Machine learning system, machine learning device, machine learning method, and program
The system optimizes data flow and model transfer paths within a network to address federated learning accuracy issues, achieving faster and more accurate deep learning model training across diverse datasets.
Patent Information
- Application Number
- JP2023104115
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-06-26
- Publication Date
- 2025-01-15
AI Technical Summary
Federated learning methods face challenges in achieving high accuracy due to variations in dataset characteristics across devices, leading to inconsistent convergence.
A machine learning system that optimizes data flow rates and model parameter transfer paths within a network to minimize training time and enhance accuracy by using statistical information and processing capacity of each server, thereby improving convergence during federated learning.
The system enables the creation of highly accurate deep learning models by optimizing data flow and model transfer paths, reducing training time and enhancing model accuracy.
Smart Images
Figure 2025004429000001_ABST
Abstract
Description
[Technical field]
[0001] The present disclosure relates to a machine learning system, a machine learning device, a machine learning method, and a program. [Background technology]
[0002] In recent years, a method called federated learning has been proposed as a method for learning a distributed deep learning model (for example, Non-Patent Document 1). In federated learning, common model parameters are distributed to devices capable of learning a deep learning model, and each device learns individually using only its own learning data, while periodically synchronizing the model parameters learned by each device to progress the learning. [Prior art documents] [Non-patent literature]
[0003] [Non-Patent Document 1] WYB Lim et al.,"Federated learning in mobile edge networks: A comprehensive survey" IEEE Communications Surveys & Tutorials, 2020. Summary of the Invention [Problem to be solved by the invention]
[0004] However, federated learning has the problem that its convergence varies depending on the characteristics of the dataset held by the device that trains the deep learning model, and it may not be possible to obtain a deep learning model with high accuracy.
[0005] The present disclosure has been made in consideration of the above points, and provides a technology that can obtain a highly accurate deep learning model through federated learning. [Means for solving the problem]
[0006] A machine learning system according to one embodiment of the present disclosure is a machine learning system that learns a machine learning model through federated learning, and includes a first receiving unit that receives statistical information about data generated by each router included in a target network, a second receiving unit that receives the processing capacity of each server included in the target network and a local learning time that represents the time required to learn the machine learning model on each server, and an optimization unit that uses the statistical information, the processing capacity, and the local learning time to optimize a parameter that represents the ratio of the amount of data held by the server and the ratio of the product of the data set distance between the data set held by the server and a data set that virtually combines all data into one, so as to minimize the sum, for the server, of the product of the data set distance between the data set held by the server and a data set that virtually combines all data into one. Effect of the Invention
[0007] A technology is provided that can obtain a highly accurate deep learning model through federated learning. [Brief description of the drawings]
[0008] [Figure 1] FIG. 2 is a diagram illustrating an example of a configuration of a target network. [Diagram 2] FIG. 2 is a diagram illustrating an example of a hardware configuration of a device according to the present embodiment. [Diagram 3] FIG. 2 is a diagram illustrating an example of a functional configuration of a calculation function unit according to the present embodiment. [Figure 4] FIG. 2 is a diagram illustrating an example of a functional configuration of a router function unit according to the present embodiment. [Diagram 5] FIG. 2 is a diagram illustrating an example of a functional configuration of a server function unit according to the embodiment. [Figure 6] FIG. 2 is a diagram illustrating an example of a functional configuration of a central server function unit according to the present embodiment. [Figure 7] 13 is a flowchart illustrating an example of a parameter and model transfer path determination process. [Figure 8] 1 is a flowchart illustrating an example of a learning process of a deep learning model. [Figure 9] 1 is a flowchart showing an example of an inference process using a deep learning model. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0009] Hereinafter, one embodiment of the present invention will be described.
[0010] <Target network> In the following embodiment, as an example, a network 1 (hereinafter also referred to as a target network 1) shown in Fig. 1 is taken as a target. The target network 1 shown in Fig. 1 includes a plurality of devices 10. The devices 10 are, for example, various computers or computer systems such as a general-purpose server or a network device.
[0011] One of the devices 10 has a calculation function unit 110 that solves an optimization problem described below and determines optimal parameters. The remaining devices 10 have a router function unit 120 that collects and transfers data. Furthermore, each of the devices 10 except for one of the devices 10 that has the router function unit 120 has a server function unit 130 that manages a data set for learning the deep learning model and performs learning (hereinafter also referred to as local learning) and inference of the deep learning model. Meanwhile, the one device 10 has a central server function unit 140 that distributes (model transfers) and synchronizes model parameters of the deep learning model. The deep learning model may be called a machine learning model or the like.
[0012] Hereinafter, it is assumed that each router function unit 120 is given a number in sequence starting from 1, and the i-th router function unit 120 will also be referred to as the "router i." Similarly, it is assumed that each server function unit 130 is given a number in sequence starting from 1, and the j-th server function unit 130 will also be referred to as the "server j." Additionally, the central server function unit 140 will also be referred to as the "central server."
[0013] Furthermore, when considering the target network 1 as a graph, each router function unit 120, each server function unit 130, and the central server function unit 140 are regarded as nodes, and physical or logical communication paths connecting the nodes are regarded as links. However, in the following, for simplicity, it is assumed that the bandwidth of the link connecting the router function unit 120 and the server function unit 130 or the central server function unit 140 in the same device 10 is infinite (or the maximum value that can be taken in the implementation).
[0014] The device 10 having the calculation function unit 110 does not necessarily need to be located within the target network 1, and may be located outside the target network 1, for example. Also, for example, the calculation function unit 110 may be included in any one of the devices 10 having the server function unit 130 or the central server function unit 140. Also, for example, the central server function unit 140 may further have the same function as the server function unit 130.
[0015] <Formulation of optimization problem> Below, we will explain a formulation for minimizing the time required to learn a deep learning model and obtaining a highly accurate deep learning model when performing federated learning in each server function unit 130.
[0016] <Definition of symbols> First, the symbols used hereinafter will be defined.
[0017] λ ij (y): The amount of data with label y generated by router i out of the amount of data collected per unit time by server j (Mbps) λ i (y): The amount of data with label y generated per unit time at router i (Mbps) λ j (y): The amount of data with label y received by server j per unit time (Mbps) λ(y)=Σ i λ i(y): The amount of data labeled y generated per unit time in the entire target network 1 (Mbps) ·c ab : The link bandwidth (Mbps) for traffic flowing from a to b between adjacent routers a and b (hereafter, the bandwidth vector is defined as C={c ab}) R=r i [ab]∈P: The ratio of data flowing from router i to link ab in the direction from a to b (data ratio) · L: A collection of links L j : the set of routes between the central server and server j l j : The set of links on a path between the central server and server j N a : A set of nodes directly connected to node a N: A set of router function units 120 N S : A set of server function units 130 ε: Tolerance ·v ab : Traffic volume due to data transfer flowing through link ab [Mbps] ·n ab : The number of scr-dst pairs (i.e., source-destination pairs) passing through link ab in model transfer A ab : Available bandwidth of link ab when transferring model M: Capacity of model parameters for deep learning model [Mbyte] · τ: Data aggregation interval for learning [s] T j : Time required for one learning cycle on server j [s] T j c : Local learning time of server j [s] T j t : Model transfer time of server j [s] p j : Processing power of server j [FLOPS] m: Size of training data (i.e., size of one record in the training dataset) However, when considering all labels, λ ij (y), λ i (y), λ j (y), λ(y) and λ ij , λ i , λ j , denoted as λ.
[0018] Note that a label is a value used as training data when learning a deep learning model, such as a binary label that indicates whether the data belongs to a certain class or not. P is the data ratio r i The set of all [ab]. Data ratio r i Since [ab] is the parameter to be optimized, r i [ab] will also be called a parameter.
[0019] <<Method of determining parameters and forwarding route>> Below, the optimal parameter r i [ab] is determined, and for each server j, L j This section explains a method for determining the optimal route (model transfer route) from among the above. This optimizes the data transfer route from each router i to each server j and the model transfer route from each server j to the central server, thereby minimizing the time required for learning the deep learning model and acquiring a highly accurate deep learning model.
[0020] According to references 1 and 2, the final model parameter vector w f and the model parameter vector w when the deep learning model is trained in one place. c The difference between is expressed by the following formula (1).
[0021]
number
[0022] The first is (i) the number of data of each label held by each server j and the number of data of each label per server when all data is distributed to each server j on average, and the Kullback-Leibler divergence (KL-div) of these two values approximated by a categorical distribution is expressed as the distance between the data sets ||Dis j Let's say ||.
[0023] The second is (ii) the approximation of the Kullback-Leibler divergence (KL-div) for each label between the data distribution of the dataset held by each server j and the data distribution of all the data collected virtually under the assumption that the data of each label follows a Gaussian distribution, which is called the inter-dataset distance ||Dis j Let's say ||.
[0024] First, the distance between the data sets ||Dis j The case where || is defined as (i) will be described. In this case, formula (1) is expressed as formula (2) below.
[0025]
number
[0026] Next, the distance between the data sets ||Dis jWe will explain the case where || is defined as (ii). In this case, the average μ i (y) and variance Σ i After calculating (y), we assume a Gaussian distribution based on these. Then, we express the distribution of data collected on server j and the distribution of data when all data is collected on one server as a mixed normal distribution, and approximate the KL-div between these mixed normal distributions using variational approximation, and then calculate r i Only the terms that depend on [ab] are extracted. As a result, although we will not go into the detailed derivation process, equation (1) can be expressed as the following equation (3).
[0027]
number
[0028] Since the above equations (2) and (3) represent the convergence limit of the model parameters by federated learning, the constraint is that they must be smaller than the allowable error ε. Under this constraint, we consider minimizing the total time required for one cycle of federated learning, specifically, the downloading of model parameters from the central server, model learning on each server j, and uploading of model parameters from each server j to the central server.
[0029] First, the transfer time T j t Assuming that the traffic on each link depends only on the data transfer from router i and the uploading and downloading of model parameters from each server j, the capacity of each link available for model transfer can be expressed as follows:
[0030]
number
[0031] In addition, the route between the central server and each server j is determined so as to satisfy the following formula (4).
[0032]
number
[0033] Next, the model training time (local training time) on each server j is T j c =αmτλ j / p j+β. This indicates that the model learning time depends on the amount of data collected by server j and the processing power of server j. α and β are constants, and their values are determined in advance by linear approximation through local learning. Therefore, the time required for one cycle of learning in federated learning on server j is T j can be expressed by the following equations (5) and (6).
[0034]
number
[0035] In this embodiment, the optimization target is r i [ab] and r i [ab] and λ i Using (y), λ ij (y) and λ j (y) can be expressed as follows:
[0036]
number
[0037] Therefore, taking into consideration constraints such as congestion avoidance requests and flow rate conservation in the target network 1, the solution is reduced to solving the constrained optimization problems shown in the following equations (7) to (15).
[0038]
number
[0039] The constrained optimization problem shown in the above equations (7) to (15) is expressed as (A)r i The solution is in two steps: (A) determine the model transfer time T j t Only for this, r is set so that this is minimized. i [ab] is found. At this time, a conditional nonlinear optimization problem is solved, so it is solved by using a solver for the nonlinear optimization problem. (B) After that, r i The capacity available for model transfer on each link by determining [ab] is ^c ab is determined, and the available bandwidth A during model transfer is determined based on this information. ab Calculate the model transfer path ^l so that equation (6) is minimized. j Determine.
[0040] <Example of hardware configuration for device 10> An example of the hardware configuration of device 10 according to this embodiment is shown in Fig. 2. As shown in Fig. 2, device 10 according to this embodiment has an input device 201, a display device 202, an external I / F 203, a communication I / F 204, a RAM (Random Access Memory) 205, a ROM (Read Only Memory) 206, an auxiliary storage device 207, and a processor 208. Each of these pieces of hardware is connected to each other via a bus 209 so as to be able to communicate with each other.
[0041] The input device 201 is, for example, a keyboard, a mouse, a touch panel, a physical button, etc. The display device 202 is, for example, a display, a display panel, etc. Note that the device 10 does not necessarily have to have at least one of the input device 201 and the display device 202, for example.
[0042] The external I / F 203 is an interface with an external device such as a recording medium 203a. Examples of the recording medium 203a include a CD (Compact Disc), a DVD (Digital Versatile Disk), an SD memory card (Secure Digital memory card), and a USB (Universal Serial Bus) memory card.
[0043] The communication I / F 204 is an interface for the device 10 to transmit and receive data to and from other devices 10. The RAM 205 is a volatile semiconductor memory (storage device) that temporarily stores programs and data. The ROM 206 is a non-volatile semiconductor memory (storage device) that can store programs and data even when the power is turned off. The auxiliary storage device 207 is a non-volatile storage device such as a hard disk drive (HDD), a solid state drive (SSD), or a flash memory. The processor 208 is various types of arithmetic devices such as a central processing unit (CPU).
[0044] 2 is merely an example, and the hardware configuration of the device 10 is not limited thereto. The device 10 may have, for example, multiple auxiliary storage devices 207 and multiple processors 208, may not have some of the hardware shown in the figure, or may have various hardware other than the hardware shown in the figure.
[0045] <Example of functional configuration of the calculation function unit 110, the router function unit 120, the server function unit 130, and the central server function unit 140> An example of the functional configuration of the calculation function unit 110, the router function unit 120, the server function unit 130, and the central server function unit 140 will be described below.
[0046] ≪Calculation function section 110≫ An example of the functional configuration of the calculation function unit 110 according to this embodiment is shown in Fig. 3. As shown in Fig. 3, the calculation function unit 110 according to this embodiment includes an optimization unit 111 and a transmission unit 112. Note that the calculation function unit 110 is realized, for example, by a process in which one or more programs installed in the device 10 are executed by the processor 208 or the like.
[0047] The optimization unit 111 optimizes the information received from each router i (the amount of data generated for each label, λ i (y), Distance between data sets ||Dis j If || is defined as (ii), then the average μ i (y) and variance Σ i (y)) and the information received from each server j (processing capacity p j and the local training time T j c (D)) to solve the constrained optimization problem shown in Eqs. (7) to (15) to obtain the optimal parameter r i [ab] and the optimal model transfer path^l j It is decided that:
[0048] The transmission unit 112 transmits the optimal parameter r determined by the optimization unit 111. i [ab] and the optimal model transfer path^l j and are sent to each router i.
[0049] <Router function unit 120> An example of the functional configuration of the router function unit 120 according to this embodiment is shown in Fig. 4. As shown in Fig. 4, the router function unit 120 according to this embodiment includes a collection unit 121, a transfer unit 122, and a statistics unit 123. Note that the router function unit 120 is realized, for example, by a process in which one or more programs installed in the device 10 are executed by the processor 208 or the like.
[0050] The collection unit 121 collects and generates data from a predetermined target (for example, a memory, a CPU, etc.). In addition, the collection unit 121 assigns labels to the data before learning of a deep learning model. Note that the labels may be assigned manually, for example.
[0051] The transfer unit 122 transfers (sends) the data collected and generated by the collection unit 121 to the corresponding server j. At this time, if a label is attached to the data, the transfer unit 122 also transfers (sends) the label together with the data. The transfer unit 122 also transfers (sends) statistical information calculated by the statistics unit 123 to the calculation function unit 110. Furthermore, the transfer unit 122 transfers (sends) model parameters downloaded and uploaded between the central server function unit 140 and each server j via the model transfer route ^l j Transfer according to.
[0052] The statistical unit 123 calculates statistical information on the data collected and generated by the collection unit 121. The statistical unit 123 calculates the amount of data generated for each label, λ i (y), Distance between data sets ||Dis j If || is defined as (ii), then the average μ i (y) and variance Σ i (y) is calculated as statistical information.
[0053] <Server function unit 130> An example of the functional configuration of the server function unit 130 according to this embodiment is shown in Fig. 5. As shown in Fig. 5, the server function unit 130 according to this embodiment includes a database management unit 131, a learning unit 132, and an inference unit 133. Note that the server function unit 130 is realized, for example, by a process in which one or more programs installed in the device 10 are executed by the processor 208 or the like.
[0054] The database management unit 131 stores the data and its label transferred from each router i in a database as learning data. The database is realized by a storage area such as the auxiliary storage device 207 of the device 10 having the server function unit 130.
[0055] The learning unit 132 learns (trains) a deep learning model through associative learning using a data set of learning data stored in a database.
[0056] In addition, the learning unit 132 performs learning when the amount of data in the data set is changed, thereby obtaining a local learning time T j c Then, the learning unit 132 calculates its own processing capacity p j and each local learning time T j c (D) is transferred (sent) to the calculation function unit 110.
[0057] The inference unit 133 performs inference on the data transferred from each router i using the trained deep learning model.
[0058] <<Central Server Function Unit 140>> An example of the functional configuration of the central server function unit 140 according to this embodiment is shown in Fig. 6. As shown in Fig. 6, the central server function unit 140 according to this embodiment includes a model transmission unit 141 and a model integration unit 142. Note that the central server function unit 140 is realized, for example, by a process in which one or more programs installed in the device 10 are executed by the processor 208 or the like.
[0059] The model transmission unit 141 transmits model parameters of the deep learning model to each server j via the transfer unit 122 of the router function unit 120 in the same device 10. The model parameters are stored in a storage area such as the auxiliary storage device 207 of the device 10 having the central server function unit 140.
[0060] When each server j receives the model parameters, the model integration unit 142 synchronizes (integrates) and stores these model parameters.
[0061] Note that the central server function unit 140 may include, for example, a database management unit 131, a learning unit 132, and an inference unit 133, similar to the server function unit .
[0062] <Processing details> Below, the optimal parameter r i [ab] and the optimal model transfer path^l j This section describes the process of determining the optimality of a deep learning model, the process of learning a deep learning model through federated learning, and the inference process using the trained deep learning model.
[0063] <Parameter and model transfer route determination process> Optimal parameter r i [ab] and the optimal model transfer path^l j The process of determining the above will be described with reference to FIG.
[0064] The statistics unit 123 of each router function unit 120 calculates statistical information (step S101). j When || is defined as (ii), the statistical unit 123 calculates the amount of data generated for each label, λ i (y) and the average μ i (y) and variance Σ i (y) is calculated as statistical information.
[0065] Next, the transfer unit 122 of each router function unit 120 transfers (sends) the statistical information calculated in the above step S101 to the calculation function unit 110 (step S102).
[0066] The learning unit 132 of each server function unit 130 performs learning when the amount of data in the data set is changed, and obtains a local learning time T j c (D) are calculated (step S103).
[0067] Next, the learning unit 132 of each server function unit 130 learns its own processing capacity pj and each local learning time T j c (D) is transferred (sent) to the calculation function unit 110 (step S104).
[0068] Next, the optimization unit 111 of the calculation function unit 110 calculates the statistical information received from each router i and the processing capacity p j and local learning time T j c (D) and solve the constrained optimization problem shown in Eqs. (7) to (15) in two steps (A) and (B) using an existing solver to obtain the optimal parameter r i [ab] and the optimal model transfer path^l j (Step S105).
[0069] Then, the transmission unit 112 of the calculation function unit 110 transmits the optimal parameter r i [ab] and the optimal model transfer path^l j to each router function unit 120 (step S106).
[0070] <Learning process of deep learning model> The process of learning a deep learning model by federated learning will be described with reference to FIG. 8.
[0071] The model transmission unit 141 of the central server function unit 140 transmits the model parameters of the deep learning model to each server j via the transfer unit 122 of the router function unit 120 in the same device 10 (step S201). As a result, the optimal model transfer path ^l j The model parameters of the deep learning model are transferred to server j by
[0072] The forwarding unit 122 of each router function unit 120 receives the optimal parameter r i In accordance with [ab], the data collected and generated by the collection unit 121 and its label are distributed to the server function unit 130 (step S202). iThe collected and generated data and its label are transferred (transmitted) to the server function unit 130 so as to satisfy [ab].
[0073] Next, the database management unit 131 of each server function unit 130 stores the data and its label received from the router function unit 120 in the database as learning data (step S203). This results in a data set consisting of learning data for the deep learning model.
[0074] Next, the learning unit 132 of each server function unit 130 learns (trains) a deep learning model by federated learning using the data set obtained in step S203 (step S204). As a result, a trained deep learning model is obtained.
[0075] Next, the learning unit 132 of each server function unit 130 transfers (sends) the model parameters of the trained deep learning model obtained in step S204 above to the central server function unit 140 via the transfer unit 122 of the router function unit 120 in the same device 10 (step S205). This creates an optimal model transfer path ^l j By this, the model parameters of the trained deep learning model trained locally on server j are transferred to the central server.
[0076] Then, the model integration unit 142 of the central server function unit 140 synchronizes (integrates) and stores the model parameters received from each server function unit 130 (step S206), thereby completing one cycle of learning in the associative learning.
[0077] <Inference processing using deep learning models> The inference process using the trained deep learning model will be described with reference to FIG. 9.
[0078] The transfer unit 122 of each router function unit 120 transfers (sends) the data collected and generated by the collection unit 121 to the server function unit 130 (step S301).
[0079] The inference unit 133 of each server function unit 130 performs inference with the trained deep learning model using the data received from the router function unit 120 (step S302). This allows, for example, anomaly detection, operational support, etc. to be realized according to the task to which the deep learning model belongs.
[0080] <Summary> As described above, in this embodiment, when a deep learning model is trained by federated learning, a parameter r, which represents the ratio of data flowing from a router i to a link ab in the direction from a to b in the target network 1, is used. i [ab] and the transfer path of the model parameters between the central server and each server j can be optimized. This improves the convergence when learning a deep learning model in each server functional unit 130 by federated learning, thereby minimizing the time required to learn a deep learning model and acquiring a deep learning model with high accuracy.
[0081] The present invention is not limited to the above-described embodiments specifically disclosed, and various modifications, changes, combinations with known technologies, etc. are possible without departing from the scope of the claims.
[0082] [References] Reference 1: Y. Zhao et al., "Federated learning with Non-IID Data" arXiv:1806.00582. Reference 2: AA Adbellatif et al., "Communication-efficient hierarchical federated learning for IoT heterogeneous systems with imbalanced data" Feture Ceneration Computer Systems 128 (2022). [Explanation of symbols]
[0083] 1 Target network 10 equipment 110 Calculation function section 111 Optimization Department 112 Transmitter 120 Router function unit 121 Collection Department 122 Transfer Department 123 Statistics Department 130 Server Function Unit 131 Database Management Department 132 Learning Department 133 Reasoning part 140 Central Server Function Unit 141 Model Transmitter 142 Model Integration Department 201 Input Device 202 Display device 203 External I / F 203a Recording media 204 Communication I / F 205 RAM 206 ROM 207 Auxiliary storage 208 processors 209 Bus
Claims
1. A machine learning system that learns a machine learning model by federated learning, comprising: A first receiving unit that receives statistical information of data generated by each router included in the target network; A second receiving unit that receives a processing capacity of each server included in the target network and a local learning time that represents a time required for each server to learn the machine learning model; an optimization unit that optimizes a parameter representing a ratio of data from the router flowing in a specific direction through links constituting the target network and a transfer route of the model parameters between a central server that manages model parameters of the machine learning model and the server, so as to minimize a ratio of the amount of data held by the server and a sum, for the server, of a product of a data set held by the server and a data set that virtually aggregates all data into one, using the statistical information, the processing capacity, and the local learning time; A machine learning system with
2. The distance between the data sets is An approximation of the Kullback-Leibler divergence for each label between the data distribution of the dataset held by the server and the data distribution when all data is virtually combined into one, under the assumption that the data follows a Gaussian distribution for each label assigned to the data; The statistical information is The machine learning system according to claim 1 , wherein the amount of data generated for each label at the router, and the average and variance of the data for each label at the router.
3. The local learning time is The time T required to train the machine learning model when the amount of data D in the training data set is changed in each server ｊ ｃ (D).
4. The optimization unit is The machine learning system according to any one of claims 1 to 3, wherein the parameters are determined by two-stage optimization under constraints representing a congestion avoidance request in the target network and constraints representing flow rate maintenance in the target network, and then the forwarding path is determined.
5. A machine learning device for learning a machine learning model by federated learning, comprising: A first receiving unit that receives statistical information of data generated by each router included in the target network; A second receiving unit that receives a processing capacity of each server included in the target network and a local learning time that represents a time required for each server to learn the machine learning model; an optimization unit that optimizes a parameter representing a ratio of data from the router flowing in a specific direction through links constituting the target network and a transfer route of the model parameters between a central server that manages model parameters of the machine learning model and the server, so as to minimize a ratio of the amount of data held by the server and a sum, for the server, of a product of a data set held by the server and a data set that virtually aggregates all data into one, using the statistical information, the processing capacity, and the local learning time; A machine learning device having the above configuration.
6. A machine learning device for learning a machine learning model by federated learning, A first receiving step of receiving statistical information of data generated by each router included in the target network; A second receiving step of receiving a processing capacity of each server included in the target network and a local learning time representing a time required for each server to learn the machine learning model; an optimization procedure for optimizing a parameter representing a ratio of data from the router flowing in a specific direction through links constituting the target network and a transfer route of the model parameters between a central server that manages model parameters of the machine learning model and the server, so as to minimize a ratio of the amount of data held by the server and a sum, for the server, of a product of a data set held by the server and a data set that virtually aggregates all data into one, using the statistical information, the processing capacity, and the local learning time; A machine learning method to perform.
7. A machine learning device for training a machine learning model by federated learning, A first receiving step of receiving statistical information of data generated by each router included in the target network; A second receiving step of receiving a processing capacity of each server included in the target network and a local learning time representing a time required for each server to learn the machine learning model; an optimization procedure for optimizing a parameter representing a ratio of data from the router flowing in a specific direction through links constituting the target network and a transfer route of the model parameters between a central server that manages model parameters of the machine learning model and the server, so as to minimize a ratio of the amount of data held by the server and a sum, for the server, of a product of a data set held by the server and a data set that virtually aggregates all data into one, using the statistical information, the processing capacity, and the local learning time; A program that executes the following.