Information processing apparatus, terminal management system, terminal management method, and program

Abstract

To select an optimal inference model for a terminal device from among a plurality of inference models.SOLUTION: An information processing apparatus (terminal device 1, cloud server 2) includes: a first acquisition unit (control unit 11, control unit 21) configured to acquire inference result information output from each of a plurality of inference models; a second acquisition unit (control unit 11, control unit 21) configured to acquire evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed by the terminal device; and an output unit (control unit 11, control unit 21) configured to output the inference result information and the evaluation result information.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information processing device, a terminal management system, a terminal management method, and a program. [Background technology]

[0002] Traditionally, inference (AI inference) has been performed using inference models that use AI (artificial intelligence) on sensor devices such as cameras and sensors (IoT devices, IoT: Internet of Things).In addition, services have been provided by performing inference using AI-based inference models on management servers (Edge devices) that manage sensing data from sensor devices. In such cases, an inference model is distributed to the sensor device or management server from a cloud server connected to the sensor device or management server, and the inference model is installed and updated.

[0003] For example, the following cases may be considered as reasons and purposes for updating an inference model: - When the AI ​​model's functionality is expanded -If there is a security risk in the AI ​​model - When the expected accuracy cannot be achieved due to the installation environment of the sensor device (change of location, season, time of day, etc.) -When the load on the AI ​​model is high and you want to improve it Users can select an inference model to be applied from among the inference models provided at their own discretion, and apply the model to perform inference on terminal devices such as IoT devices or Edge devices. Service providers can also select an inference model to be applied from among the inference models prepared by the service providers at their own discretion, and apply the model to perform inference on terminal devices.

[0004] Patent Document 1 describes that the inference accuracy of a plurality of inference models is compared, and the one with the highest accuracy is applied to the terminal device. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2023-28824 Summary of the Invention [Problem to be solved by the invention]

[0006] However, there are limitations and characteristics to performing AI inference on a terminal device, such as the following: Terminal devices have low performance values ​​such as processing speed, so they cannot quickly perform high-load processing. Terminal devices are susceptible to external influences (changes in installation location, changes in input values ​​due to seasons or time of day, etc.) Terminal devices may handle highly confidential data, including personal information (such as people's faces). Therefore, when comparing AI inference models on terminal devices with the limitations and characteristics described above, the inference model with the highest inference accuracy is not necessarily the optimal inference model for the target terminal device.

[0007] Therefore, the object of the present invention is to select the most appropriate inference model from a plurality of inference models in a terminal device. [Means for solving the problem]

[0008] In order to solve the above problem, the information processing device according to the present invention comprises: a first acquisition unit that acquires inference result information output from each of a plurality of inference models; A second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on a terminal device; an output unit that outputs the inference result information and the evaluation result information; Equipped with.

[0009] Further, the information processing device according to the present invention comprises: a first acquisition unit that acquires inference result information output from each of a plurality of inference models; A second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on a terminal device; a judgment unit that compares the inference result information and the evaluation result information and judges the optimal inference model from the plurality of inference models; Equipped with.

[0010] Furthermore, the terminal management system according to the present invention comprises: A terminal management system having a plurality of inference models, comprising a terminal device and a terminal management server, and having a function of processing predetermined information and outputting inference result information, The plurality of inference models are distributed from a terminal management server to the terminal device; a first acquisition unit that acquires inference result information output from each of the plurality of inference models; a second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on the terminal device; an output unit that outputs the inference result information and the evaluation result information; Equipped with.

[0011] Furthermore, the terminal management system according to the present invention comprises: A terminal management system having a plurality of inference models, comprising a terminal device and a terminal management server, and having a function of processing predetermined information and outputting inference result information, The plurality of inference models are distributed from a terminal management server to the terminal device; a first acquisition unit that acquires inference result information output from each of the plurality of inference models; a second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on the terminal device; a judgment unit that compares the inference result information and the evaluation result information and judges the optimal inference model from the plurality of inference models; Equipped with.

[0012] Further, the terminal management method according to the present invention comprises: A control unit of an information processing device having a plurality of inference models and having a function of processing predetermined information and outputting inference result information, a first acquisition step of acquiring inference result information output from each of the plurality of inference models; A second acquisition step of acquiring evaluation result information for specified evaluation items during inference in each of the plurality of inference models executed on a terminal device; an output step of outputting the inference result information and the evaluation result information; Includes.

[0013] Further, the terminal management method according to the present invention comprises: A control unit of an information processing device having a plurality of inference models and having a function of processing predetermined information and outputting inference result information, a first acquisition step of acquiring inference result information output from each of the plurality of inference models; A second acquisition step of acquiring evaluation result information for specified evaluation items during inference in each of the plurality of inference models executed on a terminal device; a judgment step of comparing the inference result information and the evaluation result information to determine an optimal inference model from the plurality of inference models; Includes.

[0014] In addition, the program according to the present invention is A computer of an information processing device having a plurality of inference models and having a function of processing predetermined information and outputting inference result information, a first acquisition unit that acquires inference result information output from each of the plurality of inference models; a second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on a terminal device; an output unit that outputs the inference result information and the evaluation result information; Function as.

[0015] In addition, the program according to the present invention is A computer of an information processing device having a plurality of inference models and having a function of processing predetermined information and outputting inference result information, a first acquisition unit that acquires inference result information output from each of the plurality of inference models; a second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on a terminal device; a judgment unit that compares the inference result information and the evaluation result information and determines an optimal inference model from the plurality of inference models; Function as. [Effects of the Invention]

[0016] According to the present invention, the optimal inference model can be selected from multiple inference models in a terminal device. [Brief explanation of the drawings]

[0017] [Figure 1] FIG. 2 is a block diagram showing the configuration of a terminal management system. [Figure 2] FIG. 10 is a flow diagram of a comparison process. [Figure 3] FIG. 10 is an image diagram of a comparison result display screen. [Figure 4] FIG. 10 is a flow diagram of an update process. DETAILED DESCRIPTION OF THE INVENTION

[0018] Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the scope of the present invention is not limited to the following embodiments and drawings.

[0019] <Terminal Management System 100> First, the configuration of the terminal management system 100 will be described using Fig. 1. As shown in Fig. 1, the terminal management system 100 includes a terminal device 1 (information processing device), a cloud server 2 (information processing device), an administrator terminal device 3, and an analysis device 4. The terminal device 1, the cloud server 2, the administrator terminal device 3, and the analysis device 4 transmit and receive information via a communication network N.

[0020] The terminal device 1 is an information processing device (IoT device or Edge device) that performs AI processing of predetermined information generated by a sensor unit 16 such as a camera using an inference model. The inference model is distributed to the terminal device 1 by a cloud server 2 using an AI container. Multiple inference models are also distributed to the terminal device 1, and the terminal device 1 switches between the inference models it uses. The terminal device 1 is installed in, for example, shopping malls, companies, factories, and nursing homes, and detects intruders, falls, and other abnormalities. The cloud server 2 is an information processing device that distributes AI containers containing inference models for executing the above-mentioned AI processing to the terminal device 1. Note that the cloud server 2 may also be an on-premise device. The administrator terminal device 3 is a terminal device for an administrator who manages the terminal management system 100, and is a terminal device for operating the cloud server 2. The analysis device 4 is a device that analyzes the inference model, and specifically, is a device that outputs security risk information such as the results of vulnerability diagnosis of OSS (Open Source Software) used by the inference model and the results of vulnerability diagnosis of source code. The communication network N is a LAN (Local Area Network), a WAN (Wide Area Network), the Internet, or the like.

[0021] <Terminal Device 1> Next, the configuration of the terminal device 1 will be described with reference to Fig. 1. As shown in Fig. 1, the terminal device 1 includes a control unit 11, an operation unit 12, a communication unit 13, a storage unit 14, a display unit 15, and a sensor unit 16.

[0022] The control unit 11 is configured with a CPU (Central Processing Unit), RAM (Random Access Memory), etc. The CPU of the control unit 11 reads out various programs stored in the storage unit 14, expands them in the RAM, executes various processes in accordance with the expanded programs, and controls the operation of each unit of the terminal device 1. The control unit 11 functions as a first acquisition unit that acquires inference result information output from each of the multiple inference models. Inference result information is information output by the inference model. The inference result information includes not only numerical inference values, but also information obtained by processing the inference values, such as information on whether or not there is an intruder or whether or not there has been a fall. Furthermore, the information (test data) input to the inference model may be, for example, actual measurement data from a predetermined period in the past. Specifically, the information input to the inference model may include data at each time, data for each season, and data corresponding to each discrimination target (normal / abnormal / no target), etc. The control unit 11 functions as a second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of a plurality of inference models executed on the terminal device. The evaluation result information is information about the terminal device 1 when the inference model is executed, and includes, for example, the CPU load and execution time when the inference model is executed. The control unit 11 functions as a third acquisition unit that acquires security risk information of multiple inference models. The control unit 11 functions as an output unit that outputs inference result information and evaluation result information. The control unit 11 may also output a comparison result obtained by comparing security risk information. The control unit 11 functions as a judgment unit that compares the inference result information and the evaluation result information and determines the optimal inference model from multiple inference models. The control unit 11 quantifies the acquired information based on the weight assigned to each piece of information and determines the optimal inference model based on the numerical value. The control unit 11 may not only compare the inference result information and the evaluation result information, but also compare security risk information. The control unit 11 functions as a notification control unit that notifies the user when there is no inference model whose calculated numerical value is equal to or greater than a predetermined value. The notification unit is not limited to the display unit 15, but may also be the display unit of the administrator terminal device 3. The control unit 11 may make a judgment not only based on the inference result information and evaluation result information, but also including security risk information.

[0023] The operation unit 12 is a keyboard equipped with cursor keys, numeric input keys, various function keys, etc., a pointing device such as a mouse, a touch panel laminated on the surface of the display unit 15, etc. The operation unit 12 is configured to be operable by an operator. The operation unit 12 also outputs various signals to the control unit 11 based on operations performed by the operator.

[0024] The communication unit 13 is capable of transmitting and receiving various signals and various data to and from other devices connected via the communication network N.

[0025] The storage unit 14 is configured by a non-volatile semiconductor memory, a hard disk, or the like, and stores various programs executed by the control unit 11, parameters required for executing the programs, various data, and the like. For example, the memory unit 14 stores multiple inference models that have the function of processing specified information and outputting inference result information. Multiple inference models include not only different inference models, but also upgraded versions of the same inference model.

[0026] The display unit 15 is configured with a monitor such as an LCD (Liquid Crystal Display), and displays various screens and the like according to instructions of a display signal input from the control unit 11. The display unit 15 functions as a notification unit that notifies the user, and for example, the display unit 15 displays the comparison results of the inference models described below.

[0027] <Cloud Server 2> Next, the configuration of the cloud server 2 will be described with reference to Fig. 1. As shown in Fig. 1, the cloud server 2 includes a control unit 21, a communication unit 22, and a storage unit .

[0028] The control unit 21 is configured with a CPU, RAM, etc. The CPU of the control unit 21 reads out various programs stored in the storage unit 23, expands them in the RAM, executes various processes in accordance with the expanded programs, and controls the operation of each unit of the cloud server 2. The control unit 21 includes the inference model in an AI container and distributes it to the terminal device 1 via the communication unit 22.

[0029] The communication unit 22 is capable of transmitting and receiving various signals and various data to and from other devices connected via the communication network N.

[0030] The storage unit 23 is configured by a non-volatile semiconductor memory, a hard disk, or the like, and stores various programs executed by the control unit 21, parameters required for executing the programs, various data, and the like.

[0031] <Comparison process> Next, the comparison process in the terminal device 1 will be described with reference to FIG. The comparison process is a process of comparing the performance of multiple inference models. The comparison process may be started, for example, when a new version of an inference model of an inference model already stored in the terminal device 1 is distributed. The comparison process may also be started, for example, when a signal to start the comparison process is input by the user using the operation unit 12.

[0032] First, the control unit 11 inputs test data to a plurality of inference models, causes the plurality of inference models to perform inference, and acquires inference result information (step S1; first acquisition step). In step S1, the inference in the multiple inference models does not necessarily have to be performed within the terminal device 1, but may be performed within the cloud server 2, since the purpose is to obtain inference result information.

[0033] Next, the control unit 11 acquires evaluation result information during inference of the multiple inference models executed by the terminal device 1 (step S2; second acquisition step). In step S2, the inferences for the multiple inference models are performed to obtain evaluation result information at the time of inference, and therefore must be performed within the terminal device 1. In the example of the comparison process shown in Figure 2, steps S1 and S2 are performed within the same terminal device 1, so the inferences for the multiple inference models only need to be performed once within the terminal device 1.

[0034] Next, the control unit 11 acquires security risk information from the analysis device 4 (step S3; third acquisition step).

[0035] Next, the control unit 11 generates a comparison result from the inference result information, the evaluation result information, and the security risk information, and outputs the comparison result to the display unit 15 (step S4; output step). The control unit 11 may output the comparison result to the administrator terminal device 3.

[0036] Here, the comparison result to be output will be described. For example, the comparison result P1 shown in FIG. Area A1 is an area that displays the changes over time in the inferred values ​​from each inference model. Area A2 is an area that displays the results of comparing each inference model for each item. Items include, for example, average error, maximum error, average CPU usage, maximum CPU usage, execution time, presence or absence of security risk, security risk details, etc. The average error and maximum error are generated by the control unit 11 comparing the inference result information for each inference model. The error is the difference between the expected output information when test data is input and the actually output information. In the example shown in Figure 3, the output information from the currently used inference model A is considered to be correct, and the difference between that output information and the output information from inference models B to C is considered to be the error.

[0037] This allows the user to check the comparison results and select the most suitable inference model for the terminal device 1 from multiple inference models.

[0038] The control unit 11 may determine the optimal inference model based on the information obtained from the inference result information, the evaluation result information, and the security risk information, and may indicate this in the comparison result. For example, the control unit 11 may determine that the model with the shortest execution time is the optimal inference model. Furthermore, the control unit 11 may weight the information obtained from each of the inference result information, the evaluation result information, and the security risk information, as described below, and then determine the optimal inference model. For example, the control unit 11 may weight the error average obtained from the inference result information and place importance on the error average when making a determination.

[0039] <Update process> Next, the update process in the terminal device 1 will be described with reference to FIG. The update process involves comparing the performance of multiple inference models and updating to the most optimal inference model. Similar to the comparison process described above, the update process may be initiated, for example, when a new version of an inference model of an inference model already stored in the terminal device 1 is distributed. The update process may also be initiated, for example, when a signal to start the update process is input by the user using the operation unit 12.

[0040] Steps S11 to S13 are the same as steps S1 to S3 in the comparison process, and therefore a description thereof will be omitted.

[0041] Next, the control unit 11 determines whether there is an inference model that satisfies a predetermined condition (step S14). If there is (step S14; YES), the control unit 11 proceeds to the update process at step S15. If there is not (step S14; NO), the control unit 11 proceeds to the update process at step S16.

[0042] Here, the determination method will be specifically described. The control unit 11 assigns scores based on a predetermined threshold for each item, such as the average error, maximum error, average CPU usage, maximum CPU usage, execution time, and presence or absence of security risk, as described in the comparison process, and further recalculates and sums the scores based on the assigned weights to quantify the evaluation for each inference model. For example, the control unit 11 can weight the average error obtained from the inference result information and make a judgment with an emphasis on the average error. Then, the control unit 11 determines whether there is an inference model whose calculated numerical value is equal to or greater than a predetermined value.

[0043] Next, the control unit 11 updates the inference model (step S15; judgment step). As described above, the control unit 11 determines the optimal inference model based on the information obtained from each of the inference result information, evaluation result information, and security risk information. The control unit 11 then updates the inference model. For example, the control unit 11 may determine that the model with the best performance is the optimal inference model based on the calculated numerical values.

[0044] This allows the terminal device 1 to automatically select the most suitable inference model from multiple inference models. In addition, the control unit 11 can acquire an inference model suitable for each time, period, and season in advance, and the control unit 11 can switch the inference model to be executed depending on the time, period, and season.

[0045] <Other> In the above description, the control unit 11 of the terminal device 1 functions as a first acquisition unit, a second acquisition unit, an output unit, a third acquisition unit, a determination unit, and a notification control unit, but this is not limited to this. For example, the control unit 21 of the cloud server 2 may have the functions of a first acquisition unit, a second acquisition unit, an output unit, a third acquisition unit, a determination unit, and a notification control unit. In other words, any control unit included in the terminal management system 100 can have the functions of a first acquisition unit, a second acquisition unit, an output unit, a third acquisition unit, a determination unit, and a notification control unit.

[0046] <Effects> As described above, the information processing device (terminal device 1, cloud server 2) comprises a first acquisition unit (control unit 11, control unit 21) that acquires inference result information output from each of a plurality of inference models, a second acquisition unit (control unit 11, control unit 21) that acquires evaluation result information for specified evaluation items during inference in each of a plurality of inference models executed by the terminal device 1, and an output unit (control unit 11, control unit 21) that outputs the inference result information and evaluation result information. Therefore, by checking the comparison results, the user can select the most suitable inference model for the terminal device 1 from among multiple inference models.

[0047] In addition, the information processing device (terminal device 1, cloud server 2) is equipped with a first acquisition unit (control unit 11, control unit 21) that acquires inference result information output from each of the multiple inference models, a second acquisition unit (control unit 11, control unit 21) that acquires evaluation result information for specified evaluation items at the time of inference in each of the multiple inference models executed by the terminal device 1, and a judgment unit (control unit 11, control unit 21) that compares the inference result information and the evaluation result information and determines the optimal inference model from the multiple inference models. Therefore, the most suitable inference model is automatically selected from multiple inference models in the terminal device 1.

[0048] The terminal management system 100 is a terminal management system having a terminal device 1 and a terminal management server (cloud server 2), and has a plurality of inference models with the function of processing predetermined information and outputting inference result information, wherein the plurality of inference models are distributed from the terminal management server (cloud server 2) to the terminal device 1, and includes a first acquisition unit (control unit 11, control unit 21) that acquires inference result information output from each of the plurality of inference models, a second acquisition unit (control unit 11, control unit 21) that acquires evaluation result information for predetermined evaluation items at the time of inference in each of the plurality of inference models executed by the terminal device 1, and an output unit (control unit 11, control unit 21) that outputs the inference result information and the evaluation result information. Equipped with. Therefore, by checking the comparison results, the user can select the most suitable inference model for the terminal device 1 from among multiple inference models.

[0049] In addition, the terminal management system 100 is a terminal management system having a terminal device 1 and a terminal management server (cloud server 2), and has a plurality of inference models that have the function of processing specified information and outputting inference result information, and the plurality of inference models are distributed from the terminal management server (cloud server 2) to the terminal device 1, and the terminal management system is equipped with a first acquisition unit (control unit 11, control unit 21) that acquires inference result information output from each of the plurality of inference models, a second acquisition unit (control unit 11, control unit 21) that acquires evaluation result information for specified evaluation items at the time of inference in each of the plurality of inference models executed by the terminal device 1, and a judgment unit (control unit 11, control unit 21) that compares the inference result information and the evaluation result information and determines the optimal inference model from the plurality of inference models. Therefore, the most suitable inference model is automatically selected from multiple inference models in the terminal device 1.

[0050] In addition, the terminal management method (comparison process) includes a first acquisition step (step S1) in which a control unit of an information processing device having a plurality of inference models, which has the function of processing specified information and outputting inference result information, acquires inference result information output from each of the plurality of inference models, a second acquisition step (step S2) in which evaluation result information for specified evaluation items at the time of inference in each of the plurality of inference models executed by the terminal device 1, and an output step (step S4) in which the inference result information and the evaluation result information are output. Therefore, by checking the comparison results, the user can select the most suitable inference model for the terminal device 1 from among multiple inference models.

[0051] In addition, the terminal management method (update processing) includes a first acquisition step (step S11) in which a control unit of an information processing device having a plurality of inference models, which has the function of processing specified information and outputting inference result information, acquires inference result information output from each of the plurality of inference models, a second acquisition step (step S12) in which evaluation result information for specified evaluation items at the time of inference in each of the plurality of inference models executed by the terminal device 1 is acquired, and a judgment step (step S15) in which the inference result information and the evaluation result information are compared and the optimum inference model is determined from the plurality of inference models. Therefore, the most suitable inference model is automatically selected from multiple inference models in the terminal device 1.

[0052] The program also causes a computer of an information processing device (terminal device 1, cloud server 2) having multiple inference models and having the function of processing specified information and outputting inference result information to function as a first acquisition unit (control unit 11, control unit 21) that acquires inference result information output from each of the multiple inference models, a second acquisition unit (control unit 11, control unit 21) that acquires evaluation result information for specified evaluation items at the time of inference in each of the multiple inference models executed on the terminal device 1, and an output unit (control unit 11, control unit 21) that outputs the inference result information and the evaluation result information. Therefore, by checking the comparison results, the user can select the most suitable inference model for the terminal device 1 from among multiple inference models.

[0053] The program also causes the computer of an information processing device (terminal device 1, cloud server 2) having multiple inference models and having the function of processing specified information and outputting inference result information to function as a first acquisition unit (control unit 11, control unit 21) that acquires inference result information output from each of the multiple inference models, a second acquisition unit (control unit 11, control unit 21) that acquires evaluation result information for specified evaluation items at the time of inference for each of the multiple inference models executed on the terminal device 1, and a judgment unit (control unit 11, control unit 21) that compares the inference result information and the evaluation result information and determines the optimal inference model from the multiple inference models. Therefore, the most suitable inference model is automatically selected from multiple inference models in the terminal device 1.

[0054] The present invention has been specifically described above based on an embodiment, but it goes without saying that the present invention is not limited to the above embodiment and can be modified as appropriate within the scope of the gist of the present invention. For example, in the above description, examples have been disclosed in which a hard disk or a semiconductor nonvolatile memory is used as a computer-readable medium for the program according to the present invention, but the present invention is not limited to this example. Other computer-readable media may also be portable recording media such as a CD-ROM.

[0055] In addition, the detailed configuration and operation of each device can be modified as appropriate without departing from the spirit of the invention. [Explanation of symbols]

[0056] 100 Terminal Management System 1. Terminal equipment 11 control unit (first acquisition unit, second acquisition unit, output unit, third acquisition unit, determination unit, notification control unit) 12 Control section 13 Communications Department 14 Storage section 15 Display section (notification section) 16 Sensor section 2. Cloud Server 21 control unit (first acquisition unit, second acquisition unit, output unit, third acquisition unit, determination unit, notification control unit) 21 Communications Department 21 storage section 3. Administrator terminal device (notification section) 4 Analysis device N Communication Network

Claims

1. a first acquisition unit that acquires inference result information output from each of a plurality of inference models; a second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on a terminal device; an output unit that outputs the inference result information and the evaluation result information; An information processing device comprising:

2. a first acquisition unit that acquires inference result information output from each of a plurality of inference models; a second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on a terminal device; a judgment unit that compares the inference result information and the evaluation result information and judges the optimal inference model from the plurality of inference models; An information processing device comprising:

3. The information processing device according to claim 1 or claim 2, wherein the evaluation result information includes CPU load and execution time when the inference model is executed.

4. Further comprising a third acquisition unit that acquires security risk information of the plurality of inference models, The information processing device according to claim 1 , wherein the output unit additionally outputs a comparison result of the security risk information.

5. Further comprising a third acquisition unit that acquires security risk information of the plurality of inference models, The information processing device according to claim 2 , wherein the determination unit compares the inference result information, the evaluation result information, and the security risk information, and determines an optimal inference model from the plurality of inference models.

6. The information processing device according to claim 2 , wherein the determination unit converts the acquired information into a numerical value based on a weight assigned to each piece of information, and determines an optimal inference model based on the numerical value.

7. The information processing device according to claim 2 , wherein the judgment unit includes a notification control unit that notifies the user if there is no inference model whose calculated numerical value is greater than or equal to a predetermined value.

8. A terminal management system having a plurality of inference models, comprising a terminal device and a terminal management server, and having a function of processing predetermined information and outputting inference result information, The plurality of inference models are distributed from a terminal management server to the terminal device; a first acquisition unit that acquires inference result information output from each of the plurality of inference models; a second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on the terminal device; an output unit that outputs the inference result information and the evaluation result information; A terminal management system comprising:

9. A terminal management system having a plurality of inference models, comprising a terminal device and a terminal management server, and having a function of processing predetermined information and outputting inference result information, The plurality of inference models are distributed from a terminal management server to the terminal device; a first acquisition unit that acquires inference result information output from each of the plurality of inference models; a second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on the terminal device; a judgment unit that compares the inference result information and the evaluation result information and judges the optimal inference model from the plurality of inference models; A terminal management system comprising:

10. A control unit of an information processing device having a plurality of inference models and having a function of processing predetermined information and outputting inference result information, a first acquisition step of acquiring inference result information output from each of the plurality of inference models; A second acquisition step of acquiring evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on a terminal device; an output step of outputting the inference result information and the evaluation result information; A device management method including:

11. A control unit of an information processing device having a plurality of inference models and having a function of processing predetermined information and outputting inference result information, a first acquisition step of acquiring inference result information output from each of the plurality of inference models; A second acquisition step of acquiring evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on a terminal device; a judgment step of comparing the inference result information and the evaluation result information to determine an optimal inference model from the plurality of inference models; A device management method including:

12. A computer of an information processing device having a plurality of inference models and having a function of processing predetermined information and outputting inference result information, a first acquisition unit that acquires inference result information output from each of the plurality of inference models; a second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on a terminal device; an output unit that outputs the inference result information and the evaluation result information; A program that functions as a

13. A computer of an information processing device having a plurality of inference models and having a function of processing predetermined information and outputting inference result information, a first acquisition unit that acquires inference result information output from each of the plurality of inference models; a second acquisition unit that acquires evaluation result information for a predetermined evaluation item during inference in each of the plurality of inference models executed on a terminal device; a judgment unit that compares the inference result information and the evaluation result information and determines an optimal inference model from the plurality of inference models; A program that functions as a

Images