above, the main element in the information processing apparatus includes a person.

[0337] (15) In the information processing apparatus according to (14) above, the tag includes an indicator representing a person’s physical or mental condition.

[0338] (16) In the information processing apparatus according to (14) above, the tag includes an indicator representing a person’s emotions.

[0339] (17) A processor-executed image processing method is provided, comprising: clustering hierarchical data based on multiple inference models distributed from an information processing server; performing learning using an inference model corresponding to each cluster; and sending intermediate results generated during learning for each cluster to the information processing server. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements.

[0340] (18) A non-transitory computer-readable storage medium is provided, storing a program that enables a computer to be used as an information processing apparatus, the information processing apparatus comprising: a learning unit that clusters hierarchical data based on a plurality of inference models distributed from an information processing server, and performs learning using an inference model corresponding to each cluster; and a communication unit that sends intermediate results generated for each cluster during the learning process of the learning unit to the information processing server. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements.

[0341] (19) An information processing server is provided, comprising: a learning unit that generates multiple inference models corresponding to multiple clusters respectively; and a communication unit that sends information about the multiple inference models generated by the learning unit to multiple information processing devices. The communication unit receives from the multiple information processing devices intermediate results generated by learning based on hierarchical data clustered according to the multiple inference models and the inference model corresponding to each cluster. The learning unit updates the multiple inference models based on the multiple intermediate results. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements.

[0342] (20) A processor-executed information processing method is provided, comprising: generating multiple inference models corresponding to multiple clusters respectively; sending information about the generated multiple inference models to multiple information processing devices; receiving intermediate results generated by learning from the multiple information processing devices based on hierarchical data clustered according to the multiple inference models and inference models corresponding to each cluster; and updating the multiple inference models based on the multiple intermediate results. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements.

[0343] (21) A non-transitory computer-readable storage medium is provided, storing a program that enables a computer to be used as an information processing server, the information processing server comprising: a learning unit that generates multiple inference models corresponding to multiple clusters respectively; and a communication unit that transmits information about the multiple inference models generated by the learning unit to multiple information processing devices. The communication unit receives from the multiple information processing devices intermediate results generated by learning based on hierarchical data clustered according to the multiple inference models and an inference model corresponding to each cluster, the learning unit updating the multiple inference models based on the multiple intermediate results, the hierarchical data including information for specifying principal elements and logs collected or generated in association with the principal elements.

[0344] Reference Symbol List

[0345] 1 System

[0346] 10. Information processing device

[0347] 110 sensor unit

[0348] 120 Learning Units

[0349] 130 Communication Units

[0350] 20 Information Processing Servers

[0351] 210 Learning Unit

[0352] 220 Communication Units

[0353] 40 access points

Claims

1. An information processing apparatus, comprising: The learning unit clusters hierarchical data associated with different principal elements based on multiple inference models distributed from an information processing server, and performs learning using an inference model corresponding to each cluster. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements, the logs including feature values ​​and labels. as well as The communication unit sends the intermediate results generated for each cluster during the learning process of the learning unit to the information processing server. The intermediate results include values ​​calculated based on the feature values ​​and labels belonging to each cluster, and as the number of hierarchical data belonging to the cluster increases, it becomes more difficult to recover the original feature values ​​and labels from the intermediate results.

2. The information processing apparatus according to claim 1, wherein, The learning unit performs clustering, such that hierarchical data related to the same principal element are classified into the same cluster.

3. The information processing apparatus according to claim 1, wherein, The learning unit performs clustering, such that hierarchical data associated with different principal elements are classified into the same cluster.

4. The information processing apparatus according to claim 1, wherein, The learning unit infers the label based on the feature value and the inference model.

5. The information processing apparatus according to claim 1, wherein, The communication unit receives information about the inference model and transmits the information to the learning unit, the inference model being updated based on intermediate results received by the information processing server from multiple devices.

6. The information processing apparatus according to claim 4, wherein, The main elements include devices that communicate with the communication unit.

7. The information processing apparatus according to claim 6, wherein, The main elements include an access point that communicates with the communication unit.

8. The information processing apparatus according to claim 7, wherein, The label includes metrics that indicate the communication quality associated with the access point.

9. The information processing apparatus according to claim 1, wherein, The key elements include the product category.

10. The information processing apparatus according to claim 9, wherein, The label includes indicators related to purchasing products belonging to that category.

11. The information processing apparatus according to claim 1, wherein, The main element is people.

12. The information processing apparatus according to claim 11, wherein, The label includes indicators representing the person's physical or mental condition.

13. The information processing apparatus according to claim 11, wherein, The labels include indicators representing the person's emotions.

14. An image processing method executed by a processor, comprising: The hierarchical data associated with different principal elements is clustered based on multiple inference models distributed from an information processing server, and learning is performed using an inference model corresponding to each cluster. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements, the logs including feature values ​​and labels. as well as The intermediate results generated for each cluster during the learning process will be sent to the information processing server. The intermediate results include values ​​calculated based on the feature values ​​and labels belonging to each cluster, and as the number of hierarchical data belonging to the cluster increases, it becomes more difficult to recover the original feature values ​​and labels from the intermediate results.

15. A non-transitory computer-readable storage medium storing a program for causing a computer to perform an image processing method, the image processing method comprising: Clustering of hierarchical data associated with different principal elements is performed based on multiple inference models distributed from an information processing server, and learning is performed using the inference model corresponding to each cluster. The hierarchical data includes information specifying principal elements and logs collected or generated in association with said principal elements, the logs including feature values ​​and labels. The intermediate results generated for each cluster during the learning process will be sent to the information processing server. The intermediate results include values ​​calculated based on the feature values ​​and labels belonging to each cluster, and as the number of hierarchical data belonging to the cluster increases, it becomes more difficult to recover the original feature values ​​and labels from the intermediate results.

16. An information processing server, comprising: The learning unit generates multiple inference models, each corresponding to a different cluster. as well as The communication unit transmits information about the multiple inference models generated by the learning unit to multiple information processing devices. The communication unit receives from multiple information processing devices hierarchical data related to different principal elements clustered according to multiple inference models and intermediate results generated by learning the inference model corresponding to each cluster. The hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements, the logs including feature values ​​and labels. The learning unit updates multiple inference models based on multiple intermediate results, and The intermediate results include values ​​calculated based on the feature values ​​and labels belonging to each cluster, and as the number of hierarchical data belonging to the cluster increases, it becomes more difficult to recover the original feature values ​​and labels from the intermediate results.

17. An information processing method executed by a processor, comprising: Generate multiple inference models, each corresponding to a different cluster; Information about the generated multiple inference models is sent to multiple information processing devices; The information processing devices receive intermediate results generated by learning hierarchical data associated with different principal elements based on clusters of the multiple inference models and the inference model corresponding to each cluster, wherein the hierarchical data includes information for specifying principal elements and logs collected or generated in association with the principal elements, the logs including feature values ​​and labels; as well as The inference models are updated based on the intermediate results. The intermediate results include values ​​calculated based on the feature values ​​and labels belonging to each cluster, and as the number of hierarchical data belonging to the cluster increases, it becomes more difficult to recover the original feature values ​​and labels from the intermediate results.

18. A non-transitory computer-readable storage medium storing a program for causing a computer to perform an information processing method, the information processing method comprising: Generate multiple inference models, each corresponding to a different cluster; as well as Information about the multiple inference models generated is sent to multiple information processing devices. The system receives from multiple information processing devices hierarchical data associated with different principal elements, clustered according to multiple inference models, and intermediate results generated through learning from the inference model corresponding to each cluster. The hierarchical data includes information specifying principal elements and logs collected or generated in association with the principal elements, the logs including feature values ​​and labels. The multiple inference models are updated based on the multiple intermediate results, and The intermediate results include values ​​calculated based on the feature values ​​and labels belonging to each cluster, and as the number of hierarchical data belonging to the cluster increases, it becomes more difficult to recover the original feature values ​​and labels from the intermediate results.

Images