JP2018206200A - Management device, method for management, and program - Google Patents

Abstract

To pay a contributor an appropriate commission according to usage of a learning model.SOLUTION: A management device 1 includes: a storage unit 11, which relates a learning model created by one or more contributors and the contribution rate of each contributor to creation of the learning model to each other and holds the learning model and the contribution rates to each other; a usage reception unit 15, which receives a request to use the learning model from a user terminal 3 via a network; and a commission calculation unit 19, which calculates each commission in a case where the price paid for usage of the learning model by the user terminal 3 is distributed to each contributor according to the contribution rate.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a management apparatus, a management method, and a program for distributing a consideration paid according to use of a learning model to a contributor.

  Conventionally, a learning model for predicting output data with respect to input data has been created based on machine learning (see, for example, Patent Document 1).

Japanese Patent Laying-Open No. 2015-060237

  The learning model has a plurality of layers created by learning using a neural network. In creating a learning model, a learning model may be created by adding a new layer to a learning model created by another person. In this case, the learning model is created by a contributor as a plurality of creators.

  By the way, it is required to pay a contributor appropriately for the use of the learning model. However, conventionally, paying a contributor appropriately for use of a learning model has not been considered.

  Therefore, the present invention has been made in view of these points, and provides a management device, a management method, and a program capable of appropriately paying a contributor according to the use of a learning model. With the goal.

  The management apparatus according to the first aspect of the present invention includes a storage unit that associates and holds a learning model created by one or more contributors and a contribution rate related to creation of the learning model of each contributor, via a network A receiving unit that receives a request for use of the learning model from another device, and a consideration paid in return for the use of the learning model by the other device, each of which is distributed to each contributor according to the contribution rate A reward calculation unit for calculating a reward.

  The learning model is a weight of each of a plurality of layers generated by learning using a neural network, and the management device calculates the contribution rate based on a ratio of the number of layers generated by each contributor to the total number of layers. You may further provide a contribution rate management part.

The contribution rate management unit may calculate the contribution rate based on a data amount used by the contributor to learn the neural network in order to generate the learning model.
The contribution rate management unit may update the contribution rate when the learning model is updated.

The management apparatus further includes a hash generation unit that generates a data string including a hash value of a weight of each layer of the learning model as an element, and the contribution rate management unit is configured to change the contribution when the data string is changed. The rate may be updated.
The management device may further include a quantization unit that quantizes the weight of each layer, and the hash generation unit may calculate a hash value of the weight quantized by the quantization unit.

  In the management method according to the second aspect of the present invention, the processor holds contribution rate data in which a learning model created by one or more contributors and a contribution rate related to creation of the learning model of each contributor are associated with each other. A step of reading out the contribution rate data from the storage unit, a step of accepting a use request for the learning model from another device via a network, and a consideration paid in return for the use of the learning model by the other device. And a step of calculating each reward when distributing to each contributor in accordance with the contribution rate.

  The program according to the third aspect of the present invention stores, in a computer, contribution rate data in which a learning model created by one or more contributors and a contribution rate related to creation of the learning model of each contributor are associated with each other. A function for reading out the contribution rate data, a function for receiving a request to use the learning model from another device via a network, and a consideration paid in return for the use of the learning model by the other device, And a function of calculating each reward when distributing to each contributor in accordance with the contribution rate.

  According to the present invention, there is an effect that a reward can be appropriately paid to a contributor in accordance with the use of a learning model.

It is a figure which shows the outline | summary of the management apparatus which concerns on this embodiment. It is a figure which shows the structure of the management apparatus which concerns on this embodiment. It is a figure which shows an example of the learning model information which concerns on this embodiment. It is a figure which shows an example of the license information which concerns on this embodiment. It is a figure which shows the structure of the model registration part which concerns on this embodiment. It is a flowchart which shows the flow of the process which concerns on the registration or update of the learning model which concerns on this embodiment. It is a flowchart which shows the flow of the process which concerns on use of the learning model which concerns on this embodiment, and reward calculation.

[Outline of Management Device 1]
FIG. 1 is a diagram illustrating an overview of a management apparatus 1 according to the present embodiment. The management device 1 is a device that accepts registration of a learning model of a neural network and accepts use of the registered learning model for a fee. The management device 1 is communicably connected to the contributor terminal 2 and the user terminal 3 via a communication network such as a LAN (Local Area Network) or the Internet.

  In describing this embodiment, first, a neural network will be briefly described. The basic configuration of the neural network is expressed by superposition (or graph structure) of a plurality of types of layers.

  FIG. 1 shows a forward propagation type neural network. A forward-propagation neural network is composed of three layers: an input layer, a hidden layer, and an output layer, and is directed in one direction from the input layer to the output layer. Propagate. The hidden layer can be composed of a plurality of layers in a graph. Each layer has a plurality of units (neurons). In each layer, the parameter of the function connected from the unit in the front layer to the unit in the rear layer is referred to as “weight”. Learning in the present embodiment means calculating an appropriate “weight” as a parameter of this function. Moreover, the learning model in this embodiment is the weight of each of a plurality of layers, and is used to realize a task.

  Learning in the neural network means that the weight of each layer is optimally updated using an error between the output value from the output layer for the training data and the label of the training data. In order to calculate the error, a “loss function” is defined. The error propagates one after another from the output layer side to the input layer side by the “error back propagation method”, and the weight of each layer is updated little by little. A learning model is created by executing a convergence calculation that adjusts the weight of each layer to an appropriate value so that the error is finally reduced.

  By the way, a user who creates a learning model may create a learning model by updating an existing weight by re-learning a learning model created in advance or adding a new layer. In this case, a plurality of users contribute to the creation of the learning model. Hereinafter, the user who contributed to the creation of the learning model is referred to as “contributor”.

  In the example shown in FIG. 1, a learning model M is generated. The learning model M includes weights W1 to Wn. Here, as shown in FIG. 1, the contributor A creates weights W1 to Wi corresponding to the input layer and a part of the hidden layer, and the weights Wi + 1 to Wn corresponding to the other part of the hidden layer are the contributor B. Is created. In this case, contributor A and contributor B contribute to the creation of learning model M.

  The management apparatus 1 according to the present embodiment associates the learning model with the contribution rate of each contributor to the creation of the learning model in order to appropriately distribute the consideration paid in return for the use of the learning model to each contributor. (1 in FIG. 1).

  In addition, the management apparatus 1 accepts license registration of the learning model from the user terminal 3 used by the user who wants to use the learning model ((2) in FIG. 1). Here, the license is managed by associating a model ID as model identification information for identifying the learning model with a user ID for identifying the user of the user who uses the learning model. The model identification information is generated based on hash values of one or more subsets obtained by dividing the learning model. By doing in this way, since the learning model created using the other learning model includes the model ID corresponding to the other learning model, the management apparatus 1 can include one or more contributors who created the learning model. Can be specified.

Thereafter, the management apparatus 1 accepts a learning model use request ((3) in FIG. 1). When receiving the use request, the management device 1 authenticates whether the user can use the learning model. When the authentication is successful, the management device 1 executes a task using the learning model and outputs the execution result to the user terminal 3 ((4) in FIG. 1). The management device 1 calculates a consideration for using the learning model, and charges the user of the user terminal 3 for payment of the consideration. Thereafter, the management device 1 calculates a reward to be paid to each contributor based on the consideration paid by the user and the contribution rate associated with the learning model ((5) in FIG. 1). By doing in this way, the management apparatus 1 can pay a reward appropriately to a contributor according to the learning model being used.
Hereinafter, the configuration of the management apparatus 1 will be described.

[Configuration Example of Management Device 1]
FIG. 2 is a diagram illustrating a configuration of the management apparatus 1 according to the present embodiment.
The management device 1 includes a storage unit 11 and a control unit 12.

  The storage unit 11 is, for example, a ROM (Read Only Memory) and a RAM (Random Access Memory). The storage unit 11 stores various programs for causing the management device 1 to function. For example, the storage unit 11 controls the control unit 12 of the management apparatus 1 into a model registration unit 13, a license registration unit 14, a use reception unit 15, an authentication unit 16, a learning execution unit 17, a consideration calculation unit 18, and a reward calculation which will be described later. A management program that functions as the unit 19 is stored. In addition, the storage unit 11 stores learning model information including a learning model and learning model license information.

  FIG. 3 is a diagram illustrating an example of learning model information according to the present embodiment. As illustrated in FIG. 3, the learning model information includes a registrant, a model name, an explanatory note, a learning model, a model ID, a post-quantization data string, a contributor, and a contribution rate.

  The registrant indicates the contributor who registered the learning model, and stores a contributor ID for identifying the contributor. The model name is information for identifying the model. In the present embodiment, the model name is different for each of all learning models.

  The explanatory text is a character string for explaining the contents of the model. The model ID is information for identifying the learning model. The post-quantization data string is a hash value after the model ID is quantized. The contributor indicates the contributor who contributed to the creation of the learning model, and the contributor ID is stored. Here, one or a plurality of contributors and contribution rates are associated with one learning model.

  FIG. 4 is a diagram showing an example of license information according to the present embodiment. As illustrated in FIG. 4, the license information includes a user ID of a user who uses the learning model, a model name, and a model ID.

  Returning to FIG. 2, the controller 12 is, for example, a CPU (Central Processing Unit). The control unit 12 controls functions related to the management device 1 by executing various programs stored in the storage unit 11. The control unit 12 functions as a model registration unit 13, a license registration unit 14, a use reception unit 15, an authentication unit 16, a learning execution unit 17, a price calculation unit 18, and a reward calculation unit 19 by executing a management program. .

[Register and update learning model]
The model registration unit 13 receives a new registration or update of a learning model from the contributor terminal 2 used by the contributor. FIG. 5 is a diagram illustrating a configuration of the model registration unit 13 according to the present embodiment. As illustrated in FIG. 5, the model registration unit 13 includes an acquisition unit 131, a division unit 132, a first hash generation unit 133, a quantization unit 134, a second hash generation unit 135, and a contribution rate management unit 136. And a storage control unit 137.

  The acquisition unit 131 acquires a learning model from the contributor terminal 2. Specifically, when accepting new registration of a learning model, the acquisition unit 131 receives a learning model, a learning model name, a description of the learning model, a learning model label, and a learning model from the contributor terminal 2. Get the contributor ID of the contributor who created Here, the learning model includes an index assigned to each of the plurality of layers and a plurality of weight values indicating the weights of the plurality of layers.

  Moreover, the acquisition part 131 acquires contributor ID from the contributor terminal 2 first, when receiving the update of a learning model. The acquisition unit 131 refers to the learning model information stored in the storage unit 11, identifies the model name associated with the contributor ID, and accepts selection of any one of the model names. Then, the acquisition unit 131 acquires the updated learning model from the contributor terminal 2. The acquisition unit 131 may acquire the label of the learning model even when the learning model is updated.

  The dividing unit 132 divides the learning model into one or a plurality of subsets. The dividing unit 132 divides a plurality of layers, and sets a weight corresponding to each layer as a subset. Here, the dividing unit 132 generates a subset including the label of the learning model acquired by the acquiring unit 131 together with the learning model.

  The first hash generation unit 133 generates a model ID for identifying a learning model by generating a hash value for each of the one or more subsets generated by the dividing unit 132. Specifically, first, the first hash generation unit 133 generates a hash value for each of a plurality of weight values by applying a hash function to each of a plurality of weight values indicating one or a plurality of subsets. To do. And the 1st hash production | generation part 133 connects the hash value produced | generated with respect to each of several weight value, for example in the order of the index provided to each layer, By this, the data sequence which uses these hash values as an element As a model ID.

  The quantization unit 134 quantizes each of the subsets divided by the dividing unit 132, which is the weight of each layer of the learning model. For example, the quantization unit 134 divides the range that the weight value can take into a plurality of ranges, and sets a representative value for each of the plurality of ranges. The quantization unit 134 quantizes the subset by converting the plurality of weight values into representative values corresponding to ranges to which the weight values belong. By doing in this way, the management apparatus 1 can reduce a processing load, without changing a contribution rate, when the detailed update of a learning model generate | occur | produces.

  The second hash generation unit 135 generates a post-quantization data string having the hash value of the weight of each layer of the learning model as an element. Specifically, first, the second hash generation unit 135 applies a hash function to each of a plurality of weight values indicating the subset quantized by the quantization unit 134, thereby each of the plurality of weight values. Calculate a hash value. Then, the second hash generation unit 135 concatenates the hash values generated for each of the plurality of weight values in the order of the indexes assigned to the respective layers, thereby quantizing the data having these hash values as elements. Generate a column.

  The contribution rate management unit 136 calculates the contribution rate of each contributor related to the creation of the learning model acquired by the acquisition unit 131. When newly registering a learning model, the contribution rate management unit 136 refers to the learning model information stored in the storage unit 11, adds the model ID generated by the first hash generation unit 133 to the learning model information. It is determined whether or not the stored model ID of another learning model is included.

  If the model ID of another learning model is not included, the contribution rate management unit 136 determines that the newly registered learning model is created independently by the contributor who performs registration, and determines the contribution rate of the contributor. Calculated as 1 (100%).

  When the model ID of another learning model is included, the contribution rate management unit 136 refers to the other learning model included in the learning model information stored in the storage unit 11, and the other Identify the number of layers in the learning model. Then, the contribution rate management unit 136 subtracts the number of layers of the other learning model from the total number of layers of the newly registered learning model, and specifies the number of layers generated by the contributor that performs registration.

  Further, the contribution rate management unit 136 refers to the learning model information stored in the storage unit 11, and based on the contributors and contribution rates of other learning models and the number of layers of the other learning models, The number of layers generated by each contributor who generated other learning models is specified. Then, the contribution rate management unit 136 calculates the contribution rate based on the ratio of the number of layers generated by each contributor to the total number of layers of the newly registered learning model.

  In addition, the contribution rate management unit 136 updates the contribution rate when the learning model is updated. Specifically, the contribution rate management unit 136 refers to the learning model information stored in the storage unit 11 and identifies the quantized data string of the learning model to be updated. Then, the contribution rate management unit 136 determines that the specified post-quantization data sequence is different from the post-quantization data sequence generated by the second hash generation unit 135, that is, the post-quantization data sequence has been changed. Then, it is determined that the learning model has been updated, and the contribution rate is calculated. The calculation method of the contribution rate when the learning model is updated is the same as the calculation method of the contribution rate when the learning model is newly registered, and thus description thereof is omitted.

  The contribution rate management unit 136 may calculate the contribution rate based on the amount of data used for learning of the neural network in order for each contributor to generate a learning model. In this case, the acquisition unit 131 acquires information indicating the amount of data used for learning of the neural network to generate the learning model when the learning model is registered, and learning model information stored in the storage unit 11 In addition, information indicating the data amount is stored.

  Further, the contribution rate management unit 136 may calculate the contribution rate based on the total cost of the calculation resources used for learning. In this case, the acquisition unit 131 acquires calculation resource information indicating the amount of calculation resources consumed for learning of the neural network in order to generate the learning model when the learning model is registered, and is stored in the storage unit 11. The computing resource information is stored in the learning model information. Here, the consumption resource information includes information indicating the CPU usage time and information indicating the usage area (memory usage) of the storage unit 11.

  Moreover, in order to calculate a contribution rate based on a calculation resource, the management apparatus 1 may include a model generation unit that generates a learning model by performing learning of a neural network. Then, the model generation unit receives data used for learning the neural network from the contributor terminal 2 and learns the neural network. In addition, the model generation unit stores the contributor ID, the model ID of the generated learning model, and consumption resource information indicating the calculation resource consumed in learning in association with each other. The contribution rate management unit 136 specifies the ratio of the calculation resource used by each contributor to generate the learning model based on the calculation resource indicated by the consumption resource information stored in the storage unit 11, and contributes based on the ratio. Calculate the rate.

  The contribution rate management unit 136 refers to the learning model information and identifies information indicating the data amount corresponding to the learning model and information indicating the data amount corresponding to another learning model included in the learning model. . And the contribution rate management part 136 calculates a contribution rate based on the ratio of the data amount corresponding to a learning model, and the data amount corresponding to another learning model. In the generation of a learning model, the greater the amount of data used for learning, the higher the learning effect and the more useful the learning model. Therefore, the management apparatus 1 can set the contribution rate of the contributor who contributed to the improvement of the usefulness of a learning model by calculating a contribution rate based on the data amount used for learning.

  The storage control unit 137 stores the learning model in the storage unit 11. The storage control unit 137 identifies the contributor ID acquired by the acquisition unit 131 as a registrant. When the storage control unit 137 receives a new registration of a learning model, the storage controller 137 includes the identified registrant, the learning model acquired by the acquisition unit 131, the learning model name, the description, and the label, and the first hash generation unit 133. The generated model ID, the post-quantization data string generated by the second hash generation unit 135, and the contributor and contribution rate specified by the contribution rate management unit 136 are associated and stored as learning model information in the storage unit 11. .

  In addition, when the storage control unit 137 receives an update of the learning model, if the contribution rate management unit 136 determines that the learning model has been updated based on the post-quantization data sequence, the storage control unit 136 stores the learning stored in the learning model information. The model, model ID, post-quantization data string, contributor, and contribution rate are updated. The storage control unit 137 includes the learning model acquired by the acquisition unit 131, the model ID generated by the first hash generation unit 133, the quantized data sequence generated by the second hash generation unit 135, and the contribution rate management unit 136. The learning model information is updated based on the contributor and the contribution rate identified by.

  Further, when the storage control unit 137 receives an update of the learning model, if the contribution rate management unit 136 determines that the learning model has not been updated based on the post-quantization data sequence, the storage control unit 137 stores the learning model information. The learning model and the model ID are updated to the learning model acquired by the acquisition unit 131 and the model ID generated by the first hash generation unit 133.

[Register License Information]
Returning to FIG. The license registration unit 14 receives license registration from the user of the user terminal 3. For example, the license registration unit 14 displays a model name, a model ID, and an explanatory text on the user terminal 3, and accepts selection of a learning model used by the user. Here, for example, the license registration unit 14 may calculate the license fee of the learning model based on the total number of layers and the label of the learning model, and display the license fee on the user terminal 3.

  When the learning model is selected, the user terminal 3 transmits a license registration request including a model name, a model ID, and a user ID for identifying the user to the management apparatus 1. Upon receiving the license registration request, the license registration unit 14 associates the user ID, model name, and model ID included in the license registration request with each other and stores them in the storage unit 11 as license information.

[Use learning model]
The use accepting unit 15 accepts a learning model use request from the user terminal 3 as another device via the communication network. Specifically, the use accepting unit 15 accepts a learning model use request from the user terminal 3 by receiving a learning model use request including a model name of the learning model used by the user and a user ID. . In the present embodiment, the use accepting unit 15 accepts a learning model use request from the user terminal 3, but the present invention is not limited thereto. The use receiving unit 15 may receive a use request from the contributor terminal 2.

  When the use accepting unit 15 accepts a learning model use request from the user terminal 3, the authentication unit 16 refers to the license information stored in the storage unit 11, and the user of the user terminal 3 uses the learning model. Determine whether it is possible.

  Specifically, the authentication unit 16 determines whether the user of the user terminal 3 has the model name of the learning model included in the learning model use request and the user ID associated with each other and stored in the license information. Then, it is determined that the learning model can be used.

  In addition, when the model name of the learning model included in the learning model use request and the user ID are associated with each other and stored in the license information, the authentication unit 16 allows the user of the user terminal 3 to perform the learning Determine that the model is unusable. If the authentication unit 16 determines that the learning model is unusable, the authentication unit 16 notifies the user terminal 3 of a message indicating that the learning model is unusable.

  When the user can use the learning model, the learning execution unit 17 outputs an execution result of the learning model related to data for executing the task acquired from the user terminal 3. Specifically, first, when the authentication unit 16 determines that the user of the user terminal 3 can use the learning model, the learning execution unit 17 acquires one or a plurality of data for executing the task from the user terminal 3. . The learning execution unit 17 uses the data acquired from the user terminal 3 as input data, and acquires an execution result by executing a task using a learning model corresponding to the use request. The learning execution unit 17 outputs the acquired execution result to the user terminal 3.

[Calculation of consideration for use of learning model and reward for each contributor]
The consideration calculation unit 18 calculates the amount billed to the user based on the usage mode of the learning execution unit 17 for executing the task. For example, the consideration calculation unit 18 calculates the amount charged to the user based on the usage load of the learning execution unit 17 for executing the task. Here, the usage load of the learning execution unit 17 is the time spent by the learning execution unit 17 when the task based on the learning model used by the user is executed, that is, the CPU usage time. By doing in this way, the management apparatus 1 can calculate a charge amount fairly based on a viewpoint of utilization load.

  The consideration calculation unit 18 may calculate the amount charged to the user based on the occupation rate of the learning execution unit 17 for executing the task. Here, the occupation rate of the learning execution unit 17 is the CPU utilization rate of the learning execution unit 17 when a task based on the learning model used by the user is executed. By doing in this way, the management apparatus 1 can calculate a charge amount fairly based on CPU utilization, when a some user uses.

  The consideration calculation unit 18 notifies the calculated charge amount to the user terminal 3. The user of the user terminal 3 pays the notified bill amount to the administrator of the management apparatus 1 by a predetermined method.

  The reward calculation unit 19 calculates each reward when the consideration paid in return for the use of the learning model by the user terminal 3 is distributed to each contributor according to the contribution rate related to creation of the learning model.

  Specifically, first, the reward calculation unit 19 refers to the learning model information stored in the storage unit 11 and specifies the contribution rate of each contributor of the learning model for which the consideration has been calculated. And the reward calculation part 19 calculates the reward of each contributor by multiplying the consideration which the consideration calculation part 18 calculated, and the contribution rate of each identified contributor.

  Here, the remuneration calculation unit 19 may calculate the remuneration of the manager of the management device 1 and subtract the remuneration of the administrator from the consideration calculated by the consideration calculation unit 18. And the reward calculation part 19 may calculate the reward of each contributor by multiplying the consideration after subtraction, and the contribution rate of each specified contributor. The reward calculation unit 19 pays the calculated charge amount to each contributor by a predetermined method.

[Flow of processing related to learning model registration]
Next, the flow of processing in the management apparatus 1 will be described. First, the flow of processing related to registration or update of a learning model will be described. FIG. 6 is a flowchart showing a flow of processing related to registration or update of a learning model according to the present embodiment.

First, the acquisition unit 131 acquires a learning model (S10).
Subsequently, the dividing unit 132 divides the learning model to generate a subset (S20).

Subsequently, the first hash generation unit 133 generates a hash value of the subset generated in S20, and generates a model ID of the learning model based on the hash value (S30).
Subsequently, the quantization unit 134 quantizes the subset (S40). Then, the second hash generation unit 135 generates a hash value of the quantized subset, and generates a quantized data string based on the hash value (S50).

Subsequently, the contribution rate management unit 136 calculates the contribution rate of each contributor related to creation of the learning model (S60).
Subsequently, the storage control unit 137 updates the learning model information by registering or updating the learning model (S70).

[Flow of processing related to learning model usage and reward calculation]
Next, the flow of processing related to use of the learning model and calculation of reward will be described. FIG. 7 is a flowchart showing a flow of processing relating to use of the learning model and reward calculation according to the present embodiment.

First, the use receiving unit 15 receives a learning model use request (S110). Here, the learning model use request includes the model name of the learning model and the user ID.
Subsequently, the authentication unit 16 authenticates whether or not the learning model for which the use request has been accepted is usable (S120).

Subsequently, the learning execution unit 17 acquires input data for executing a task using the learning model in response to successful authentication by the authentication unit 16 (S130).
Subsequently, the learning execution unit 17 executes a task using the acquired input data and the learning model corresponding to the use request (S140). Thereby, the learning execution part 17 acquires the execution result of a task.

Subsequently, the consideration calculating unit 18 calculates a billing amount related to the use of the learning model based on the usage mode of the learning model (S150).
Subsequently, the reward calculating unit 19 calculates a reward for each contributor based on the contribution rate stored in the learning model information corresponding to the used learning model (S160).

[Effect in this embodiment]
As described above, the management device 1 according to the present embodiment stores the learning model created by one or more contributors and the contribution rate related to the creation of the learning model of each contributor in association with each other, and the user terminal 3 via the network. A learning model use request is received from the user, and each reward is calculated when the consideration paid in return for the use of the learning model by the user terminal 3 is distributed to each contributor according to the contribution rate corresponding to the learning model.

  By doing in this way, the management apparatus 1 can pay a reward appropriately with respect to the contributor who contributed to preparation of a learning model. In particular, when a learning model is created by a plurality of contributors, the reward of each contributor can be calculated fairly based on the contribution rate related to creation of the learning model of each contributor.

  Further, the management device 1 according to the present embodiment divides the learning model into one or a plurality of subsets, generates a hash value of each subset, generates a model ID for specifying the learning model, and generates the generated model ID And license information that associates the learning model with the user ID that can use the learning model. By doing in this way, the model ID corresponding to the other learning model is included in the learning model created using at least a part of the other learning model. Therefore, the management apparatus 1 can manage whether the learning model is created using another learning model, and can identify one or a plurality of contributors who created the learning model.

  As mentioned above, although this invention was demonstrated using embodiment, the technical scope of this invention is not limited to the range as described in the said embodiment. It will be apparent to those skilled in the art that various modifications or improvements can be added to the above embodiment. In particular, the specific embodiments of the distribution / integration of the devices are not limited to those illustrated above, and all or a part thereof may be arbitrarily selected according to various additions or according to the functional load. It can be configured to be functionally or physically distributed and integrated in units.

  For example, in the above-described embodiment, the management apparatus 1 accesses each of the plurality of communication devices 4 and acquires the logs generated by each of the plurality of communication devices 4, but is not limited thereto. For example, a log acquisition device that acquires a log generated by each of a plurality of communication devices 4 and a management device 1 are provided, and the management device 1 generates each of the plurality of communication devices 4 from the log acquisition device. A log may be acquired.

  For example, in the above-described embodiment, the case where the contribution rate management unit 136 calculates the contribution rate based on the ratio of the number of layers generated by each contributor to the total number of layers of the newly registered learning model has been described. Instead, the contribution rate management unit 136 may calculate the contribution rate based on the ratio of the weights included in the layers generated by each contributor to the total data amount of the weights included in each layer of the newly registered learning model. Good. Since the amount of weight data included in each layer may differ depending on the layer, the amount of weight data actually generated by each contributor can be reflected in the contribution rate.

DESCRIPTION OF SYMBOLS 1 ... Management apparatus, 11 ... Memory | storage part, 12 ... Control part, 13 ... Model registration part, 131 ... Acquisition part, 132 ... Dividing part, 133 ... 1st hash Generation unit 134... Quantization unit 135 135 Second hash generation unit 136. Contribution rate management unit 137 Storage control unit 14 License registration unit 15 Use accepting unit, 16 ... authentication unit, 17 ... learning execution unit, 18 ... consideration calculation unit, 19 ... reward calculation unit, 2 ... contributor terminal, 3 ... user terminal

Claims (8)

A storage unit that associates and holds a learning model created by one or more contributors and a contribution rate related to creation of the learning model of each contributor;
A receiving unit that receives a request to use the learning model from another device via a network;
A reward calculation unit for calculating respective rewards when the consideration paid in return for the use of the learning model by the other device is distributed to each contributor according to the contribution rate;
A management device comprising: The learning model is a weight of each of a plurality of layers generated by learning using a neural network,
A contribution rate management unit that calculates the contribution rate based on a ratio of the number of layers generated by each contributor to the total number of layers;
The management apparatus according to claim 1. The contribution rate management unit calculates the contribution rate based on the amount of data used by the contributor to learn the neural network in order to generate the learning model.
The management device according to claim 2. The contribution rate management unit updates the contribution rate when the learning model is updated.
The management device according to claim 2 or 3. A hash generation unit that generates a data string having a hash value of a weight of each layer of the learning model as an element;
The contribution rate management unit updates the contribution rate when the data string is changed.
The management apparatus as described in any one of Claim 2 to 4. A quantization unit for quantizing the weight of each layer;
The hash generation unit calculates a hash value of the weight quantized by the quantization unit;
The management device according to claim 5. Processor
Reading the contribution rate data from a storage unit holding contribution rate data associated with a learning model created by one or more contributors and a contribution rate related to creation of the learning model of each contributor;
Receiving a request to use the learning model from another device via a network;
Calculating each reward when distributing the consideration paid in return for the use of the learning model by the other device to each contributor according to the contribution rate;
Management method to execute. On the computer,
A function of reading the contribution rate data from a storage unit that holds contribution rate data associated with a learning model created by one or more contributors and a contribution rate related to creation of the learning model of each contributor;
A function of accepting a request to use the learning model from another device via a network;
A function of calculating each reward when distributing the consideration paid in return for the use of the learning model by the other device to each contributor according to the contribution rate;
A program that realizes

Images